Results of the search

No new studies were identified for inclusion in this 2021 update. See Figure 1.

Figure 1

---

---

Study flow diagram.

We identified four additional reports relating to the previously included study Robin 2016 (Robin 2017a; Robin 2017b; Robin 2018; Robin 2020); and one additional report each for Carrier 2015 (NCT00773448) and Prandoni 2016 (NCT00361647). One new study was assessed as excluded (Kraaijpoel 2018). We identified two new ongoing studies (EUCTR2018‐003958‐25‐ES; EUCTR2020‐002210‐41‐FR).

Included studies

Four studies fulfilled the eligibility criteria for inclusion in this review (Carrier 2015; Piccioli 2004b; Prandoni 2016; Robin 2016).

See Characteristics of included studies table.

The first study was a randomised multicentre study of 201 apparently cancer‐free people with acute unprovoked VTE (Piccioli 2004b). Extensive investigations for occult malignant disease were compared with testing at the physician's discretion. Ninety‐nine participants were randomised to the extensive screening group and 102 were randomised to the control group. Participants in the extensive investigations group were offered ultrasound and CT scans of the abdomen and pelvis, double contrast barium swallowing, colonoscopy or sigmoidoscopy followed by a barium enema, haemoccult test, sputum cytology and tumour markers including carcinoembryonic antigen (CEA), alpha‐fetoprotein (α‐FP) and CA125. Women also underwent mammography and Papanicolaou (Pap) smears while men had transabdominal ultrasound of the prostate and a total prostate‐specