Scolaris Content Display Scolaris Content Display

区分胰腺局灶性病变的成像模式

Contraer todo Desplegar todo

摘要

研究背景

每年都检测到了越来越多的胰腺偶发病变。能否准确地将胰腺病变分为良性、癌前病变和癌性病变是决定是否进行治疗或监测的关键。区分良性病变与癌前病变和癌性病变可以防止患者进行不必要的大手术。尽管准确区分胰腺病变很重要,但尚无明确的胰腺局灶性病变管理流程。

研究目的

本系统综述旨在确定和比较各种成像模式在检测胰腺局灶性病变患者的癌性和癌前病变方面的诊断准确性。

检索策略

我们检索了CENTRAL、MEDLINE、Embase和科学引文索引(Science Citation Index),截至2016年7月19日。我们检索了纳入研究的参考文献以确定进一步的研究。检索时,我们没有限制语言或出版情况,以及资料是前瞻性收集的还是回顾性收集的。

纳入排除标准

我们计划纳入报告了胰腺病变患者指标测试横断面信息((计算机断层扫描(computed tomography,CT)、磁共振成像(magnetic resonance imaging,MRI)、正电子发射断层扫描(positron emission tomography,PET)、内镜超声(endoscopic ultrasound,EUS)、EUS弹性成像和EUS引导下活检或细针穿刺抽吸(fine‐needle aspiration,FNA))和参考标准(通过手术切除对整个病灶进行组织病理学检查确认病灶性质,或对指标测试阴性的人活检以进行癌前病变或癌变的组织病理学检查确认并至少进行六个月的临床随访)的研究,不限制语言或出版状态,也不限制资料是前瞻性还是回顾性收集的。

资料收集与分析

两位综述作者独立检索参考文献以确定相关研究并提取资料。我们计划使用双变量分析来计算95%置信区间的总敏感性和特异性以及分层总受试者工作特征(hierarchical summary receiver operating characteristic,HSROC)来比较测试并评价异质性,但因为数据稀疏,在适当的时候使用更简单的模型(例如单变量随机效应模型和单变量固定效应模型)合并研究。由于数据稀疏,我们无法使用正规的统计方法比较测试的诊断性能。

主要结果

我们纳入了54项研究,共涉及3196名研究对象,评价了各种指标测试的诊断准确性。在这54项研究中,确定了八种不同的目标疾病,最终诊断为良性、癌前病变和癌性病变。没有研究具有高方法学质量。纳入的单一研究中没有一个比较组具有足够高的方法学质量,不能提供重要结果。为了区分癌性病变与良性或癌前病变,我们在每项指标测试中仅确定一项研究。第二项分析,对象是区分癌性病变与良性病变的研究,提供了可以进行meta分析的三项测试。诊断癌症的敏感性和特异性是:EUS‐FNA:敏感性0.79(95% 置信区间(confidence interval,CI) [0.07,1.00]),特异性1.00(95% CI [0.91,1.00]);EUS:敏感性0.95(95% CI [0.84,0.99]),特异性0.53(95% CI [0.31,0.74]);PET:敏感性0.92(95% CI [0.80,0.97]),特异性0.65(95% CI [0.39,0.84])。第三项分析,对象是区分癌前病变或癌性病变与良性病变的研究,仅提供了一项测试(EUS‐FNA),并进行了meta分析。EUS‐FNA对诊断癌前病变或癌性病变具有中等敏感性(敏感性0.73(95% CI [0.01,1.00])和高特异性0.94(95% CI [0.15,1.00]),极宽的置信区间反映研究之间有异质性)。第四项分析,对象是区分癌性(浸润性癌)和癌前病变(异型增生)的研究,提供了三项测试并进行了meta分析。诊断浸润性癌的敏感性和特异性是:CT:敏感性0.72(95% CI [0.50,0.87]),特异性0.92(95% CI [0.81,0.97]);EUS:敏感性0.78(95% CI [0.44,0.94]),特异性0.91(95% CI [0.61,0.98]);EUS‐FNA:敏感性0.66(95% CI [0.03, 0.99]),特异性0.92(95% CI [0.73,0.98])。第五项分析,对象是区分癌性(高度异型增生或浸润性癌)与癌前病变(低度或中度异型增生)的研究,提供了六项测试并进行了meta分析。诊断癌症(高度异型增生或浸润性癌)的敏感性和特异性为:CT:敏感性0.87(95% CI [0.00,1.00]),特异性0.96(95% CI [0.00,1.00]);EUS:敏感性0.86(95% CI [0.74,0.92]),特异性0.91(95% CI [0.83,0.96]);EUS‐FNA:敏感性0.47(95% CI [0.24,0.70]),特异性0.91(95% CI [0.32,1.00]);EUS‐FNA癌胚抗原200ng/mL:敏感性0.58(95% CI [0.28,0.83]),特异性0.51(95% CI [0.19,0.81]);MRI:敏感性0.69(95% CI [0.44,0.86]),特异性0.93(95% CI [0.43,1.00]);PET:敏感性0.90(95% CI [0.79, 0.96]),特异性0.94(95% CI [0.81,0.99])。第六项分析,对象是区分癌性(浸润性癌)和癌前病变(低度异型增生)的研究,没有提供可进行meta分析的测试。第七项分析,对象是区分癌前病变或癌性病变(中度或高度异型增生或浸润性癌)与癌前病变(低度异型增生)的研究,提供了两项测试并进行了meta分析。诊断癌症的敏感性和特异性为:CT:敏感性0.83(95% CI [0.68,0.92]),特异性0.83(95% CI [0.64,0.93])和MRI:敏感性0.80(95% CI [0.58,0.92]),特异性0.81(95% CI [0.53,0.95])。第八项分析,对象是区分癌前病变或癌性病变(中度或高度异型增生或浸润性癌)与癌前病变(低度异型增生)或良性病变的研究,没有提供可进行meta分析的测试。

胰腺囊性局灶性病变的亚组分析没有重大变化(42项研究;2086名研究对象)。纳入的研究均未评价EUS弹性成像和序贯测试。

作者结论

由于研究作者将胰腺局灶性病变分为癌性、癌前病变和良性病变的方式不同;每项比较中纳入的研究少,置信区间宽;研究的方法学质量差以及比较组中估计值的异质性,我们无法得出任何确定的结论。

简语概要

确定胰腺肿瘤是否癌变的不同成像技术的准确性

系统综述背景

胰腺是腹部的一个器官,它分泌助消化的胰液,并控制产生重要激素(如胰岛素)的细胞。越来越多接受常规扫描(例如超声波扫描或计算机断层扫描(computed tomography,CT))的人发现了胰腺异常,以所谓的“阴影”形式呈现,可描述为胰腺局灶性病变、胰腺肿块、胰腺肿瘤、胰腺囊肿或胰腺结节。很大一部分胰腺局灶性病变是不需要治疗的良性(非癌性)病变。手术切除肿瘤是治疗癌前病变(即胰腺局灶性病变不是完全成熟的癌,没有像癌一样具有扩散的能力,但可以变成癌)和癌性胰腺局灶性病变的主要方法。正在开发治疗癌前病变的新方法,例如使用“热”来破坏肿瘤。对于局灶性胰腺癌患者,手术切除仍然是唯一的可能治愈的方法。因此,重要的是要确定胰腺局灶性病变是非癌性病变、癌前病变还是癌性病变。很多扫描都可用于区分胰腺局灶性病变的性质,如下所示。

•计算机断层扫描(Computed tomography,CT):从不同角度收集的一系列X射线,然后使用计算机重建。

•磁共振成像(Magnetic resonance imaging,MRI):使用强磁力的磁体产生身体不同组织的图像。

•正电子发射断层扫描(Positron emission tomography,PET):使用少量放射性葡萄糖(糖)来区分不同的组织。这利用了癌细胞比正常细胞消耗更多葡萄糖的特性。

•内镜超声(也称为超声内镜或(endoscopic ultrasound, EUS)):使用内窥镜,将照相机放入体腔以观察身体内部。内窥镜末端的超声波(高能声波)探头用于区分组织。

•EUS弹性成像:测量病灶的硬度,用于确定病变是否癌变。

•EUS引导下活检:取出细胞或组织在显微镜下检查或对细胞或组织进行其他测试。

目前尚不清楚不同扫描方法在区分胰腺局灶性病变方面的有效性。

研究特征

我们对报告不同扫描准确性的研究进行了彻底的文献检索,截至2016年7月19日。我们确定了54项研究共报告了3196名胰腺局灶性病变患者的信息。这些研究评价了一项或多项上述测试,并将这些测试结果与通过手术切除病灶并在显微镜下检查提供的最终诊断进行了比较。没有关注EUS弹性成像的诊断测试准确性研究,也没有关注多项扫描而不是单项扫描的研究。

主要结果

研究在定义癌前病变和癌性病变上存在的差异意味着我们无法合并资料来提供这些测试的总体结果。由于以下原因,我们无法得出任何确定的结论。

•研究作者将胰腺局灶性病变分为癌性、癌前病变和良性病变的方式在不同研究中并不一致。

•这些研究纳入的研究对象很少,导致结果存在很大的不确定性。

•这些研究的方法学质量差,给结果带来了额外的不确定性。

•即使在以相似方式将局灶性胰腺病变区分为癌性、癌前病变和良性病变的研究中,结果也不一致。

证据质量

所有研究的方法学质量都低,这可能会导致得出错误的结论。

Authors' conclusions

Implications for practice

We were unable to arrive at any firm conclusions because of the differences in the way that study authors classified focal pancreatic lesions into cancerous, precancerous, and benign lesions; the inclusion of few studies with wide confidence intervals for each comparison; poor methodological quality in the studies; and heterogeneity in the estimates within comparisons.

Implications for research

Further studies of high methodological quality are necessary. Future research should be conducted in a prospective manner, however most importantly the definition of benign and cancerous lesions in the analysis of studies should be standardised according to World Health Organization (WHO) classification. The threshold for positivity of endoscopic ultrasound‐guided fine‐needle aspiration cancer markers should be prespecified. Future studies should avoid any inappropriate exclusions to ensure that true diagnostic accuracy can be determined. Long‐term follow‐up of participants with negative tests will help in understanding the implications of false‐negative results and will aid clinical decision‐making.

Summary of findings

Open in table viewer
Summary of findings Imaging modalities for characterising focal pancreatic lesions

Name of test

Number of studies (number of participants)

Sensitivity (95% CI)

Specificity (95% CI)

Post‐test probability of positive test* (95% CI)

Post‐test probability of negative test* (95% CI)

Number of false positives per 100 positive index test results (95% CI)

Number of false negatives per 100 negative index test results (95% CI)

Risk of bias

Applicability concerns

Uncertainty (due to inconsistency or inability to assess inconsistency, and random errors because of overall small sample size)

Cancerous versus benign or precancerous (median pre‐test probability: 63%)

EUS‐FNA (cytology)

1 (45)

0.79 (0.60 to 0.91)

1.00 (0.85 to 1.00)

98% (79% to 100%)

26% (14% to 43%)

2 (0 to 21)

26 (14 to 43)

Unclear

High

High

EUS‐FNA (CEA > 500 ng/mL)

1 (24)

0.93 (0.70 to 0.99)

0.33 (0.12 to 0.65)

70% (59% to 79%)

25% (4% to 73%)

30 (21 to 41)

25 (4 to 73)

High

High

High

PET (criteria unspecified)

1 (76)

0.85 (0.73 to 0.92)

0.91 (0.72 to 0.97)

94% (81% to 98%)

21% (12% to 34%)

6 (2 to 19)

21 (12 to 34)

Unclear

High

High

Cancerous versus benign (median pre‐test probability: 70%)

EUS

2 (133)

0.95 (0.84 to 0.99)

0.53 (0.31 to 0.74)

82% (74% to 88%)

18% (6% to 45%)

18 (12 to 26)

18 (6 to 45)

Unclear or high

High

High

EUS‐FNA (cytology)

3 (147)

0.79 (0.07 to 1.00)

1.00 (0.91 to 1.00)

99% (90% to 100%)

32% (2% to 92%)

0 (0 to 9)

32 (2 to 92)

High

High

High

PET (criteria unspecified)

3 (99)

0.92 (0.80 to 0.97)

0.65 (0.39 to 0.85)

86% (75% to 92%)

22% (9% to 44%)

14 (8 to 25)

22 (9 to 44)

High

High

High

PET (SUVmax > 3.5)

1 (80)

0.96 (0.87 to 0.99)

0.62 (0.43 to 0.78)

85% (78% to 90%)

12% (3% to 36%)

15 (10 to 22)

12 (3 to 36)

High

High

High

CT

2 (123)

0.98 (0.00 to 1.00)

0.76 (0.02 to 1.00)

90% (17% to 100%)

6% (0% to 100%)

10 (0 to 83)

6 (0 to 100)

Unclear or high

High

High

MRI

1 (29)

0.80 (0.58 to 0.92)

0.89 (0.57 to 0.98)

94% (72% to 99%)

34% (17% to 56%)

6 (1 to 28)

34 (17 to 56)

High

High

High

Precancerous or cancerous versus benign (median pre‐test probability: 71%)

EUS

1 (34)

0.92 (0.74 to 0.98)

0.60 (0.31 to 0.83)

85% (72% to 92%)

25% (7% to 58%)

15 (8 to 28)

25 (7 to 58)

High

High

High

EUS‐FNA (cytology)

2 (52)

0.73 (0.01 to 1.00)

0.94 (0.15 to 1.00)

97% (25% to 100%)

41% (1% to 98%)

3 (0 to 75)

41 (1 to 98)

Unclear or high

High

High

EUS‐FNA (CEA > 50 ng/mL)

1 (11)

0.29 (0.08 to 0.64)

0.25 (0.05 to 0.70)

48% (20% to 77%)

87% (54% to 98%)

52 (23 to 80)

87 (54 to 98)

High

High

High

PET (SUVmax 2.4)

1 (32)

0.94 (0.74 to 0.99)

0.93 (0.69 to 0.99)

97% (83% to 100%)

13% (2% to 49%)

3 (0 to 17)

13 (2 to 49)

High

High

High

CT

1 (48)

0.62 (0.45 to 0.76)

0.64 (0.39 to 0.84)

81% (66% to 90%)

59% (44% to 72%)

19 (10 to 34)

59 (44 to 72)

Unclear

High

High

MRI

1 (27)

0.93 (0.69 to 0.99)

0.85 (0.58 to 0.96)

94% (80% to 98%)

17% (3% to 58%)

6 (2 to 20)

17 (3 to 58)

High

High

High

Cancerous (invasive carcinoma) versus precancerous (dysplasia) (median pre‐test probability: 27%)

EUS

5 (156)

0.78 (0.45 to 0.94)

0.91 (0.61 to 0.98)

75% (37% to 94%)

8% (3% to 22%)

25 (6 to 63)

8 (3 to 22)

Unclear or high

High

High

EUS‐FNA (cytology)

3 (158)

0.66 (0.03 to 0.99)

0.92 (0.73 to 0.98)

75% (29% to 95%)

12% (1% to 69%)

25 (5 to 71)

12 (1 to 69)

Unclear or high

High

High

EUS‐FNA (CEA > 200 ng/mL)

1 (41)

1.00 (0.57 to 1.00)

0.64 (0.48 to 0.78)

51% (40% to 61%)

Not estimable

49 (39 to 60)

Not estimable

High

High

High

CT

6 (326)

0.72 (0.50 to 0.87)

0.92 (0.81 to 0.97)

78% (57% to 91%)

10% (5% to 18%)

22 (9 to 43)

10 (5 to 18)

Unclear or high

High

High

MRI

1 (32)

0.75 (0.30 to 0.95)

0.93 (0.77 to 0.98)

80% (48% to 94%)

9% (2% to 35%)

20 (6 to 52)

9 (2 to 35)

High

High

High

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) (median pre‐test probability: 45%)

EUS

4 (196)

0.86 (0.74 to 0.92)

0.91 (0.83 to 0.96)

89% (80% to 94%)

11% (7% to 19%)

11 (6 to 20)

11 (7 to 19)

High

High

High

EUS‐FNA (cytology)

3 (310)

0.47 (0.24 to 0.70)

0.91 (0.32 to 1.00)

81% (19% to 99%)

32% (22% to 45%)

19 (1 to 81)

32 (22 to 45)

Unclear or high

High

High

EUS‐FNA (CEA > 200 ng/mL)

3 (160)

0.58 (0.28 to 0.83)

0.51 (0.19 to 0.81)

49% (28% to 70%)

40% (19% to 65%)

51 (30 to 72)

40 (19 to 65)

High

High

High

EUS‐FNA (CA 19‐9 > 1000 U/mL)

1 (41)

0.90 (0.60 to 0.98)

0.42 (0.26 to 0.59)

56% (47% to 65%)

16% (3% to 57%)

44 (35 to 53)

16 (3 to 57)

High

High

High

EUS‐FNA (CEA > 692.8 ng/mL)

1 (20)

0.80 (0.49 to 0.94)

0.90 (0.60 to 0.98)

87% (50% to 98%)

15% (5% to 39%)

13 (2 to 50)

15 (5 to 39)

Unclear

High

High

PET (SUVmax > 2 to 2.5)

4 (124)

0.90 (0.79 to 0.96)

0.94 (0.81 to 0.99)

93% (78% to 98%)

8% (4% to 16%)

7 (2 to 22)

8 (4 to 16)

High

High

High

CT

3 (139)

0.87 (0.00 to 1.00)

0.96 (0.00 to 1.00)

95% (0% to 100%)

10% (0% to 100%)

5 (0 to 100)

10 (0 to 100)

Unclear or high

High

High

MRI

3 (189)

0.69 (0.44 to 0.86)

0.93 (0.43 to 1.00)

89% (35% to 99%)

21% (12% to 36%)

11 (1 to 65)

21 (12 to 36)

High

High

High

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) (median pre‐test probability: 21%)

EUS

1 (51)

0.77 (0.50 to 0.92)

0.89 (0.76 to 0.96)

67% (43% to 84%)

7% (3% to 16%)

33 (16 to 57)

7 (3 to 16)

Unclear

High

High

CT

1 (46)

0.50 (0.22 to 0.78)

0.95 (0.83 to 0.99)

72% (36% to 92%)

13% (7% to 22%)

28 (8 to 64)

13 (7 to 22)

High

High

High

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) (median pre‐test probability: 59%)

CT

3 (106)

0.83 (0.68 to 0.92)

0.83 (0.64 to 0.93)

89% (56% to 98%)

33% (18% to 52%)

11 (2 to 44)

33 (18 to 52)

High

High

High

MRI

2 (71)

0.80 (0.58 to 0.92)

0.81 (0.53 to 0.95)

86% (67% to 95%)

27% (13% to 47%)

14 (5 to 33)

27 (13 to 47)

High

High

High

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign (median pre‐test probability: 43%)

EUS

1 (70)

0.97 (0.83 to 0.99)

0.40 (0.26 to 0.55)

55% (48% to 61%)

6% (1% to 31%)

45 (39 to 52)

6 (1 to 31)

High

High

High

*Post‐test probability was calculated at the median pre‐test probability.

Abbreviations:

CA 19‐9: carbohydrate antigen 19‐9
CEA: carcinoembryonic antigen
CI: confidence interval
CT: computed tomography
EUS: endoscopic ultrasound
FNA: fine‐needle aspiration
MRI: magnetic resonance imaging
PET: positron emission tomography
SUVmax: maximum standardised uptake values

Background

(Please see the glossary in Appendix 1 for terms that have not been described in the main text.)

A 'shadow’ identified in the pancreas on imaging may be variously described as a focal pancreatic lesion, pancreatic mass, pancreatic tumour, pancreatic cyst, or pancreatic nodule. This phrasing refers to focal lesions, as opposed to diffuse changes of the pancreas, and includes solid and cystic lesions of the pancreas. In the Western world, the prevalence of focal pancreatic lesions is approximately 1.2% and is increasing steadily (by approximately 8%) each year, with smaller and asymptomatic lesions being identified more frequently (Gaujoux 2011; Spinelli 2004). An incidental pancreatic lesion is one that is detected in the pancreas of a patient who undergoes radiological investigations for an unrelated medical condition (Sachs 2009). Such asymptomatic incidental lesions represent 55% to 60% of pancreatic tumours (Gaujoux 2011; Spinelli 2004). Some focal pancreatic lesions may be associated with symptoms, depending upon their size and nature. The symptoms of pancreatic cancer, which generally refers to pancreatic adenocarcinoma, can include obstructive jaundice (yellowish discolouration of the skin and the whites of the eyes with dark urine and pale stool due to blockage of bile duct (National Cancer Institute 2011a), a tube that transports the bile from the liver), loss of appetite, and abdominal pain (Holly 2004). The symptoms of pancreatic neuroendocrine tumours (tumours arising from cells that secrete hormones), some of which may be cancer, are related to the excessive secretion of hormones (by the tumour) such as insulin, glucagon, gastrin, somatostatin, and vasoactive peptide resulting in hypoglycaemia (decreased blood sugar), hyperglycaemia (increased blood sugar, a rare cause of diabetes), and gastrointestinal disturbances such as peptic ulcer and diarrhoea (Batcher 2011). The symptoms of chronic pancreatitis (chronic inflammation of the pancreas that can result in alteration in the structure and function of the pancreas) are abdominal and back pain and those symptoms related to pancreatic insufficiency, which include steatorrhoea, malabsorption, vitamin deficiency, diabetes, or weight loss (Braganza 2011; Nair 2007). About 40% of people with focal pancreatic lesions have chronic pancreatitis (Spinelli 2004). In the remaining 60% of people with focal pancreatic lesions, the remaining pancreas is normal.

Focal pancreatic lesions can be benign (serous pancreatic cystadenoma, acinar cell cystadenoma, papillary cysts, lymphoepithelial cysts, simple cysts), precancerous (intraductal papillary mucinous neoplasm (IPMN) with dysplasia but without invasive cancer, mucinous cystic neoplasm (MCN), benign neuroendocrine tumours), or cancer (ductal adenocarcinoma, acinar cell carcinoma, IPMN with invasive carcinoma, cystadenocarcinoma, pancreatoblastoma, solid pseudo‐papillary neoplasm, cancer neuroendocrine tumours) (Luttges 2011; Sachs 2009; Spinelli 2004; WHO 2016). Dysplasia can be low grade, intermediate grade, or high grade (WHO 2016). About 80% of benign lesions, 50% of precancerous lesions, and 20% of cancerous lesions are asymptomatic (Spinelli 2004). Focal pancreatic lesions can be solid or cystic or mixed solid and cystic tumours (Cho 2011).

Surgical resection is generally considered to be the only curative treatment for pancreatic cancer. Worldwide, only 15% to 20% of people with pancreatic cancer undergo potentially curative resection (Conlon 1996; Engelken 2003; Katz 2009; Michelassi 1989; Shahrudin 1997; Smith 2008). In the remaining patients, the cancers are not resected because of infiltration of local structures or disseminated disease. Early diagnosis of pancreatic cancer might enable resection of the pancreatic cancer before it is too late to resect. Pancreatic resection is a major surgery, with an approximately 1% to 25% risk of perioperative death reported worldwide (Conlon 1996; Katz 2009; Michelassi 1989; Shahrudin 1997; van Oost 2006). High‐volume centres show a lower perioperative mortality of less than 5% compared to low‐volume centres, which are associated with a perioperative mortality of up to 25% (Gurusamy 2013; van Oost 2006). Pancreatic resection is also associated with an about 40% morbidity rate (Gurusamy 2013; van der Gaag 2010). Only 5% to 25% of patients survive for five years (Conlon 1996; Katz 2009; Michelassi 1989; Shahrudin 1997). Surgery is generally offered if there are features suggestive of precancerous or cancerous lesions (Lee 2005c), although some clinicians prefer sequential follow‐up (by imaging) of precancerous lesions to surgical resection (Irie 2004). Surgery is offered when there is an increase in the size or morphology (the way the lesion appears) of the lesion in sequential imaging (Gaujoux 2011). Surgery is also offered when there is considerable uncertainty as to the nature of the lesion. In some ways, surgery can be considered as a diagnostic test for characterisation of the lesion and as a treatment for people with cancerous and precancerous lesions. Histological confirmation of the lesion by percutaneous biopsy is generally not performed because of difficulty in accessing the lesion percutaneously and because of dissemination of cancer cells.

Target condition being diagnosed

  1. Cancerous versus benign or precancerous lesions.

  2. Precancerous or cancerous (including the type of cancerous lesion) versus benign lesions.

Index test(s)

Computed tomography (CT) scan

This involves a series of X‐rays taken from different angles, which are then reconstructed using a computer (National Cancer Institute 2011a). Morphological features of the lesion, such as density, regularity of margins, vascularity, and the diameter of the pancreatic duct, are taken into account to characterise the lesion. The main side effect of CT scan is the ionising radiation (radioactivity) associated with it. Everyone is exposed to very small amounts of radiation (background radiation). One CT scan of the abdomen is equivalent to approximately three years of background radiation (Fred 2004). In addition, the contrast material (dye used to view the structures better) can cause allergic reactions, such as difficulty breathing, or kidney damage, particularly in people with pre‐existing kidney disease (Namasivayam 2006).

Magnetic resonance imaging (MRI)

This involves the use of a powerful magnet to produce images of different tissues of the body. Magnetic resonance imaging is also known as nuclear magnetic resonance imaging (NMRI) (National Cancer Institute 2011b). Similar features as those employed in CT scan are used to characterise the lesion. While MRI does not use radiation, it is contraindicated in people with metallic implants such as artificial joints, those with cardiac pacemakers (devices used to control heart rhythm), and those with claustrophobia (fear of closed spaces) (Dill 2008). Some of the contrasts used can also cause kidney damage (Dill 2008).

Positron emission tomography (PET)

This involves the use of a small amount of radioactive glucose (sugar) to differentiate between different tissues. It takes advantage of the tendency of cancer cells to use more glucose than normal cells. Positron emission tomography is also known as PET scan (National Cancer Institute 2011c). Cancerous lesions appear as areas of increased uptake. Positron emission tomography also uses ionising radiation (Leide‐Svegborn 2010). The radiation exposure to one PET scan is similar to that in one CT scan of abdomen (Fred 2004; Leide‐Svegborn 2010).

Endoscopic ultrasound (EUS)

This involves the use of an endoscope, a camera introduced into a body cavity to view the inside of the body. An ultrasound (high‐energy sound waves) probe at the end of the endoscope is used to differentiate tissues. Endoscopic ultrasound is also known as endosonography (National Cancer Institute 2011d). Features such as echogenicity and regularity of margins are taken into account and used to characterise the lesion. Complications following EUS are rare and include perforation (Benson 2010; Niv 2011).

EUS elastography

This measures the stiffness of the lesion, which can be used to identify whether the lesion is benign or cancerous (Iglesias‐Garcia 2010). The complications associated with EUS elastography are the same as with EUS.

EUS‐guided biopsy

This is the removal of cells or tissues for examination by a pathologist. The pathologist may study the tissue under a microscope or perform other tests on the cells or tissue. There are many different types of biopsy procedures. The most common types include:

  1. incisional biopsy, in which only a sample of tissue is removed;

  2. excisional biopsy, in which an entire lump or suspicious area is removed; and

  3. needle biopsy, in which a sample of tissue or fluid is removed with a needle. When a wide needle is used the procedure is called a core biopsy. When a thin needle is used the procedure is called a fine‐needle aspiration biopsy (FNAB) (National Cancer Institute 2011e).

Because of the risk of dissemination from cancer, EUS‐guided biopsy is preferable to percutaneous (image‐guided) biopsy (Micames 2003). The examinations under the microscope used may include the routine haemotoxylin and eosin stain for core biopsy and special staining for FNAB (Mehta 2010). Immunocytochemistry and proteomic profiling to identify the presence of biomarkers in the tissue may also be used in the diagnosis (Cui 2009; Mehta 2010). A positive core biopsy can confirm cancer, but a negative core biopsy cannot rule out cancer. Cytology results are not quite as reliable as core biopsy as false‐positive cytology has been reported (Hancke 1984).

Complications associated with EUS‐guided biopsy include those associated with EUS as well as bleeding (Benson 2010; Niv 2011).

Of these index tests, the commonly available tests are CT scan and MRI. The remaining tests (PET, EUS, EUS elastography, and EUS‐guided biopsy) are available in major tertiary centres only.

Clinical pathway

There is no standard algorithm in the diagnosis or management of focal pancreatic lesions. The algorithm may vary from one centre to another and even within the same centre (Gaujoux 2011; Goh 2006b). One possible diagnostic clinical pathway is shown in Figure 1. As noted in Figure 1, an increase in the size of or change of morphological features is one of the reasons that surgeons recommend surgical excision, as this may indicate that the lesion was malignant in the first instance (without features suggestive of malignancy in the original scan) or has transformed into a malignant lesion. The interval for sequential scans is variable. Our local protocol advises sequential scanning in one year in the absence of malignant features. It is important to distinguish whether the focal pancreatic lesion is benign with no cancer potential so that unnecessary surgery and anxiety can be avoided. It is also important to know whether the lesion is precancerous or cancerous so that an informed decision about surgery can be made after weighing the potential benefits and harms. In addition, new alternative treatments such as radiofrequency ablation (destruction of tissue using radiofrequency waves) are being evaluated for precancerous lesions (Pereira 2015). It is also necessary to differentiate the different types of cancer, since different malignancies carry different prognoses (Klempnauer 1995). Some surgeons follow the single‐test strategy, that is making decisions based on the features of the lesion in a single test, while others follow repeated testing (repeating the imaging modality or using a different imaging modality), particularly if the nature of the lesion is indeterminate. The optimal interval between the tests in the repeated‐testing strategy is unknown.


Clinical pathway.Abbreviations:Ca 19‐9: carbohydrate antigen 19‐9
 CT: computed tomography
 EUS: endoscopic ultrasound
 MRI: magnetic resonance imaging
 PET: positron emission tomography

Clinical pathway.

Abbreviations:

Ca 19‐9: carbohydrate antigen 19‐9
CT: computed tomography
EUS: endoscopic ultrasound
MRI: magnetic resonance imaging
PET: positron emission tomography

Prior test(s)

The tests that occur prior to pancreatic imaging depend on how the patient presents. The investigation may be targeted if the patient presents with abdominal symptoms, however it is equally possible that the pancreatic lesion is an incidental finding on an abdominal scan for an alternative reason. As pancreatic cancer is relatively late presenting (Porta 2005), the number of incidental lesions found will be high comparative to other cancers where symptoms will primarily drive discovery. Whilst CT, MRI, and PET may identify incidental lesions, EUS and EUS‐guided fine‐needle aspiration (EUS‐FNA) are the likely second test for known lesions of symptomatic individuals.

Role of index test(s)

All of the index tests described are used primarily to characterise pancreatic lesions as either benign or cancerous, or more importantly as needing significant or more conservative treatment. The location of the pancreas makes percutaneous biopsy dangerous because of the risk of cancer spread, therefore determination of cancer stage and consequently required treatment must be made non‐invasively by the imaging techniques and by EUS‐FNA.

Alternative test(s)

Computed tomography is usually part of a standard algorithm for assessing focal pancreatic lesions (Gaujoux 2011). If the incidental lesion is detected on CT scan, then CT scan can be the only investigation, since the added value of the other tests is not known. One or more of the above tests may be used in addition to, or instead of, CT scan. Diagnostic laparoscopy and laparoscopic ultrasound are other tests that may be used in the differential diagnosis of focal pancreatic lesions; however, these tests are not used routinely. Serum carbohydrate antigen 19‐9 (CA 19‐9) is a substance released into the bloodstream by both cancer cells and normal cells and is used as a type of tumour marker (National Cancer Institute 2011f). Excessive CA 19‐9 in the blood can be a sign of pancreatic cancer or other types of cancer or conditions. The amount of CA 19‐9 in the blood can be used to measure how effective cancer treatments are or if cancer has returned. It can be used in conjunction with other imaging modalities in the assessment of the focal pancreatic lesion.

Rationale

The various imaging modalities use different methods to differentiate normal and diseased tissues. Endoscopic ultrasound is closer to the tissues and therefore high‐frequency ultrasound waves can be used, which have better resolution but poorer penetration than an external ultrasound. Image‐guided biopsy can be performed and the tissue can be examined under the microscope to differentiate between types of focal pancreatic lesion.

Accurate characterisation of lesions will help in patient management. Patients with cancerous lesions will be offered surgery if there is no distant spread of cancer and assuming they are fit for major surgery. Patients with cancerous lesions who are not eligible for surgery because of distant spread of cancer or lack of fitness for major surgery will be offered other treatments such as chemotherapy. Patients with precancerous lesions may also undergo surgery or ablation depending upon the clinician and patient preferences. Unnecessary major surgery can be avoided in patients with benign lesions.

There is currently no Cochrane review of studies assessing the diagnostic accuracy of different imaging modalities in the assessment of focal pancreatic lesions.

Objectives

To determine and compare the diagnostic accuracy of various imaging modalities in detecting cancerous and precancerous lesions in people with focal pancreatic lesions.

Secondary objectives

We planned to explore the following sources of heterogeneity.

  1. Studies at low risk of bias versus those at unclear or high risk of bias (as assessed by the revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS) tool as recommended by the Cochrane Screening and Diagnostic Tests Methods Group) (Whiting 2006). In particular, we considered the studies classified as 'yes' for the items differential verification, un‐interpretable results, and withdrawals as the most important sources of heterogeneity.

  2. Full‐text publications versus abstracts (this might be indicative of publication bias, since there may be an association between the results of the study and the study reaching full publication) (Eloubeidi 2001).

  3. Prospective studies versus retrospective studies.

  4. Different types of reference standard.

  5. Symptomatic versus asymptomatic lesions (the presence of symptoms may increase the pre‐test probability).

  6. Solid versus cystic lesions (as the diagnostic accuracy of the imaging modalities may vary depending upon whether the lesion is solid or cystic).

  7. Participants with chronic pancreatitis versus those without chronic pancreatitis.

  8. Different criteria used by the authors to classify the lesions.

  9. Single imaging versus sequential imaging (repeated imaging).

  10. Different intervals of sequential imaging (e.g. imaging every six months versus annual review).

Methods

Criteria for considering studies for this review

Types of studies

We included studies reporting on cross‐sectional information of the index test and reference test in the appropriate patient population (see below), irrespective of language or publication status or whether the data were collected prospectively or retrospectively. However, we excluded case series in which only true‐positive results or true‐negative results were reported without any information on the other participants who underwent the test.

Participants

Adults with focal pancreatic lesions.

Index tests

CT scan, MRI scan, PET scan, EUS, EUS elastography, and EUS‐guided biopsy either alone or in combination as replacement for major surgery for diagnostic purposes.

We accepted the criteria stated by the authors to classify the lesion as benign, precancerous, and cancerous for different imaging modalities.

There is no standard algorithm in the diagnosis or management of focal pancreatic lesions. Other tests that may be used in the diagnosis of focal pancreatic lesions include diagnostic laparoscopy, laparoscopic ultrasound, serum levels of CA 19‐9, and surgical resection (surgical resection may be considered diagnostic when the diagnosis is uncertain after all other diagnostic modalities have been attempted).

Target conditions

  1. Benign versus precancerous and cancerous lesions (including the type of cancerous lesion).

  2. Benign and precancerous versus cancerous lesions.

Reference standards

We accepted the following reference standards.

  • Histopathological examination of the entire lesion by surgical resection (gold standard). This classified the lesion as benign, precancerous, or cancerous.

  • Histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in people with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging but using appropriate criteria such as metastases or confirmation of cancer by biopsy or death of participants due to cancer) of all participants with negative test for a period of at least six months and for a maximum period of 24 months. Until a definitive diagnosis is available, percutaneous biopsy is generally avoided because of the fear of seeding of cancer cells in potentially resectable cancers. As anticipated, the tissues obtained for histopathological examination were obtained from surgical resection. It is unlikely that patients with a low likelihood for cancer based on clinical symptoms and signs and test results (may include the results of index test) are subject to surgery or biopsy. Even if a biopsy is performed in such patients, a cancerous or precancerous lesion cannot be ruled out because of sampling error. Consequently, such patients are usually followed up clinically with sequential imaging. Pancreatic adenocarcinoma will cause clinical deterioration or increase in tumour size during a period of six months, and so we accepted clinical follow‐up or sequential follow‐up imaging (irrespective of the modality of the imaging) of all participants with a negative biopsy or no biopsy for at least six months as one of the reference standards. However, we accepted clinical follow‐up as a reference standard only when the criteria used for diagnosis were appropriate (e.g. identification of metastases, later biopsy of the lesion confirming the nature of the lesion, and death of participants due to cancer). The choice of a maximum period of 24 months was an arbitrary choice based on the low probability of precancerous lesions becoming cancerous during 24 months. Clinical follow‐up of patients is unlikely to classify precancerous lesions correctly since patients are unlikely to develop metastases or die within this interval.

Search methods for identification of studies

Electronic searches

We searched the following databases.

  1. Cochrane Central Register of Controlled Trials (CENTRAL) in the Cochrane Library (Issue 7, 2016) (Appendix 2).

  2. MEDLINE via PubMed (January 1946 to 19 July 2016) (Appendix 3).

  3. Embase via OvidSP (January 1947 to 19 July 2016) (Appendix 4).

  4. Science Citation Index Expanded via ISI Web of Knowledge (January 1980 to 19 July 2016) (Appendix 5).

Searching other resources

We searched the references of included studies to identify further studies (Horsley 2011). We also searched for additional articles related to the included studies by performing the 'related search' function in MEDLINE (PubMed) and Embase (OvidSP) and 'citing reference' search (search the articles that cited the included articles) in Science Citation Index Expanded and Embase (OvidSP) (Sampson 2008).

Data collection and analysis

Selection of studies

Two review authors independently searched the references to identify relevant studies. We obtained the full text of references that at least one of the review authors consider relevant and used these full texts to further exclude irrelevant references. We selected references to studies that met the inclusion criteria for data extraction. Any differences in study selection were arbitrated by review author BR Davidson.

Data extraction and management

Two review authors independently extracted the following data from each included study.

  • First author of report.

  • Year of publication of report.

  • Study design (prospective or retrospective; cross‐sectional studies or randomised clinical trials).

  • Inclusion and exclusion criteria for individual studies.

  • Total number of participants.

  • Number of females.

  • Mean age of the participants.

  • Criteria used for classification of lesions.

  • Preoperative tests carried out prior to index test.

  • Index test.

  • Reference standard.

  • True positive (TP), false positive (FP), true negative (TN), and false negative (FN) data.

Main analysis

The unit of analysis was the participant. We extracted the TP, FP, TN, and FN information for each index test for the following situations (when data were available).

  1. Precancerous or cancerous lesions (positive test) versus benign lesions with no cancer potential (negative test) (this helps determine whether the patient needs further follow‐up).

  2. Cancerous lesions (positive test) versus non‐cancerous lesions (negative test) (this helps determine whether the patient needs immediate surgery).

  3. In the group of participants with precancerous or cancerous lesions (i.e. those with positive test in the analysis of benign lesions with no cancer potential (negative test) versus precancerous or cancerous lesions (positive test)), we extracted the TP, FP, TN, and FN information for cancerous lesions (positive test) versus precancerous lesions (negative test) (this helps in assessing whether or not surgery is appropriate; surgery is the only curative option for cancerous lesions, while follow‐up may be an option for precancerous lesions).

We extracted the information on indeterminate results separately from the TP, FP, TN, and FN data. There is no standard algorithm of management of patients with indeterminate results in the first scan. Some surgeons may recommend surgical resection for indeterminate lesions, while others may advise additional scans or sequential follow‐up imaging. We therefore planned to perform a sensitivity analysis as described in Sensitivity analyses.

For tests performed for sequential follow‐up imaging (repeated‐testing strategy), we planned to extract the TP, FP, TN, and FN data for the strategy as a whole. We considered increase in size or change in the lesion on sequential follow‐up imaging (performed within 12 months) a positive index test. If the lesion remained static (or decreased in size) without any change in the characteristics of the lesion, we considered this a negative index test. The majority of surgeons will recommend further follow‐up imaging or no follow‐up if the sequential follow‐up image shows no change in the lesion, and there is no clinical deterioration for the comparison between precancerous and cancerous lesions. We therefore considered indeterminate results on sequential follow‐up imaging as negative results for this comparison.

We sought further information from study authors where necessary. Any differences between the review authors were resolved by discussion.

Assessment of methodological quality

Two review authors independently assessed the quality of the studies using the QUADAS‐2 assessment tool (Whiting 2006; Whiting 2011). We resolved any differences in assessment using the QUADAS‐2 assessment algorithm published in the protocol. We sought further information from the authors of the studies in order to accurately assess the methodological quality of the studies.

We assessed the quality items derived from the QUADAS‐2 tool using the methodology stated in Table 1.

Open in table viewer
Table 1. QUADAS‐2 classification

Domain

Signalling question

Signalling question

Signalling question

Risk of bias

Concerns for
applicability

1: Patient sampling

Was a consecutive or random sample of patients enrolled?

Was a case‐control design avoided?

Did the study avoid inappropriate exclusions?

Could the selection of participants have introduced bias?

Are there concerns that the included participants and setting do not match the review question?

Yes: all consecutive patients or random sample of patients with focal pancreatic lesions were enrolled

No: selected patients were enrolled

Unclear: this was not clear from the report

Yes: case‐control design was avoided

No: case‐control design was not avoided

Unclear: this was not clear from the report

Yes: the study avoided inappropriate exclusions (i.e. difficult‐to‐diagnose patients)

No: the study excluded patients inappropriately

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 signalling question

Low concern: the selected participants represent the patients in whom the tests will be used in clinical practice (please see diagnostic pathway (Figure 1))

High concern: there is high concern that participant selection was performed in such a way that the included participants did not represent the patients in whom the tests will be used in clinical practice

2: Index test(s)

Were the index test results interpreted without knowledge of the results of the reference standard?

If a threshold was used, was it prespecified?

Could the conduct or interpretation of the index test have introduced bias?

Are there concerns that the index test, its conduct, or its interpretation differ from the review question?

Yes: index test results were interpreted without knowledge of the results of the reference standard

No: index test results were interpreted with knowledge of the results of the reference standard

Unclear: this was not clear from the report

Yes: if the criteria for a positive test were prespecified

No: if the criteria for a positive test were not prespecified

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 of the 2 signalling questions

High concern: there is high concern that the conduct or interpretation of the index test differs from the way it is likely to be used in clinical practice

Low concern: there is low concern that the conduct or interpretation of the index test differs from the way it is likely to be used in clinical practice

3: Target condition and reference standard(s)

Is the reference standard likely to classify the target condition correctly?

Were the reference standard results interpreted without knowledge of the results of the index tests?

Could the reference standard, its conduct, or its interpretation have introduced bias?

Are there concerns that the target condition as defined by the reference standard does not match the review question?

Yes: histopathological examination of the entire lesion by surgical resection

No: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

Unclear: this was not clear from the report. Such studies will be excluded

Yes: reference standard results were interpreted without knowledge of the results of the index test

No: reference standard results were interpreted with knowledge of the results of the index test

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 of the 2 signalling questions

Low concern: histopathological examination of the entire lesion by surgical resection

High concern: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

4: Flow and timing

Was there an appropriate interval between index test and reference standard?

Did all patients receive the same reference standard?

Were all patients included in the analysis?

Could the patient flow have introduced bias?

Yes:

histopathological examination of the entire lesion (gold standard) ‐ performed within 2 months (chosen arbitrarily).

Histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) performed within 2 months and clinical follow‐up (including sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months

No:

the histopathological examination was performed beyond 2 months of the index tests.

The clinical follow‐up (including sequential follow‐up imaging) was performed less than 6 months after the index test, because some tumours may be slow‐growing

Unclear: this was not clear from the report

Yes: histopathological examination of the entire lesion by surgical resection

No: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

Unclear: this was not clear from the report. Such studies will be excluded

Yes: all patients meeting the selection criteria (selected participants) were included in the analysis, or data on all of the selected participants were available so that a 2 x 2 table including all selected participants could be constructed

No: not all patients meeting the selection criteria were included in the analysis, or the 2 x 2 table could not be constructed using data on all selected participants

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 signalling question

Statistical analysis and data synthesis

We have plotted study estimates of sensitivity and specificity on forest plots and in receiver operating characteristic (ROC) space to explore between‐study variation in the performance of each test. To estimate the summary sensitivity and specificity of each test, we planned to perform the meta‐analysis by fitting the bivariate model (Chu 2006; Reitsma 2005), which accounts for between‐study variability in estimates of sensitivity and specificity through the inclusion of random effects for the logit sensitivity and logit specificity parameters of the bivariate model. As there was lack of convergence due to sparse data, we tried other alternate models suggested by Takwoingi 2015 and colleagues. These included the random‐effects model, ignoring the inverse correlation between sensitivities and specificities in the different studies due to intrinsic threshold effect, and the fixed‐effect model for either sensitivity or specificity or both after visualising the forest plots and summary receiver operating characteristics (SROC) plots (Takwoingi 2015). We based our choice between the different models on the distribution of sensitivities and specificities as noted in the forest plots or ROC space. We also used the model fit as indicated by the ‐2 log likelihood and considered the model with the lower ‐2 log likelihood to be the better model.

We planned to compare the diagnostic accuracy of the tests by including covariate terms for test type (CT scan, MRI, PET, EUS, EUS‐FNA, EUS elastography) in the bivariate model to estimate differences in the sensitivity and specificity of the tests. We planned to allow both the sensitivity and specificity to vary by covariate. In addition, we planned to permit the variances of the random effects and their covariance to also depend on test type, thus allowing the variances to differ between tests. We planned to use likelihood ratio tests to compare the model with and without covariate (test type). We planned to use a P value of less than 0.05 for the likelihood ratio test to indicate differences in the diagnostic accuracy between the tests. If studies that reported different tests in the same study population were available from at least four studies, we planned to perform a direct head‐to‐head comparison by limiting the test comparison to such studies. We planned to calculate the relative sensitivities and specificities for each pair‐wise comparison of tests.

We performed the meta‐analysis using the NLMIXED command in SAS version 9.4 (SAS Institute Inc, Cary, North Carolina, USA) (Takwoingi 2012). The post‐test probabilities were calculated using these pre‐test probabilities and the summary positive and negative likelihood ratios. We calculated the summary likelihood ratios and their confidence intervals from the functions of the parameter estimates from the model that we fitted to estimate the summary sensitivities and specificities. Post‐test probability associated with a positive test is the probability of having the target condition (e.g. precancer or cancer) on the basis of a positive test result (e.g. positive CT) and is the same as the term 'positive predictive value' used in a single diagnostic accuracy study. Post‐test probability associated with a negative test is the probability of having the target condition (e.g. precancer or cancer) on the basis of a negative test result (e.g. negative CT) and is 1 ‐ 'negative predictive value'. 'Negative predictive value' is the term used in a single diagnostic accuracy study to indicate the chance that the participant has no target condition when the test is negative. We have reported the summary sensitivity, specificity, and post‐test probabilities for the median pre‐test probabilities whenever possible.

Investigations of heterogeneity

We visually inspected forest plots of sensitivity and specificity and the ROC curve to identify heterogeneity. We planned to explore heterogeneity by using the different sources of heterogeneity as covariates in the METADAS macro (Takwoingi 2012), but due to the sparseness of the data we were unable to do this. We planned to assess whether there was a statistically significant difference in the likelihood ratios in order to identify heterogeneity. Although we did not formally compare the diagnostic test accuracy of different index tests between solid and cystic lesions, we have presented a subgroup analysis of solid and cystic lesions, since some clinicians consider the diagnostic test accuracies to differ between the two.

Sensitivity analyses

In the presence of indeterminate results (for any reason) for the initial test, we planned to consider two scenarios: the participants with indeterminate results as positive for the test, as some surgeons will recommend surgical resection for indeterminate lesions; and the indeterminate results as negative for the test, as some surgeons will recommend sequential follow‐up imaging. We planned to assess the diagnostic accuracy in both of these scenarios. However, due to sparse data and few studies reporting indeterminate results we did not perform the above.

We also planned to assess the comparative performance of tests by direct comparison (i.e. the tests performed in the same participant) versus indirect comparison (the tests performed in different participants across studies).

Results

Results of the search

We identified 33,795 references through electronic searches of the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, Embase, and Science Citation Index. We were left with 24,799 references after removing duplicate references. We excluded 23,879 clearly irrelevant references through reading the abstracts. We sought the full text for 920 references for further assessment. We did not identify any additional references to studies through other searches. We excluded 866 references for the reasons described in the Characteristics of excluded studies tables. Fifty‐two studies (54 references) met the inclusion criteria. Two studies reported the diagnostic test data on solid and cystic lesions separately (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid)therefore, we considered them as separate studies. We thus included a total of 54 studies in the review (Brand 2000; Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Choi 2003; Correa‐Gallego 2009; de Jong 2012; Doi 2002; Erkan 2012; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Grieser 2010; Harrison 1999; Higashi 1997; Hong 2010; Hu 2013; Jafarimehr 2010; Jang 2014a; Jang 2014b; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kato 1995; Kim 2015; Klau 2011; Kobayashi 2012; Kubo 2001; Kucera 2012; Le Baleur 2011a; Lee 2014; Maire 2008; McHenry 2002; Nakagawa 2009; Nara 2009; Ogawa 2008; Ogawa 2014; Otomi 2014; Pais 2007; Sahani 2006; Saito 2013; Salla 2007; Sedlack 2002; Smith 2016; Takanami 2011; Takeshita 2008; Tan 2009; Taouli 2000; Tomimaru 2010; Yamao 2001; Zhan 2011; Zhan 2013). The reference flow diagram is shown in Figure 2.


Study flow diagram.

Study flow diagram.

Characteristics of included studies

For a summary of the characteristics of included studies see the Characteristics of included studies table.

We included a total of 54 studies involving 31,196 participants in this systematic review. The studies reported investigation of eight different target conditions:

  • cancerous versus benign or precancerous lesions;

  • cancerous versus benign lesions;

  • precancerous or cancerous lesions versus benign lesions;

  • cancerous (invasive carcinoma) versus precancerous (dysplasia) lesions;

  • cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) lesions;

  • cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions;

  • precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions; and

  • precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign lesions.

The variation in target condition was due to different definitions of what constitutes a benign, precancerous, and cancerous lesion. For example, the World Health Organization pancreatic tumour classification system classifies intraductal papillary mucinous neoplasms' (IPMNs) as precancerous tumours regardless of dysplasia (Luttges 2011). However, many of the included studies considered IPMNs to be benign lesions or even classified them as benign or cancerous based on the grade of dysplasia. This meant that the index tests were actually used for differentiating between very different populations of cancerous and benign tumours, and therefore the combination of all studies as simply cancer versus benign would have been inappropriate. In addition, different surgeons will have different thresholds for recommending surgery. Consequently, we have presented the results for all of the various definitions used by authors to classify a lesion as benign, precancerous, or cancerous.

Three studies reported data on tests differentiating cancerous from benign or precancerous lesions. Of these three studies, one reported the performance of EUS‐FNA using cytology (McHenry 2002); another reported the performance of EUS‐FNA using a carcinoembryonic antigen (CEA) threshold of 500 ng/mL (Kalha 2003); and the third reported the performance of PET to differentiate between benign or precancerous and cancerous lesions (Jafarimehr 2010). The median pre‐test probability of a cancerous lesion in these studies was 0.625 or 62.5% (minimum 0.533, maximum 0.711).

Twelve studies reported data on tests differentiating cancerous from benign lesions. Of these 12 studies, two reported the performance of EUS (Brand 2000; Harrison 1999); three reported the performance of EUS‐FNA (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cherian 2010); three reported the performance of PET (Erkan 2012; Higashi 1997; Kato 1995); one reported the performance of PET with a standard uptake value (SUV) maximum of greater than 3.5 as its threshold for positivity (Hu 2013); two reported the performance of CT (Grieser 2010; Harrison 1999); and one reported the performance of MRI to differentiate between cancerous and benign lesions (Klau 2011). The median pre‐test probability of a cancerous lesion in these studies was 0.697 or 69.7% (minimum 0.231, maximum 0.889).

Six studies reported data on tests differentiating precancerous or cancerous from benign lesions, with one study providing data for multiple imaging modalities (Sedlack 2002). One study reported the performance of EUS (Sedlack 2002); three studies reported the performance of EUS‐FNA (Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Sedlack 2002); one study reported the performance of EUS‐FNA using a CEA threshold of 50 ng/mL (Sedlack 2002); one study reported the performance of PET with an SUV maximum threshold of greater than 2.4 as its threshold for positivity (Otomi 2014); one study reported the performance of CT (Fisher 2008); and one study reported the performance of MRI (Jang 2014a). The median pre‐test probability of a precancerous or cancerous lesion in these studies was 0.706 or 70.6% (minimum 0.519, maximum 0.75).

Twelve studies reported data on tests differentiating cancerous invasive carcinomas from precancerous dysplastic lesions, with some studies reporting the diagnostic test accuracy or more than one index test. Five studies reported the performance of EUS (Cellier 1998; de Jong 2012; Nakagawa 2009; Yamao 2001; Zhan 2011); three studies reported the performance of EUS‐FNA (Jin 2015; Pais 2007; Salla 2007); and one study reported the performance of EUS‐FNA using a CEA threshold of 200 ng/mL (Maire 2008). Six studies reported the performance of CT (Cellier 1998; Nakagawa 2009; Nara 2009; Ogawa 2008; Taouli 2000; Yamao 2001), and one study reported the performance of MRI (de Jong 2012). The median pre‐test probability of a cancerous invasive carcinoma was 0.270 or 27% (minimum 0.122, maximum 0.618).

Eighteen studies reported data on tests differentiating cancerous lesions defined by high‐grade dysplasia or invasive carcinoma from precancerous lesions with a low or intermediate grade of dysplasia, with some studies reporting the diagnostic test accuracy or more than one index test. Four studies reported the performance of EUS (Doi 2002; Kobayashi 2012; Lee 2014; Yamao 2001). Three studies reported the performance of EUS‐FNA (Jin 2013a; Smith 2016; Zhan 2013). Three studies reported the performance of EUS‐FNA using a CEA threshold of 200 ng/mL (Correa‐Gallego 2009; Kucera 2012; Maire 2008). One study reported the performance of EUS‐FNA using a carbohydrate antigen 19‐9 threshold of greater than 1000 U/mL (Maire 2008). One study reported the performance of EUS‐FNA using a CEA threshold of 692.8 ng/mL (Zhan 2013). Four studies reported the performance of PET with an SUVmax value between 2 and 2.5 as their threshold for positivity (Hong 2010; Saito 2013; Takanami 2011; Tomimaru 2010). Three studies reported the performance of CT (Hong 2010; Le Baleur 2011a; Yamao 2001). Three studies reported the performance of MRI (Jang 2014b; Kim 2015; Ogawa 2014). The median pre‐test probability of a cancerous lesion defined by high‐grade dysplasia or invasive carcinoma in these studies was 0.449 or 44.9% (minimum 0.167, maximum 0.875).

Two studies reported data on tests differentiating cancerous invasive carcinomas from precancerous lesions with a low grade of dysplasia. One study reported the performance of EUS (Kubo 2001), and one study reported the performance of CT (Takeshita 2008). The median pre‐test probability of cancerous invasive carcinoma in these studies was 0.214 or 21.4% (minimum 0.174, maximum 0.255).

Five studies reported data on tests differentiating precancerous or cancerous lesions that may be moderately or highly dysplastic or invasive carcinomas from precancerous lesions with a low grade of dysplasia. Three studies reported the performance of CT (Ogawa 2008; Sahani 2006; Tan 2009), and two studies reported the performance of MRI (Choi 2003; Sahani 2006). None of the studies reported the diagnostic accuracy of EUS elastography or sequential testing. The median pre‐test probability of a cancerous lesion that may be moderately or highly dysplastic or an invasive carcinoma in these studies was 0.593 or 59.3% (minimum 0.574, maximum 0.68).

One study reported data on tests differentiating precancerous or cancerous lesions that may be moderately or highly dysplastic or invasive carcinomas from benign or precancerous lesions with a low grade of dysplasia. This study reported the performance of EUS. The median pre‐test probability of a cancerous lesion that may be moderately or highly dysplastic or an invasive carcinoma in this study was 0.429 or 42.9%.

Forty‐six studies were full‐text publications (Brand 2000; Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Cherian 2010; Choi 2003; Correa‐Gallego 2009; de Jong 2012; Doi 2002; Fisher 2008; Grieser 2010; Harrison 1999; Higashi 1997; Hong 2010; Hu 2013; Jang 2014a; Jang 2014b; Jin 2015; Kamata 2016a; Kato 1995; Kim 2015; Klau 2011; Kobayashi 2012; Kubo 2001; Kucera 2012; Le Baleur 2011a; Lee 2014; Maire 2008; Nakagawa 2009; Nara 2009; Ogawa 2008; Ogawa 2014; Otomi 2014; Pais 2007; Sahani 2006; Saito 2013; Salla 2007; Sedlack 2002; Smith 2016; Takanami 2011; Takeshita 2008; Tan 2009; Taouli 2000; Tomimaru 2010; Yamao 2001; Zhan 2013). The remaining studies were abstracts (Erkan 2012; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Jafarimehr 2010; Jin 2013a; Kalha 2003; McHenry 2002; Zhan 2011). Three studies were prospective (Brand 2000; de Jong 2012; Erkan 2012); 39 were retrospective (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Cherian 2010; Correa‐Gallego 2009; Doi 2002; Fisher 2008; Grieser 2010; Harrison 1999; Hong 2010; Hu 2013; Jafarimehr 2010; Jang 2014a; Jang 2014b; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kim 2015; Klau 2011; Kobayashi 2012; Kubo 2001; Kucera 2012; Lee 2014; Maire 2008; McHenry 2002; Nakagawa 2009; Ogawa 2008; Otomi 2014; Pais 2007; Sahani 2006; Saito 2013; Salla 2007; Sedlack 2002; Smith 2016; Takanami 2011; Taouli 2000; Zhan 2011; Zhan 2013); and 12 did not state whether they were prospective or retrospective (Choi 2003; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Higashi 1997; Kato 1995; Le Baleur 2011a; Nara 2009; Ogawa 2014; Takeshita 2008; Tan 2009; Tomimaru 2010; Yamao 2001).

None of the studies reported data on symptomatic and asymptomatic participants separately. Forty‐two studies (2086 participants) reported on cystic pancreatic lesions (Brandwein 2001 ‐ Cystic; Cellier 1998; Choi 2003; Correa‐Gallego 2009; de Jong 2012; Doi 2002; Fischer 2009 ‐ Cystic; Fisher 2008; Hong 2010; Hu 2013; Jang 2014a; Jang 2014b; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kim 2015; Kobayashi 2012; Kubo 2001; Kucera 2012; Le Baleur 2011a; Lee 2014; Maire 2008; McHenry 2002; Nakagawa 2009; Nara 2009; Ogawa 2008; Ogawa 2014; Pais 2007; Sahani 2006; Saito 2013; Salla 2007; Sedlack 2002; Smith 2016; Takanami 2011; Takeshita 2008; Tan 2009; Taouli 2000; Tomimaru 2010; Yamao 2001; Zhan 2011; Zhan 2013). Four studies reported on solid pancreatic lesions (Brandwein 2001 ‐ Solid; Cherian 2010; Fischer 2009 ‐ Solid; Klau 2011). The remaining eight studies either did not mention whether the lesions were cystic or solid, or did not report this information separately (Brand 2000; Erkan 2012; Grieser 2010; Harrison 1999; Higashi 1997; Jafarimehr 2010; Kato 1995; Otomi 2014). None of the studies reported data on people with chronic pancreatitis separately.

Overall, 12 studies reported data on EUS results (Brand 2000; Cellier 1998; de Jong 2012; Doi 2002; Harrison 1999; Kamata 2016a; Kobayashi 2012; Kubo 2001; Lee 2014; Nakagawa 2009; Sedlack 2002; Yamao 2001); 19 studies reported data on EUS‐FNA (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cherian 2010; Correa‐Gallego 2009; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Jin 2013a; Jin 2015; Kalha 2003; Kucera 2012; Maire 2008; McHenry 2002; Pais 2007; Salla 2007; Sedlack 2002; Smith 2016; Zhan 2011; Zhan 2013); 10 studies reported data on PET (Erkan 2012; Higashi 1997; Hong 2010; Hu 2013; Jafarimehr 2010; Kato 1995; Otomi 2014; Saito 2013; Takanami 2011; Tomimaru 2010); 13 studies reported data on CT (Cellier 1998; Grieser 2010; Harrison 1999; Hong 2010; Le Baleur 2011a; Nakagawa 2009; Nara 2009; Ogawa 2008; Sahani 2006; Takeshita 2008; Tan 2009; Taouli 2000; Yamao 2001); and eight studies reported data on MRI (Choi 2003; de Jong 2012; Jang 2014a; Jang 2014b; Kim 2015; Klau 2011; Ogawa 2014; Sahani 2006).

The criteria for a positive test result varied widely by study and are described in detail in Characteristics of included studies. The reference standards in all of the included studies was surgical excision.

Methodological quality of included studies

The methodological quality of the included studies is summarised in Figure 3 and Figure 4. None of the included studies was of high methodological quality.


Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.

Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.


Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.

Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.

Participant selection domain

In the participant selection domain, nine studies had a low risk of bias (Cherian 2010; Correa‐Gallego 2009; Fisher 2008; Kamata 2016a; Klau 2011; Nara 2009; Otomi 2014; Saito 2013; Salla 2007). All of the studies had high applicability concerns because of concerns that the participants did not match the review question. The review question was to find out the diagnostic accuracy of these index tests in people with focal lesions. However, all of the studies meeting the inclusion criteria for this review except Cherian 2010 used surgical excision as the reference standard, suggesting that the surgeons considered these patients to be at high risk of malignancy based on the results of the index tests or the tests that patients had prior to or subsequent to the index test. Cherian 2010 was also at high risk of applicability concern because it excluded participants with resectable lesions on CT scan and included only those equivocal lesions on CT scan.

Index test domain

In the index test domain, nine studies were at low risk of bias (Correa‐Gallego 2009; Hong 2010; Jang 2014b; Kim 2015; Kubo 2001; Nara 2009; Ogawa 2014; Tan 2009; Taouli 2000). Of the remaining studies, 31 were at unclear risk of bias because it was unclear whether the index test results were interpreted without knowledge of the results of the reference standard (Brand 2000; Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Choi 2003; Cherian 2010; de Jong 2012; Doi 2002; Erkan 2012; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Harrison 1999; Jafarimehr 2010; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kato 1995; Kobayashi 2012; Kucera 2012; Le Baleur 2011a; McHenry 2002; Ogawa 2008; Pais 2007; Salla 2007; Sedlack 2002; Smith 2016; Yamao 2001; Zhan 2011; Zhan 2013). Fifteen studies were at high risk of bias because the threshold for the index test was not prespecified (Grieser 2010; Higashi 1997; Hu 2013; Jang 2014a; Klau 2011; Lee 2014; Maire 2008; Nakagawa 2009; Otomi 2014; Sahani 2006; Saito 2013; Takanami 2011; Takeshita 2008; Tomimaru 2010; Zhan 2013). Twenty‐eight studies had low applicability concerns (Brand 2000; Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Choi 2003; Cherian 2010; Correa‐Gallego 2009; de Jong 2012; Doi 2002; Hong 2010; Hu 2013; Jang 2014b; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kim 2015; Kubo 2001; Le Baleur 2011a; Nara 2009; Ogawa 2008; Ogawa 2014; Pais 2007; Sahani 2006; Sedlack 2002; Smith 2016; Yamao 2001; Zhan 2013), and the remaining 27 studies had high applicability concerns because of concerns that the index test, its conduct, or interpretation differed from the review question (Erkan 2012; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Grieser 2010; Harrison 1999; Higashi 1997; Jafarimehr 2010; Jang 2014a; Kato 1995; Klau 2011; Kobayashi 2012; Kucera 2012; Lee 2014; Maire 2008; McHenry 2002; Nakagawa 2009; Otomi 2014; Saito 2013; Salla 2007; Sedlack 2002; Takanami 2011; Takeshita 2008; Tan 2009; Taouli 2000; Tomimaru 2010; Zhan 2011).

Reference standard domain

In the reference standard domain, three studies were at low risk of bias (Correa‐Gallego 2009; Grieser 2010; Hu 2013). Two studies were at high risk of bias because the reference standard results were not interpreted without knowledge of the index test results (Pais 2007), or because radiological and clinical follow‐up was used in some of the participants as the reference standard. The remaining 49 studies were at unclear risk of bias as it was unclear if the reference standard was interpreted without knowledge of the results of index tests (Brand 2000; Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cellier 1998; Choi 2003; de Jong 2012; Doi 2002; Erkan 2012; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Harrison 1999; Higashi 1997; Hong 2010; Jafarimehr 2010; Jang 2014a; Jang 2014b; Jin 2013a; Jin 2015; Kalha 2003; Kamata 2016a; Kato 1995; Kim 2015; Klau 2011; Kobayashi 2012; Kubo 2001; Kucera 2012; Le Baleur 2011a; Lee 2014; Maire 2008; McHenry 2002; Nakagawa 2009; Nara 2009; Ogawa 2008; Ogawa 2014; Otomi 2014; Sahani 2006; Saito 2013; Salla 2007; Sedlack 2002; Smith 2016; Takanami 2011; Takeshita 2008; Tan 2009; Taouli 2000; Tomimaru 2010; Yamao 2001; Zhan 2011; Zhan 2013). All studies were at low concern for applicability, as we considered the definition of the target condition by the reference standard to match the review question.

Flow and timing domain

None of the studies were at low risk of bias in the flow and timing domain. Thirty studies were at high risk of bias because not all of the participants were included in the analysis, or there was an inappropriate interval between the index test and reference standard (Brand 2000; Brandwein 2001 ‐ Cystic; Cellier 1998; Choi 2003; Correa‐Gallego 2009; de Jong 2012; Doi 2002; Erkan 2012; Hong 2010; Jang 2014a; Jang 2014b; Jin 2015; Kalha 2003; Kamata 2016a; Kato 1995; Kim 2015; Kobayashi 2012; Kucera 2012; Lee 2014; Nakagawa 2009; Ogawa 2008; Ogawa 2014; Pais 2007; Sedlack 2002; Smith 2016; Takanami 2011; Tan 2009; Taouli 2000; Tomimaru 2010; Yamao 2001). One study was at high risk of bias because the reference standards that participants received were dependent on the index test results. The remaining 23 studies were at unclear risk of bias because it was either unclear if there was an appropriate interval between the index test and reference standard or if all participants were included in the analysis, or both (Brandwein 2001 ‐ Solid; Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Fisher 2008; Grieser 2010; Harrison 1999; Higashi 1997; Hu 2013; Jafarimehr 2010; Jin 2013a; Klau 2011; Kubo 2001; Le Baleur 2011a; Maire 2008; McHenry 2002; Nara 2009; Otomi 2014; Sahani 2006; Saito 2013; Salla 2007; Takeshita 2008; Zhan 2011; Zhan 2013).

Findings

The results are summarised in the summary of findings Table. The overall sensitivities and specificities for different tests for different target conditions are tabulated in Table 2. A detailed description is given below.

Open in table viewer
Table 2. Summary sensitivity and specificity of different tests for different target conditions

Comparison

Name of test

Sensitivity

Specificity

Cancerous versus benign or precancerous

EUS‐FNA
(cytology)

0.79 (95% CI 0.60 to 0.91)

1.00 (95% CI 0.85 to 1.00)

Cancerous versus benign or precancerous

EUS‐FNA
(CEA > 500 ng/mL)

0.93 (95% CI 0.70 to 0.99)

0.33 (95% CI 0.12 to 0.65)

Cancerous versus benign or precancerous

PET
(criteria unspecified)

0.85 (95% CI 0.73 to 0.92)

0.91 (95% CI 0.72 to 0.97)

Cancerous versus benign

EUS

0.95 (95% CI 0.84 to 0.99)

0.53 (95% CI 0.31 to 0.74)

Cancerous versus benign

EUS‐FNA
(cytology)

0.79 (95% CI 0.07 to 1.00)

1.00 (95% CI 0.91 to 1.00)

Cancerous versus benign

PET (criteria unspecified)

0.92 (95% CI 0.80 to 0.97)

0.65 (95% CI 0.39 to 0.85)

Cancerous versus benign

PET (SUVmax > 3.5)

0.96 (95% CI 0.87 to 0.99)

0.62 (95% CI 0.43 to 0.78)

Cancerous versus benign

CT

0.98 (95% CI 0.00 to 1.00)

0.76 (95% CI 0.02 to 1.00)

Cancerous versus benign

MRI

0.80 (95% CI 0.58 to 0.92)

0.89 (95% CI 0.57 to 0.98)

Precancerous or cancerous versus benign

EUS

0.92 (95% CI 0.74 to 0.98)

0.60 (95% CI 0.31 to 0.83)

Precancerous or cancerous versus benign

EUS‐FNA (cytology)

0.73 (95% CI 0.01 to 1.00)

0.94 (95% CI 0.15 to 1.00)

Precancerous or cancerous versus benign

EUS‐FNA
(CEA > 50 ng/mL)

0.29 (95% CI 0.08 to 0.64)

0.25 (95% CI 0.05 to 0.70)

Precancerous or cancerous versus benign

PET
(SUVmax 2.4)

0.94 (95% CI 0.74 to 0.99)

0.93 (95% CI 0.69 to 0.99)

Precancerous or cancerous versus benign

CT

0.62 (95% CI 0.45 to 0.76)

0.64 (95% CI 0.39 to 0.84)

Precancerous or cancerous versus benign

MRI

0.93 (95% CI 0.69 to 0.99)

0.85 (95% CI 0.58 to 0.96)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS

0.78 (95% CI 0.45 to 0.94)

0.91 (95% CI 0.61 to 0.98)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS‐FNA
(cytology)

0.66 (95% CI 0.03 to 0.99)

0.92 (95% CI 0.73 to 0.98)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS‐FNA
(CEA > 200 ng/mL)

1.00 (95% CI 0.57 to 1.00)

0.64 (95% CI 0.48 to 0.78)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

CT

0.72 (95% CI 0.50 to 0.87)

0.92 (95% CI 0.81 to 0.97)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

MRI

0.75 (95% CI 0.30 to 0.95)

0.93 (95% CI 0.77 to 0.98)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS

0.86 (95% CI 0.74 to 0.92)

0.91 (95% CI 0.83 to 0.96)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(cytology)

0.47 (95% CI 0.24 to 0.70)

0.91 (95% CI 0.32 to 1.00)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CEA > 200 ng/mL)

0.58 (95% CI 0.28 to 0.83)

0.51 (95% CI 0.19 to 0.81)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CA 19‐9 > 1000 U/mL)

0.90 (95% CI 0.60 to 0.98)

0.42 (95% CI 0.26 to 0.59)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CEA > 692.8 ng/mL)

0.80 (95% CI 0.49 to 0.94)

0.90 (95% CI 0.60 to 0.98)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

PET
(SUVmax > 2 to 2.5)

0.90 (95% CI 0.79 to 0.96)

0.94 (95% CI 0.81 to 0.99)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

CT

0.87 (95% CI 0.00 to 1.00)

0.96 (95% CI 0.00 to 1.00)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

MRI

0.69 (95% CI 0.44 to 0.86)

0.93 (95% CI 0.43 to 1.00)

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia)

EUS

0.77 (95% CI 0.50 to 0.92)

0.89 (95% CI 0.76 to 0.96)

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia)

CT

0.50 (95% CI 0.22 to 0.78)

0.95 (95% CI 0.83 to 0.99)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)

CT

0.83 (95% CI 0.68 to 0.92)

0.83 (95% CI 0.64 to 0.93)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)

MRI

0.80 (95% CI 0.58 to 0.92)

0.81 (95% CI 0.53 to 0.95)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign

EUS

0.97 (95% CI 0.83 to 0.99)

0.40 (95% CI 0.26 to 0.55)

Cystic lesion subgroup analysis

Cancerous versus benign ‐ EUS‐FNA (cytology)

0.50 (95% CI 0.19 to 0.81)

1.00 (95% CI 0.84 to 1.00)

Cystic lesion subgroup analysis

Cancerous versus benign ‐ PET
(SUVmax > 3.5)

0.96 (95% CI 0.87 to 0.99)

0.62 (95% CI 0.43 to 0.78)

Cystic lesion subgroup analysis

Precancerous or cancerous versus benign ‐ EUS‐FNA
(cytology)

0.43 (95% CI 0.19 to 0.71)

1.00 (95% CI 0.74 to 1.00)

CA 19‐9: carbohydrate antigen 19‐9
CEA: carcinoembryonic antigen
CI: confidence interval
CT: computed tomography
EUS: endoscopic ultrasound
FNA: fine‐needle aspiration
MRI: magnetic resonance imaging
PET: positron emission tomography
SUVmax: maximum standardised uptake values

Cancerous versus benign or precancerous

EUS‐FNA cytology: We included one study reporting data on 45 participants for this test (McHenry 2002). The sensitivity and specificity for diagnosing cancer were 0.79 (95% confidence interval (CI) 0.60 to 0.91) and 1.00 (95% CI 0.85 to 1.00), respectively.

EUS‐FNA (CEA > 500 ng/mL): We included one study reporting data on 24 participants for this test (Kalha 2003). The sensitivity and specificity for diagnosing cancer were 0.93 (95% CI 0.70 to 0.99) and 0.33 (95% CI 0.12 to 0.65), respectively.

PET (criteria: not specified): We included one study reporting data on 76 participants for this test (Jafarimehr 2010). The sensitivity and specificity for diagnosing cancer were 0.85 (95% CI 0.73 to 0.92) and 0.91 (95% CI 0.72 to 0.97), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'cancerous versus benign or precancerous' studies is shown in Figure 5.


Forest plot ‐ Cancerous versus benign or precancerous.

Forest plot ‐ Cancerous versus benign or precancerous.

Cancerous versus benign

EUS: Two studies reporting data on 133 participants were included for this test, allowing meta‐analysis to be performed (Brand 2000; Harrison 1999). The summary sensitivity and summary specificity for diagnosing cancer were 0.95 (95% CI 0.84 to 0.99) and 0.53 (95% CI 0.31 to 0.74), respectively.

EUS‐FNA cytology: Three studies reporting data on 147 participants were included for this test (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid; Cherian 2010). The sensitivity and specificity for diagnosing cancer were 0.79 (95% CI 0.07 to 1.00) and 1.00 (95% CI 0.91 to 1.00), respectively.

PET (criteria: not specified): Three studies reporting data on 99 participants were included for this test, allowing meta‐analysis to be performed (Erkan 2012; Higashi 1997; Kato 1995). The summary sensitivity and summary specificity for diagnosing cancer were 0.92 (95% CI 0.80 to 0.97) and 0.65 (95% CI 0.39 to 0.85), respectively.

PET (SUVmax > 3.5): We included one study reporting data on 80 participants for this test (Hu 2013). The sensitivity and specificity for diagnosing cancer were 0.96 (95% CI 0.87 to 0.99) and 0.62 (95% CI 0.43 to 0.78), respectively.

CT: Two studies reporting data on 123 participants were included for this test, allowing meta‐analysis to be performed (Grieser 2010; Harrison 1999). The summary sensitivity and summary specificity for diagnosing cancer were 0.98 (95% CI 0.00 to 1.00) and 0.76 (95% CI 0.02 to 1.00), respectively.

MRI: We included one study reporting data on 29 participants for this test (Klau 2011). The sensitivity and specificity for diagnosing cancer were 0.80 (95% CI 0.58 to 0.92) and 0.89 (95% CI 0.57 to 0.98), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'cancerous versus benign' studies is shown in Figure 6.


Forest plot ‐ Cancerous versus benign.

Forest plot ‐ Cancerous versus benign.

Precancerous or cancerous versus benign

EUS: We included one study reporting data on 34 participants for this test (Sedlack 2002). The sensitivity and specificity for diagnosing cancer or precancer were 0.92 (95% CI 0.74 to 0.98) and 0.60 (95% CI 0.31 to 0.83), respectively.

EUS‐FNA cytology: We included three studies, reporting data on 52 participants for this test (Fischer 2009 ‐ Cystic; Fischer 2009 ‐ Solid; Sedlack 2002). The summary sensitivity and summary specificity for diagnosing cancer or precancer were 0.73 (95% CI 0.01 to 1.00) and 0.94 (95% CI 0.15 to 1.00), respectively.

EUS‐FNA (CEA > 50 ng/mL): We included one study reporting data on 11 participants for this test (Sedlack 2002). The sensitivity and specificity for diagnosing cancer or precancer were 0.29 (95% CI 0.08 to 0.64) and 0.25 (95% CI 0.05 to 0.70), respectively.

PET (SUVmax > 2.4): We included one study reporting data on 32 participants for this test (Otomi 2014). The sensitivity and specificity for diagnosing cancer or precancer were 0.94 (95% CI 0.74 to 0.99) and 0.93 (95% CI 0.69 to 0.99), respectively.

CT: We included one study reporting data on 48 participants for this test (Fisher 2008). The sensitivity and specificity for diagnosing cancer or precancer were 0.62 (95% CI 0.45 to 0.76) and 0.64 (95% CI 0.39 to 0.84), respectively.

MRI: We included one study reporting data on 27 participants for this test (Jang 2014a). The sensitivity and specificity for diagnosing cancer or precancer were 0.93 (95% CI 0.69 to 0.99) and 0.85 (95% CI 0.58 to 0.96), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'precancerous or cancerous versus benign' studies is shown in Figure 7.


Forest plot ‐ Precancerous or cancerous versus benign.

Forest plot ‐ Precancerous or cancerous versus benign.

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS: Five studies reporting data on 156 participants were included for this test, allowing meta‐analysis to be performed (Cellier 1998; de Jong 2012; Nakagawa 2009; Yamao 2001; Zhan 2011). The summary sensitivity and summary specificity for diagnosing invasive cancer were 0.78 (95% CI 0.45 to 0.94) and 0.91 (95% CI 0.61 to 0.98), respectively.

EUS‐FNA cytology: Three studies reporting data on 158 participants were included for this test, allowing meta‐analysis to be performed (Jin 2013a; Pais 2007; Salla 2007). The summary sensitivity and summary specificity for diagnosing invasive cancer were 0.66 (95% CI 0.03 to 0.99) and 0.92 (95% CI 0.73 to 0.98), respectively.

EUS‐FNA (CEA > 200 ng/mL): We included one study reporting data on 41 participants for this test (Maire 2008). The sensitivity and specificity for diagnosing invasive cancer were 1.00 (95% CI 0.57 to 1.00) and 0.64 (95% CI 0.48 to 0.78), respectively.

CT: Six studies reporting data on 326 participants were included for this test, allowing meta‐analysis to be performed (Cellier 1998; Nakagawa 2009; Nara 2009; Ogawa 2008; Taouli 2000; Yamao 2001). The summary sensitivity and summary specificity for diagnosing invasive cancer were 0.72 (95% CI 0.50 to 0.87) and 0.92 (95% CI 0.81 to 0.97), respectively.

MRI: We included one study reporting data on 32 participants for this test (de Jong 2012). The sensitivity and specificity for diagnosing invasive cancer were 0.75 (95% CI 0.30 to 0.95) and 0.93 (95% CI 0.77 to 0.98), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'cancer (invasive carcinoma) versus precancerous (dysplasia)' studies is shown in Figure 8.


Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (dysplasia).

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (dysplasia).

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS: Four studies reporting data on 196 participants were included for this test, allowing meta‐analysis to be performed (Doi 2002; Kobayashi 2012; Lee 2014; Yamao 2001). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.86 (95% CI 0.74 to 0.92) and 0.91 (95% CI 0.83 to 0.96), respectively.

EUS‐FNA cytology: Three studies reporting data on 310 participants were included for this test, allowing meta‐analysis to be performed (Jin 2013a; Smith 2016; Zhan 2013). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.47 (95% CI 0.24 to 0.70) and 0.91 (95% CI 0.32 to 1.00), respectively.

EUS‐FNA (CEA > 200 ng/mL): Three studies reporting data on 160 participants were included for this test, allowing meta‐analysis to be performed (Correa‐Gallego 2009; Kucera 2012; Maire 2008). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.58 (95% CI 0.28 to 0.83) and 0.51 (95% CI 0.19 to 0.81), respectively.

EUS‐FNA (carbohydrate antigen 19‐9 > 1000 U/mL): We included one study reporting data on 41 participants for this test (Maire 2008). The sensitivity and specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.90 (95% CI 0.60 to 0.98) and 0.42 (95% CI 0.26 to 0.59), respectively.

EUS‐FNA (CEA > 692.8 ng/mL): We included one study reporting data on 20 participants for this test (Zhan 2013). The sensitivity and specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.80 (95% CI 0.49 to 0.94) and 0.90 (95% CI 0.60 to 0.98), respectively.

PET (SUVmax 2 to 2.5): Four studies reporting data on 124 participants were included for this test, allowing meta‐analysis to be performed (Hong 2010; Saito 2013; Takanami 2011; Tomimaru 2010). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.90 (95% CI 0.79 to 0.96) and 0.94 (95% CI 0.81 to 0.99), respectively.

CT: Three studies reporting data on 139 participants were included for this test, allowing meta‐analysis to be performed (Hong 2010; Le Baleur 2011a; Yamao 2001). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.87 (95% CI 0.00 to 1.00) and 0.96 (95% CI 0.00 to 1.00), respectively.

MRI: Three studies reporting data on 189 participants were included for this test, allowing meta‐analysis to be performed (Jang 2014b; Kim 2015; Ogawa 2014). The summary sensitivity and summary specificity for diagnosing high‐grade dysplasia or invasive cancer were 0.69 (95% CI 0.44 to 0.86) and 0.93 (95% CI 0.43 to 1.00), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'cancer (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)' studies is shown in Figure 9.


Forest plot ‐ Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia).

Forest plot ‐ Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia).

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia)

EUS: We included one study reporting data on 51 participants for this test (Kubo 2001). The sensitivity and specificity for diagnosing invasive cancer were 0.77 (95% CI 0.50 to 0.92) and 0.89 (95% CI 0.76 to 0.96), respectively.

CT: We included one study reporting data on 46 participants for this test (Takeshita 2008). The sensitivity and specificity for diagnosing invasive cancer were 0.50 (95% CI 0.22 to 0.78) and 0.95 (95% CI 0.83 to 0.99), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'cancer (invasive carcinoma) versus precancerous (low‐grade dysplasia)' studies is shown in Figure 10.


Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia).

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia).

Precancerous or cancer (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)

CT: Three studies reporting data on 106 participants were included for this test, allowing meta‐analysis to be performed (Ogawa 2008; Sahani 2006; Tan 2009). The summary sensitivity and summary specificity for diagnosing intermediate‐ or high‐grade dysplasia or invasive cancer were 0.83 (95% CI 0.68 to 0.92) and 0.83 (95% CI 0.64 to 0.93), respectively.

MRI: Two studies reporting data on 71 participants were included for this test, allowing meta‐analysis to be performed (Choi 2003; Takeshita 2008). The summary sensitivity and specificity for diagnosing intermediate‐ or high‐grade dysplasia or invasive cancer were 0.80 (95% CI 0.58 to 0.92) and 0.81 (95% CI 0.53 to 0.95), respectively.

The results including sensitivities and specificities and post‐test probabilities at median pre‐test probabilities are summarised in summary of findings Table. A forest plot summarising all of the sensitivity and specificity data for the 'precancerous or cancer (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)' studies is shown in Figure 11.


Forest plot ‐ Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia).

Forest plot ‐ Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia).

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign

EUS: We included one study reporting data on 70 participants for this test (Kamata 2016a). The sensitivity and specificity for diagnosing intermediate‐ or high‐grade dysplasia or invasive carcinoma were 0.97 (95% CI 0.83 to 0.99) and 0.40 (95% CI 0.26 to 0.55), respectively.

The results including sensitivity and specificity and post‐test probability at median pre‐test probability are summarised in summary of findings Table. A forest plot summarising the sensitivity and specificity data for the 'precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign' study is shown in Figure 12.


Forest plot of 33 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.

Forest plot of 33 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.

Subgroup analyses

We assessed the performance of the tests excluding any studies investigating participants with solid lesions and those in which information for solid and cystic lesions was not reported separately. All of the studies assessing the ability of different imaging modalities to differentiate precancerous versus cancerous lesions regardless of the definitions used by authors for precancer and cancer (Analysis 4 to Analysis 8) included participants with cystic focal pancreatic lesions only, therefore all the results reported are for cystic focal pancreatic lesions only.

In the analysis assessing the ability of different imaging modalities to differentiate benign or precancerous versus cancerous lesions, we excluded one study because if did not specify the type (solid or cystic) of lesions for which included participants were investigated (Jafarimehr 2010). However, as this study did not contribute to a meta‐analysis, there were no changes to the analysis.

Cancerous versus benign

In the analysis assessing the ability of different imaging modalities to differentiate benign versus cancerous lesions, we excluded eight studies because they did not explicitly include participants with cystic lesions (Brand 2000; Erkan 2012; Grieser 2010; Harrison 1999; Higashi 1997; Hu 2013; Kato 1995; Klau 2011), and one study that only had one component included (Brandwein 2001 ‐ Cystic; Brandwein 2001 ‐ Solid). This left two remaining studies, which did not contribute to a meta‐analysis due to the exclusion of the other studies (Brandwein 2001 ‐ Cystic; Hu 2013). We therefore performed no meta‐analyses for this group. The new findings for benign versus cancerous lesions are described below.

EUS‐FNA: We included one study reporting data on 26 participants for this test (Brandwein 2001 ‐ Cystic). The sensitivity and specificity for diagnosing cancer were 0.50 (95% CI 0.19 to 0.81) and 1.00 (95% CI 0.84 to 1.00), respectively.

PET: We included one study reporting data on 80 participants for this test (Hu 2013). The sensitivity and specificity for diagnosing cancer were 0.96 (95% CI 0.87 to 0.99) and 0.62 (95% CI 0.43 to 0.78), respectively.

A forest plot summarising all the sensitivity and specificity data for the cystic subgroup analysis of 'cancerous versus benign lesion' studies is shown in Figure 13.


Forest plot ‐ Cystic lesion subgroup analysis: Cancerous versus benign.

Forest plot ‐ Cystic lesion subgroup analysis: Cancerous versus benign.

Precancerous or cancerous versus benign

In the analysis assessing the ability of different imaging modalities to differentiate precancerous or cancerous versus benign lesions, we excluded a component of one study because the participants had solid pancreatic lesions (Fischer 2009 ‐ Solid). We therefore re‐performed the meta‐analysis for precancerous or cancerous versus benign lesions ‐ EUS‐FNA without these data. The remaining tests for this target condition did not have any studies excluded and were therefore not redone. The new findings are described below.

EUS‐FNA: Two studies reporting data on 34 participants were included for this test, allowing meta‐analysis to be performed (Fischer 2009 ‐ Cystic; Sedlack 2002). The summary sensitivity and summary specificity for diagnosing precancer or cancer were 0.43 (95% CI 0.19 to 0.71) and 1.00 (95% CI 0.74 to 1.00), respectively.

A forest plot summarising the sensitivity and specificity data for the cystic subgroup analysis of 'precancerous or cancerous versus benign lesion' study is shown in Figure 14.


Forest plot ‐ Cystic lesion subgroup analysis: Precancerous or cancerous versus benign ‐ EUS‐FNA.

Forest plot ‐ Cystic lesion subgroup analysis: Precancerous or cancerous versus benign ‐ EUS‐FNA.

Discussion

Summary of main results

The results are summarised in summary of findings Table.

We included 54 studies involving a total of 3196 participants that evaluated the diagnostic accuracy of various imaging modalities (EUS, EUS‐FNA, PET, CT, and MRI) for characterising focal pancreatic lesions. We identified eight different target conditions in these studies, with the studies using imaging modalities to differentiate: cancerous versus benign or precancerous lesions; cancerous versus benign lesions; precancerous or cancerous lesions versus benign lesions; cancerous (invasive carcinoma) versus precancerous (dysplasia) lesions; cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) lesions; cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions; precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions; and precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign lesions. The wide variety of tumour types that constituted benign and cancerous lesions within the studies meant that only a few meaningful meta‐analyses could be performed. None of the comparisons in which single studies were included were of sufficiently high methodological quality to warrant highlighting of the results.

For differentiation of cancerous lesions from benign or precancerous lesions, only single studies were included and therefore meta‐analysis was not performed. Overall, EUS‐FNA (cytology) had a sensitivity of 0.79 (95% CI 0.60 to 0.91) and specificity of 1.00 (95% CI 0.85 to 1.00); EUS‐FNA (CEA > 500 ng/mL) had a sensitivity of 0.93 (95% CI 0.70 to 0.99) and specificity of 0.33 (95% CI 0.12 to 0.65); and PET had a sensitivity of 0.85 (95% CI 0.73 to 0.92) and specificity of 0.91 (95% CI 0.72 to 0.97).

The second analysis, of studies differentiating cancerous versus benign lesions, provided three tests in which meta‐analysis could be performed, however the data were sparse: one of these tests contained three studies, and the remaining two tests contained two studies, meaning the meta‐analysis was of limited value. There was little difference in the diagnostic test accuracy between the imaging techniques. EUS‐FNA achieved very high specificity (of 1.00, i.e. no false negatives) but modest sensitivity (0.79; 95% CI 0.07 to 1.00). A high specificity of EUS‐FNA can be expected, since this involves physically sampling the lesion. However, the modest sensitivity may reflect that the sampling methods were inadequate. Additional guidance such as identifying the location most likely to yield the correct results or additional guidance using optical endoscopy techniques such as confocal laser microendoscopy may overcome this problem and improve the sensitivity of EUS (Giovannini 2012), but there are major challenges, such as knowing the area within the lesion that is being examined by confocal laser microendoscopy, that must be addressed before such methods can be used routinely.

The third analysis, of studies differentiating precancerous or cancerous and benign lesions, only provided one test (EUS‐FNA) for which meta‐analysis was performed. The results were unreliable due to significant heterogeneity in the results between the studies.

The fourth analysis, of studies differentiating cancerous (invasive carcinoma) versus precancerous (dysplasia) lesions, provided three tests in which meta‐analysis was performed, with one test containing five studies (EUS), one test containing three studies (EUS‐FNA), and the third test containing six studies (CT). All five of the tests included in the analysis had a similar level of accuracy according to their respective ROC curves. EUS and CT showed the highest (and similar) accuracy estimates (EUS = sensitivity 0.78 and specificity 0.91; CT = sensitivity 0.72 and specificity 0.92) and included the largest number of studies (five and six, respectively) among all comparisons.

The fifth analysis, of studies differentiating cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) lesions, provided six tests in which meta‐analysis was performed, with two tests containing four studies (EUS and PET SUVmax 2 to 2.5), one test containing two studies (EUS‐FNA), three tests containing three studies (EUS‐FNA > 200, CT, and MRI), and the remaining two tests providing single studies. PET performed with the highest accuracy (sensitivity 0.90 (95% CI 0.79 to 0.96) and specificity 0.94 (95% CI 0.81 to 0.99)).

The sixth analysis, of studies differentiating cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions, provided no tests in which meta‐analysis was performed.

The seventh analysis, of studies differentiating precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) lesions, provided two tests in which meta‐analysis was performed. The meta‐analysis results for CT (sensitivity 0.83 (95% CI 0.68 to 0.92) and specificity 0.83 (95% CI 0.64 to 0.93)) were similar to those of MRI (sensitivity 0.80 (95% CI 0.58 to 0.92) and specificity 0.81 (95% CI 0.53 to 0.95)), however lack of significant data means little can be inferred from this.

The eighth analysis, of studies differentiating precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign lesions, provided no tests in which meta‐analysis was performed.

We performed a subgroup analysis to investigate the performance of imaging modalities for cystic pancreatic lesions. This only resulted in alterations to the 'cancerous versus benign or precancerous', 'cancerous versus benign', and 'precancerous or cancerous versus benign' groups, however when re‐performed in these groups, the analysis did not result in any significant changes.

Overall, none of the tests assessed had sufficient overall diagnostic accuracy to be considered a definitive diagnostic modality. High sensitivity of the test is required so that precancer or cancer is not missed. High specificity is required to avoid major surgery. Sensitivity and specificity in excess of 90% are required to recommend the particular modality over other modalities. Only PET in differentiating precancerous (low‐ or intermediate‐grade dysplasia) versus cancer (high‐grade dysplasia or invasive carcinoma) approaches this level of accuracy. Overall, modalities other than EUS‐FNA had moderate to high sensitivity but moderate specificity, while EUS‐FNA had high specificity with moderate sensitivity in distinguishing the nature of focal pancreatic lesions.

Strengths and weaknesses of the review

We conducted a thorough literature search and included full‐text publications and abstracts without any language restrictions. Two review authors independently identified and extracted data from the studies, potentially reducing the chance of error that would be associated with one person performing the data extraction. We used strict reference standards that are likely to diagnose the target condition with a high degree of accuracy. These were the major strengths of the review.

We included EUS‐FNA as part of the review. Strictly speaking, EUS‐FNA cannot be considered an imaging modality since it uses cytology criteria or levels of tumour markers in the aspirate rather than imaging features to make the diagnosis. We had mentioned at the protocol stage that we would include EUS‐FNA in this review, as the searches for EUS return EUS‐FNA as well and because EUS‐FNA along with the imaging modalities included in this review are the most widely used tests for characterising focal pancreatic lesions. Our review provides the most important information about the tests performed to characterise focal lesions in one location and hence is probably more useful for clinicians, who would otherwise have to search for another review for information on EUS‐FNA.

The major limitation in the review process was the diverse nature of the collected data, with a wide variety of definitions of benign, precancerous, and cancerous lesions. This limited the possible analysis of the data and the conclusions that could be made from our analyses. While some authors defined precancerous lesions as lesions with low‐ and intermediate‐grade dysplasia, others defined it as low‐grade dysplasia only, while yet others considered any form of dysplasia as precancerous lesions. In the comparison 'cancer versus benign', it is unclear how the study authors dealt with precancerous lesions, that is whether they included precancerous lesions in the 'cancer' group or the 'benign' group, or whether they simply excluded them, consequently undermining any conclusions that could be made for this comparison.

We could not perform a bivariate random‐effects model that takes correlation between sensitivity and specificity into account and were unable to compare the diagnostic test accuracy of index tests using formal statistical methods due to the sparseness of data for each comparison. As a result, we performed the analysis using simpler models suggested by Takwoingi 2015 and colleagues. We reported the model with the lowest ‐2 log likelihood and also visualised the forest plots and ROC plots in deciding the model to be reported. The confidence intervals were extremely wide for the following analyses: benign versus cancer: CT; benign versus cancer: EUS‐FNA; benign versus precancer or cancer: EUS‐FNA; and precancer (low‐ or intermediate‐grade dysplasia) versus cancer (high‐grade dysplasia or invasive carcinoma): CT. While fixed‐effect model provided narrower confidence intervals for some of the above analyses, such models were inappropriate for these data because of the poor overlap of confidence intervals on the forest plot. This observation (i.e. that fixed‐effect models were not appropriate) was supported by the ‐2 log likelihoods, which were higher for the univariate fixed‐effect model than those of the models presented. The alternative was not to perform a meta‐analysis at all, which is even more difficult to interpret. At least the current results allowed us to interpret that the sensitivity or specificity or both could not be estimated reliably. There was reasonable overlap of confidence intervals in the other meta‐analyses performed. With regard to the tests for which meta‐analysis could not be performed, the diagnostic test accuracy from single studies needs confirmation by other studies to assess whether the results are reproducible. Hence, we are unable to arrive at any major conclusions based on information by a single study.

A high proportion of studies were at high risk of bias and with high concern regarding applicability in all four domains of the QUADAS‐2 tool. This makes the validity of the results questionable. Of particular concern was the type of people who underwent these tests. Because of the strict but appropriate reference standard, all of the participants in all of the studies included in this review except Cherian 2010 underwent surgical resection. This suggested that the surgeons thought that these participants had high probability of having high‐grade dysplasia or cancer, either because of the results of this test or other tests performed alongside the index tests. Since most of the studies were retrospective studies, if participants were operated on on the basis of the index test, and only participants who underwent surgery were included, participants with negative index tests but who had cancer would have been excluded inappropriately. This would have resulted in overestimation of sensitivity. The studies did not report the proportion of people in whom the different tests were feasible. This is particularly important for EUS and EUS‐FNA, since the participants may have been selected to undergo EUS or EUS‐FNA based on the proximity to the stomach or duodenum. This increases the concern regarding applicability. The studies did not report the complications associated with the index test. While this is unlikely to influence the diagnostic accuracy of the index test, it may have implications in determining the balance of benefits and harms in choosing a test.

Another limitation of this review was that we have included sensitivity‐maximising diagnostic filters for searching MEDLINE and Embase databases (Haynes 2004; Wilczynski 2005), and also used terms to limit the searches in Science Citation Index. We did this because the original searches without the filters retrieved more than 60,000 references. We had to balance the possibility of missing some studies against the risk of not being able to complete the review. We decided that it is useful to have evidence from major studies rather than having no information at all. However, it must be noted that the diagnostic filters we used have a sensitivity of 98.6% for MEDLINE and 100% for Embase. Consequently, the chances that we missed some relevant diagnostic studies are extremely low. This was further reduced by performing a 'related search' and 'citing reference search', in which we found no studies that could be included in this review.

We identified six other systematic reviews on the topics included in this systematic review (Banafea 2016; Chen 2012; Fuccio 2013; Gillis 2015; Hewitt 2012; Mei 2013). These included the role of EUS‐FNA (cytology), K‐ras gene mutation analysis of FNA aspirate, and EUS elastography in focal pancreatic lesions. The diagnostic test accuracy in four of the studies showed that EUS cytology and K‐ras gene mutation analysis of FNA aspirate had a reasonably high sensitivity (0.80 to 0.86) and very high specificity (96% to 98%) in solid pancreatic lesions (Banafea 2016; Chen 2012; Fuccio 2013; Hewitt 2012). These studies accepted cytology and clinical follow‐up (without specifying the exact nature of acceptable clinical follow‐up) in addition to histopathology as reference standards (Chen 2012; Fuccio 2013; Hewitt 2012). It is likely that this methodological difference was responsible for the major differences between our observations and these systematic reviews. In addition, these systematic reviews restricted participants to those with solid pancreatic lesions (Banafea 2016; Chen 2012; Fuccio 2013; Hewitt 2012), which could be another explanation for the differences between our observations and these systematic reviews. One systematic review evaluated EUS elastography in focal pancreatic lesions and reported a high sensitivity of 0.95 and a specificity of 0.67 (Mei 2013). We did not identify any study evaluating EUS elastography that met our inclusion criteria with respect to our reference standard, therefore we are unable to comment on the observation by Mei 2013. The last systematic review evaluated the role of EUS‐FNA and EUS‐FNA molecular analysis (i.e. check for abnormal genes) in people with cystic pancreatic lesions. The authors found poor sensitivity and high specificity of EUS‐FNA, which is similar to our findings (Gillis 2015).

Applicability of findings to the review question

All studies had high applicability concerns, making the applicability of findings to the target patient population of all incidental lesions questionable. The findings are applicable only for people who are suspected to be at high risk of high‐grade dysplasia or cancer. The review question was to find out the diagnostic accuracy of these index tests in people with focal pancreatic lesions, usually detected incidentally. However, all of the studies that met the inclusion criteria for this review used surgical excision as the reference standard, suggesting that the surgeons considered these patients to have a high risk of malignancy based on the results of the index tests or any additional tests. In terms of current availability of these tests, CT scan and MRI are likely to be available in most secondary centres. EUS is likely to be available in limited secondary centres and most tertiary centres that treat pancreatic lesions. PET is likely to be available only in limited tertiary centres, although the tertiary centres are likely to have access to a PET scan. However, based on the observations in this review, there do not appear to be any major differences between the different imaging modalities. The improved sensitivity of EUS‐FNA compared to other imaging modalities is compensated by a corresponding decrease in sensitivity, consequently there do not appear to be major advantages to using EUS for characterising focal pancreatic lesions compared to other non‐invasive methodologies.

Clinical pathway.Abbreviations:Ca 19‐9: carbohydrate antigen 19‐9
 CT: computed tomography
 EUS: endoscopic ultrasound
 MRI: magnetic resonance imaging
 PET: positron emission tomography
Figuras y tablas -
Figure 1

Clinical pathway.

Abbreviations:

Ca 19‐9: carbohydrate antigen 19‐9
CT: computed tomography
EUS: endoscopic ultrasound
MRI: magnetic resonance imaging
PET: positron emission tomography

Study flow diagram.
Figuras y tablas -
Figure 2

Study flow diagram.

Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.
Figuras y tablas -
Figure 3

Risk of bias and applicability concerns graph: review authors' judgements about each domain presented as percentages across included studies.

Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.
Figuras y tablas -
Figure 4

Risk of bias and applicability concerns summary: review authors' judgements about each domain for each included study.

Forest plot ‐ Cancerous versus benign or precancerous.
Figuras y tablas -
Figure 5

Forest plot ‐ Cancerous versus benign or precancerous.

Forest plot ‐ Cancerous versus benign.
Figuras y tablas -
Figure 6

Forest plot ‐ Cancerous versus benign.

Forest plot ‐ Precancerous or cancerous versus benign.
Figuras y tablas -
Figure 7

Forest plot ‐ Precancerous or cancerous versus benign.

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (dysplasia).
Figuras y tablas -
Figure 8

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (dysplasia).

Forest plot ‐ Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia).
Figuras y tablas -
Figure 9

Forest plot ‐ Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia).

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia).
Figuras y tablas -
Figure 10

Forest plot ‐ Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia).

Forest plot ‐ Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia).
Figuras y tablas -
Figure 11

Forest plot ‐ Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia).

Forest plot of 33 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.
Figuras y tablas -
Figure 12

Forest plot of 33 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.

Forest plot ‐ Cystic lesion subgroup analysis: Cancerous versus benign.
Figuras y tablas -
Figure 13

Forest plot ‐ Cystic lesion subgroup analysis: Cancerous versus benign.

Forest plot ‐ Cystic lesion subgroup analysis: Precancerous or cancerous versus benign ‐ EUS‐FNA.
Figuras y tablas -
Figure 14

Forest plot ‐ Cystic lesion subgroup analysis: Precancerous or cancerous versus benign ‐ EUS‐FNA.

Cancerous versus benign or precancerous ‐ EUS‐FNA (cytology).
Figuras y tablas -
Test 1

Cancerous versus benign or precancerous ‐ EUS‐FNA (cytology).

Cancerous versus benign or precancerous ‐ EUS‐FNA (CEA > 500 ng/mL).
Figuras y tablas -
Test 2

Cancerous versus benign or precancerous ‐ EUS‐FNA (CEA > 500 ng/mL).

Cancerous versus benign or precancerous ‐ PET.
Figuras y tablas -
Test 3

Cancerous versus benign or precancerous ‐ PET.

Cancerous versus benign ‐ EUS.
Figuras y tablas -
Test 4

Cancerous versus benign ‐ EUS.

Cancerous versus benign ‐ EUS‐FNA.
Figuras y tablas -
Test 5

Cancerous versus benign ‐ EUS‐FNA.

Cancerous versus benign ‐ PET.
Figuras y tablas -
Test 6

Cancerous versus benign ‐ PET.

Cancerous versus benign ‐ PET (SUVmax > 3.5).
Figuras y tablas -
Test 7

Cancerous versus benign ‐ PET (SUVmax > 3.5).

Cancerous versus benign ‐ CT.
Figuras y tablas -
Test 8

Cancerous versus benign ‐ CT.

Cancerous versus benign ‐ MRI.
Figuras y tablas -
Test 9

Cancerous versus benign ‐ MRI.

Precancerous or cancerous versus benign ‐ EUS.
Figuras y tablas -
Test 10

Precancerous or cancerous versus benign ‐ EUS.

Precancerous or cancerous versus benign ‐ EUS‐FNA.
Figuras y tablas -
Test 11

Precancerous or cancerous versus benign ‐ EUS‐FNA.

Precancerous or cancerous versus benign ‐ EUS‐FNA (CEA > 50 ng/mL).
Figuras y tablas -
Test 12

Precancerous or cancerous versus benign ‐ EUS‐FNA (CEA > 50 ng/mL).

Precancerous or cancerous versus benign ‐ PET (SUVmax > 2.4).
Figuras y tablas -
Test 13

Precancerous or cancerous versus benign ‐ PET (SUVmax > 2.4).

Precancerous or cancerous versus benign ‐ CT.
Figuras y tablas -
Test 14

Precancerous or cancerous versus benign ‐ CT.

Precancerous or cancerous versus benign ‐ MRI.
Figuras y tablas -
Test 15

Precancerous or cancerous versus benign ‐ MRI.

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS.
Figuras y tablas -
Test 16

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS.

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA.
Figuras y tablas -
Test 17

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA.

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL).
Figuras y tablas -
Test 18

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL).

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ CT.
Figuras y tablas -
Test 19

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ CT.

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ MRI.
Figuras y tablas -
Test 20

Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ MRI.

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS.
Figuras y tablas -
Test 21

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS.

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA.
Figuras y tablas -
Test 22

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA.

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL).
Figuras y tablas -
Test 23

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL).

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CA 19‐9 > 1000 U/mL).
Figuras y tablas -
Test 24

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CA 19‐9 > 1000 U/mL).

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 692.8 ng/mL).
Figuras y tablas -
Test 25

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 692.8 ng/mL).

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ PET (SUVmax 2 to 2.5).
Figuras y tablas -
Test 26

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ PET (SUVmax 2 to 2.5).

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ CT.
Figuras y tablas -
Test 27

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ CT.

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ MRI.
Figuras y tablas -
Test 28

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ MRI.

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ EUS.
Figuras y tablas -
Test 29

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ EUS.

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT.
Figuras y tablas -
Test 30

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT.

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT.
Figuras y tablas -
Test 31

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT.

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ MRI.
Figuras y tablas -
Test 32

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ MRI.

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.
Figuras y tablas -
Test 33

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS.

Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ EUS‐FNA.
Figuras y tablas -
Test 34

Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ EUS‐FNA.

Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ PET.
Figuras y tablas -
Test 35

Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ PET.

Cystic lesion subgroup analysis ‐ Precancerous or cancerous versus benign ‐ EUS‐FNA.
Figuras y tablas -
Test 36

Cystic lesion subgroup analysis ‐ Precancerous or cancerous versus benign ‐ EUS‐FNA.

Summary of findings Imaging modalities for characterising focal pancreatic lesions

Name of test

Number of studies (number of participants)

Sensitivity (95% CI)

Specificity (95% CI)

Post‐test probability of positive test* (95% CI)

Post‐test probability of negative test* (95% CI)

Number of false positives per 100 positive index test results (95% CI)

Number of false negatives per 100 negative index test results (95% CI)

Risk of bias

Applicability concerns

Uncertainty (due to inconsistency or inability to assess inconsistency, and random errors because of overall small sample size)

Cancerous versus benign or precancerous (median pre‐test probability: 63%)

EUS‐FNA (cytology)

1 (45)

0.79 (0.60 to 0.91)

1.00 (0.85 to 1.00)

98% (79% to 100%)

26% (14% to 43%)

2 (0 to 21)

26 (14 to 43)

Unclear

High

High

EUS‐FNA (CEA > 500 ng/mL)

1 (24)

0.93 (0.70 to 0.99)

0.33 (0.12 to 0.65)

70% (59% to 79%)

25% (4% to 73%)

30 (21 to 41)

25 (4 to 73)

High

High

High

PET (criteria unspecified)

1 (76)

0.85 (0.73 to 0.92)

0.91 (0.72 to 0.97)

94% (81% to 98%)

21% (12% to 34%)

6 (2 to 19)

21 (12 to 34)

Unclear

High

High

Cancerous versus benign (median pre‐test probability: 70%)

EUS

2 (133)

0.95 (0.84 to 0.99)

0.53 (0.31 to 0.74)

82% (74% to 88%)

18% (6% to 45%)

18 (12 to 26)

18 (6 to 45)

Unclear or high

High

High

EUS‐FNA (cytology)

3 (147)

0.79 (0.07 to 1.00)

1.00 (0.91 to 1.00)

99% (90% to 100%)

32% (2% to 92%)

0 (0 to 9)

32 (2 to 92)

High

High

High

PET (criteria unspecified)

3 (99)

0.92 (0.80 to 0.97)

0.65 (0.39 to 0.85)

86% (75% to 92%)

22% (9% to 44%)

14 (8 to 25)

22 (9 to 44)

High

High

High

PET (SUVmax > 3.5)

1 (80)

0.96 (0.87 to 0.99)

0.62 (0.43 to 0.78)

85% (78% to 90%)

12% (3% to 36%)

15 (10 to 22)

12 (3 to 36)

High

High

High

CT

2 (123)

0.98 (0.00 to 1.00)

0.76 (0.02 to 1.00)

90% (17% to 100%)

6% (0% to 100%)

10 (0 to 83)

6 (0 to 100)

Unclear or high

High

High

MRI

1 (29)

0.80 (0.58 to 0.92)

0.89 (0.57 to 0.98)

94% (72% to 99%)

34% (17% to 56%)

6 (1 to 28)

34 (17 to 56)

High

High

High

Precancerous or cancerous versus benign (median pre‐test probability: 71%)

EUS

1 (34)

0.92 (0.74 to 0.98)

0.60 (0.31 to 0.83)

85% (72% to 92%)

25% (7% to 58%)

15 (8 to 28)

25 (7 to 58)

High

High

High

EUS‐FNA (cytology)

2 (52)

0.73 (0.01 to 1.00)

0.94 (0.15 to 1.00)

97% (25% to 100%)

41% (1% to 98%)

3 (0 to 75)

41 (1 to 98)

Unclear or high

High

High

EUS‐FNA (CEA > 50 ng/mL)

1 (11)

0.29 (0.08 to 0.64)

0.25 (0.05 to 0.70)

48% (20% to 77%)

87% (54% to 98%)

52 (23 to 80)

87 (54 to 98)

High

High

High

PET (SUVmax 2.4)

1 (32)

0.94 (0.74 to 0.99)

0.93 (0.69 to 0.99)

97% (83% to 100%)

13% (2% to 49%)

3 (0 to 17)

13 (2 to 49)

High

High

High

CT

1 (48)

0.62 (0.45 to 0.76)

0.64 (0.39 to 0.84)

81% (66% to 90%)

59% (44% to 72%)

19 (10 to 34)

59 (44 to 72)

Unclear

High

High

MRI

1 (27)

0.93 (0.69 to 0.99)

0.85 (0.58 to 0.96)

94% (80% to 98%)

17% (3% to 58%)

6 (2 to 20)

17 (3 to 58)

High

High

High

Cancerous (invasive carcinoma) versus precancerous (dysplasia) (median pre‐test probability: 27%)

EUS

5 (156)

0.78 (0.45 to 0.94)

0.91 (0.61 to 0.98)

75% (37% to 94%)

8% (3% to 22%)

25 (6 to 63)

8 (3 to 22)

Unclear or high

High

High

EUS‐FNA (cytology)

3 (158)

0.66 (0.03 to 0.99)

0.92 (0.73 to 0.98)

75% (29% to 95%)

12% (1% to 69%)

25 (5 to 71)

12 (1 to 69)

Unclear or high

High

High

EUS‐FNA (CEA > 200 ng/mL)

1 (41)

1.00 (0.57 to 1.00)

0.64 (0.48 to 0.78)

51% (40% to 61%)

Not estimable

49 (39 to 60)

Not estimable

High

High

High

CT

6 (326)

0.72 (0.50 to 0.87)

0.92 (0.81 to 0.97)

78% (57% to 91%)

10% (5% to 18%)

22 (9 to 43)

10 (5 to 18)

Unclear or high

High

High

MRI

1 (32)

0.75 (0.30 to 0.95)

0.93 (0.77 to 0.98)

80% (48% to 94%)

9% (2% to 35%)

20 (6 to 52)

9 (2 to 35)

High

High

High

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) (median pre‐test probability: 45%)

EUS

4 (196)

0.86 (0.74 to 0.92)

0.91 (0.83 to 0.96)

89% (80% to 94%)

11% (7% to 19%)

11 (6 to 20)

11 (7 to 19)

High

High

High

EUS‐FNA (cytology)

3 (310)

0.47 (0.24 to 0.70)

0.91 (0.32 to 1.00)

81% (19% to 99%)

32% (22% to 45%)

19 (1 to 81)

32 (22 to 45)

Unclear or high

High

High

EUS‐FNA (CEA > 200 ng/mL)

3 (160)

0.58 (0.28 to 0.83)

0.51 (0.19 to 0.81)

49% (28% to 70%)

40% (19% to 65%)

51 (30 to 72)

40 (19 to 65)

High

High

High

EUS‐FNA (CA 19‐9 > 1000 U/mL)

1 (41)

0.90 (0.60 to 0.98)

0.42 (0.26 to 0.59)

56% (47% to 65%)

16% (3% to 57%)

44 (35 to 53)

16 (3 to 57)

High

High

High

EUS‐FNA (CEA > 692.8 ng/mL)

1 (20)

0.80 (0.49 to 0.94)

0.90 (0.60 to 0.98)

87% (50% to 98%)

15% (5% to 39%)

13 (2 to 50)

15 (5 to 39)

Unclear

High

High

PET (SUVmax > 2 to 2.5)

4 (124)

0.90 (0.79 to 0.96)

0.94 (0.81 to 0.99)

93% (78% to 98%)

8% (4% to 16%)

7 (2 to 22)

8 (4 to 16)

High

High

High

CT

3 (139)

0.87 (0.00 to 1.00)

0.96 (0.00 to 1.00)

95% (0% to 100%)

10% (0% to 100%)

5 (0 to 100)

10 (0 to 100)

Unclear or high

High

High

MRI

3 (189)

0.69 (0.44 to 0.86)

0.93 (0.43 to 1.00)

89% (35% to 99%)

21% (12% to 36%)

11 (1 to 65)

21 (12 to 36)

High

High

High

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) (median pre‐test probability: 21%)

EUS

1 (51)

0.77 (0.50 to 0.92)

0.89 (0.76 to 0.96)

67% (43% to 84%)

7% (3% to 16%)

33 (16 to 57)

7 (3 to 16)

Unclear

High

High

CT

1 (46)

0.50 (0.22 to 0.78)

0.95 (0.83 to 0.99)

72% (36% to 92%)

13% (7% to 22%)

28 (8 to 64)

13 (7 to 22)

High

High

High

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) (median pre‐test probability: 59%)

CT

3 (106)

0.83 (0.68 to 0.92)

0.83 (0.64 to 0.93)

89% (56% to 98%)

33% (18% to 52%)

11 (2 to 44)

33 (18 to 52)

High

High

High

MRI

2 (71)

0.80 (0.58 to 0.92)

0.81 (0.53 to 0.95)

86% (67% to 95%)

27% (13% to 47%)

14 (5 to 33)

27 (13 to 47)

High

High

High

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign (median pre‐test probability: 43%)

EUS

1 (70)

0.97 (0.83 to 0.99)

0.40 (0.26 to 0.55)

55% (48% to 61%)

6% (1% to 31%)

45 (39 to 52)

6 (1 to 31)

High

High

High

*Post‐test probability was calculated at the median pre‐test probability.

Abbreviations:

CA 19‐9: carbohydrate antigen 19‐9
CEA: carcinoembryonic antigen
CI: confidence interval
CT: computed tomography
EUS: endoscopic ultrasound
FNA: fine‐needle aspiration
MRI: magnetic resonance imaging
PET: positron emission tomography
SUVmax: maximum standardised uptake values

Figuras y tablas -
Summary of findings Imaging modalities for characterising focal pancreatic lesions
Table 1. QUADAS‐2 classification

Domain

Signalling question

Signalling question

Signalling question

Risk of bias

Concerns for
applicability

1: Patient sampling

Was a consecutive or random sample of patients enrolled?

Was a case‐control design avoided?

Did the study avoid inappropriate exclusions?

Could the selection of participants have introduced bias?

Are there concerns that the included participants and setting do not match the review question?

Yes: all consecutive patients or random sample of patients with focal pancreatic lesions were enrolled

No: selected patients were enrolled

Unclear: this was not clear from the report

Yes: case‐control design was avoided

No: case‐control design was not avoided

Unclear: this was not clear from the report

Yes: the study avoided inappropriate exclusions (i.e. difficult‐to‐diagnose patients)

No: the study excluded patients inappropriately

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 signalling question

Low concern: the selected participants represent the patients in whom the tests will be used in clinical practice (please see diagnostic pathway (Figure 1))

High concern: there is high concern that participant selection was performed in such a way that the included participants did not represent the patients in whom the tests will be used in clinical practice

2: Index test(s)

Were the index test results interpreted without knowledge of the results of the reference standard?

If a threshold was used, was it prespecified?

Could the conduct or interpretation of the index test have introduced bias?

Are there concerns that the index test, its conduct, or its interpretation differ from the review question?

Yes: index test results were interpreted without knowledge of the results of the reference standard

No: index test results were interpreted with knowledge of the results of the reference standard

Unclear: this was not clear from the report

Yes: if the criteria for a positive test were prespecified

No: if the criteria for a positive test were not prespecified

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 of the 2 signalling questions

High concern: there is high concern that the conduct or interpretation of the index test differs from the way it is likely to be used in clinical practice

Low concern: there is low concern that the conduct or interpretation of the index test differs from the way it is likely to be used in clinical practice

3: Target condition and reference standard(s)

Is the reference standard likely to classify the target condition correctly?

Were the reference standard results interpreted without knowledge of the results of the index tests?

Could the reference standard, its conduct, or its interpretation have introduced bias?

Are there concerns that the target condition as defined by the reference standard does not match the review question?

Yes: histopathological examination of the entire lesion by surgical resection

No: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

Unclear: this was not clear from the report. Such studies will be excluded

Yes: reference standard results were interpreted without knowledge of the results of the index test

No: reference standard results were interpreted with knowledge of the results of the index test

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 of the 2 signalling questions

Low concern: histopathological examination of the entire lesion by surgical resection

High concern: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

4: Flow and timing

Was there an appropriate interval between index test and reference standard?

Did all patients receive the same reference standard?

Were all patients included in the analysis?

Could the patient flow have introduced bias?

Yes:

histopathological examination of the entire lesion (gold standard) ‐ performed within 2 months (chosen arbitrarily).

Histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) performed within 2 months and clinical follow‐up (including sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months

No:

the histopathological examination was performed beyond 2 months of the index tests.

The clinical follow‐up (including sequential follow‐up imaging) was performed less than 6 months after the index test, because some tumours may be slow‐growing

Unclear: this was not clear from the report

Yes: histopathological examination of the entire lesion by surgical resection

No: histopathological examination (irrespective of how the tissues were obtained for histopathological examination) in patients with positive test (for cancerous or precancerous lesions) and clinical follow‐up by a doctor (with or without sequential follow‐up with imaging) of all patients with negative test for a period of at least 6 months and for a maximum period of 24 months

Unclear: this was not clear from the report. Such studies will be excluded

Yes: all patients meeting the selection criteria (selected participants) were included in the analysis, or data on all of the selected participants were available so that a 2 x 2 table including all selected participants could be constructed

No: not all patients meeting the selection criteria were included in the analysis, or the 2 x 2 table could not be constructed using data on all selected participants

Unclear: this was not clear from the report

Low risk: 'yes' for all signalling questions

High risk: 'no' or 'unclear' for at least 1 signalling question

Figuras y tablas -
Table 1. QUADAS‐2 classification
Table 2. Summary sensitivity and specificity of different tests for different target conditions

Comparison

Name of test

Sensitivity

Specificity

Cancerous versus benign or precancerous

EUS‐FNA
(cytology)

0.79 (95% CI 0.60 to 0.91)

1.00 (95% CI 0.85 to 1.00)

Cancerous versus benign or precancerous

EUS‐FNA
(CEA > 500 ng/mL)

0.93 (95% CI 0.70 to 0.99)

0.33 (95% CI 0.12 to 0.65)

Cancerous versus benign or precancerous

PET
(criteria unspecified)

0.85 (95% CI 0.73 to 0.92)

0.91 (95% CI 0.72 to 0.97)

Cancerous versus benign

EUS

0.95 (95% CI 0.84 to 0.99)

0.53 (95% CI 0.31 to 0.74)

Cancerous versus benign

EUS‐FNA
(cytology)

0.79 (95% CI 0.07 to 1.00)

1.00 (95% CI 0.91 to 1.00)

Cancerous versus benign

PET (criteria unspecified)

0.92 (95% CI 0.80 to 0.97)

0.65 (95% CI 0.39 to 0.85)

Cancerous versus benign

PET (SUVmax > 3.5)

0.96 (95% CI 0.87 to 0.99)

0.62 (95% CI 0.43 to 0.78)

Cancerous versus benign

CT

0.98 (95% CI 0.00 to 1.00)

0.76 (95% CI 0.02 to 1.00)

Cancerous versus benign

MRI

0.80 (95% CI 0.58 to 0.92)

0.89 (95% CI 0.57 to 0.98)

Precancerous or cancerous versus benign

EUS

0.92 (95% CI 0.74 to 0.98)

0.60 (95% CI 0.31 to 0.83)

Precancerous or cancerous versus benign

EUS‐FNA (cytology)

0.73 (95% CI 0.01 to 1.00)

0.94 (95% CI 0.15 to 1.00)

Precancerous or cancerous versus benign

EUS‐FNA
(CEA > 50 ng/mL)

0.29 (95% CI 0.08 to 0.64)

0.25 (95% CI 0.05 to 0.70)

Precancerous or cancerous versus benign

PET
(SUVmax 2.4)

0.94 (95% CI 0.74 to 0.99)

0.93 (95% CI 0.69 to 0.99)

Precancerous or cancerous versus benign

CT

0.62 (95% CI 0.45 to 0.76)

0.64 (95% CI 0.39 to 0.84)

Precancerous or cancerous versus benign

MRI

0.93 (95% CI 0.69 to 0.99)

0.85 (95% CI 0.58 to 0.96)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS

0.78 (95% CI 0.45 to 0.94)

0.91 (95% CI 0.61 to 0.98)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS‐FNA
(cytology)

0.66 (95% CI 0.03 to 0.99)

0.92 (95% CI 0.73 to 0.98)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

EUS‐FNA
(CEA > 200 ng/mL)

1.00 (95% CI 0.57 to 1.00)

0.64 (95% CI 0.48 to 0.78)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

CT

0.72 (95% CI 0.50 to 0.87)

0.92 (95% CI 0.81 to 0.97)

Cancerous (invasive carcinoma) versus precancerous (dysplasia)

MRI

0.75 (95% CI 0.30 to 0.95)

0.93 (95% CI 0.77 to 0.98)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS

0.86 (95% CI 0.74 to 0.92)

0.91 (95% CI 0.83 to 0.96)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(cytology)

0.47 (95% CI 0.24 to 0.70)

0.91 (95% CI 0.32 to 1.00)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CEA > 200 ng/mL)

0.58 (95% CI 0.28 to 0.83)

0.51 (95% CI 0.19 to 0.81)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CA 19‐9 > 1000 U/mL)

0.90 (95% CI 0.60 to 0.98)

0.42 (95% CI 0.26 to 0.59)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

EUS‐FNA
(CEA > 692.8 ng/mL)

0.80 (95% CI 0.49 to 0.94)

0.90 (95% CI 0.60 to 0.98)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

PET
(SUVmax > 2 to 2.5)

0.90 (95% CI 0.79 to 0.96)

0.94 (95% CI 0.81 to 0.99)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

CT

0.87 (95% CI 0.00 to 1.00)

0.96 (95% CI 0.00 to 1.00)

Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia)

MRI

0.69 (95% CI 0.44 to 0.86)

0.93 (95% CI 0.43 to 1.00)

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia)

EUS

0.77 (95% CI 0.50 to 0.92)

0.89 (95% CI 0.76 to 0.96)

Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia)

CT

0.50 (95% CI 0.22 to 0.78)

0.95 (95% CI 0.83 to 0.99)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)

CT

0.83 (95% CI 0.68 to 0.92)

0.83 (95% CI 0.64 to 0.93)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia)

MRI

0.80 (95% CI 0.58 to 0.92)

0.81 (95% CI 0.53 to 0.95)

Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) or benign

EUS

0.97 (95% CI 0.83 to 0.99)

0.40 (95% CI 0.26 to 0.55)

Cystic lesion subgroup analysis

Cancerous versus benign ‐ EUS‐FNA (cytology)

0.50 (95% CI 0.19 to 0.81)

1.00 (95% CI 0.84 to 1.00)

Cystic lesion subgroup analysis

Cancerous versus benign ‐ PET
(SUVmax > 3.5)

0.96 (95% CI 0.87 to 0.99)

0.62 (95% CI 0.43 to 0.78)

Cystic lesion subgroup analysis

Precancerous or cancerous versus benign ‐ EUS‐FNA
(cytology)

0.43 (95% CI 0.19 to 0.71)

1.00 (95% CI 0.74 to 1.00)

CA 19‐9: carbohydrate antigen 19‐9
CEA: carcinoembryonic antigen
CI: confidence interval
CT: computed tomography
EUS: endoscopic ultrasound
FNA: fine‐needle aspiration
MRI: magnetic resonance imaging
PET: positron emission tomography
SUVmax: maximum standardised uptake values

Figuras y tablas -
Table 2. Summary sensitivity and specificity of different tests for different target conditions
Table Tests. Data tables by test

Test

No. of studies

No. of participants

1 Cancerous versus benign or precancerous ‐ EUS‐FNA (cytology) Show forest plot

1

45

2 Cancerous versus benign or precancerous ‐ EUS‐FNA (CEA > 500 ng/mL) Show forest plot

1

24

3 Cancerous versus benign or precancerous ‐ PET Show forest plot

1

76

4 Cancerous versus benign ‐ EUS Show forest plot

2

133

5 Cancerous versus benign ‐ EUS‐FNA Show forest plot

3

147

6 Cancerous versus benign ‐ PET Show forest plot

3

99

7 Cancerous versus benign ‐ PET (SUVmax > 3.5) Show forest plot

1

80

8 Cancerous versus benign ‐ CT Show forest plot

2

123

9 Cancerous versus benign ‐ MRI Show forest plot

1

29

10 Precancerous or cancerous versus benign ‐ EUS Show forest plot

1

34

11 Precancerous or cancerous versus benign ‐ EUS‐FNA Show forest plot

3

52

12 Precancerous or cancerous versus benign ‐ EUS‐FNA (CEA > 50 ng/mL) Show forest plot

1

11

13 Precancerous or cancerous versus benign ‐ PET (SUVmax > 2.4) Show forest plot

1

32

14 Precancerous or cancerous versus benign ‐ CT Show forest plot

1

48

15 Precancerous or cancerous versus benign ‐ MRI Show forest plot

1

27

16 Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS Show forest plot

5

156

17 Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA Show forest plot

3

158

18 Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL) Show forest plot

1

41

19 Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ CT Show forest plot

6

326

20 Cancerous (invasive carcinoma) versus precancerous (dysplasia) ‐ MRI Show forest plot

1

32

21 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS Show forest plot

4

196

22 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA Show forest plot

3

310

23 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 200 ng/mL) Show forest plot

3

160

24 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CA 19‐9 > 1000 U/mL) Show forest plot

1

41

25 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ EUS‐FNA (CEA > 692.8 ng/mL) Show forest plot

1

20

26 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ PET (SUVmax 2 to 2.5) Show forest plot

4

124

27 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ CT Show forest plot

3

139

28 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) ‐ MRI Show forest plot

3

189

29 Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ EUS Show forest plot

1

51

30 Cancerous (invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT Show forest plot

1

46

31 Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ CT Show forest plot

3

106

32 Precancerous or cancerous (intermediate‐ or high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐grade dysplasia) ‐ MRI Show forest plot

2

71

33 Cancerous (high‐grade dysplasia or invasive carcinoma) versus precancerous (low‐ or intermediate‐grade dysplasia) or benign ‐ EUS Show forest plot

1

70

34 Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ EUS‐FNA Show forest plot

1

26

35 Cystic lesion subgroup analysis ‐ Cancerous versus benign ‐ PET Show forest plot

1

80

36 Cystic lesion subgroup analysis ‐ Precancerous or cancerous versus benign ‐ EUS‐FNA Show forest plot

2

34

Figuras y tablas -
Table Tests. Data tables by test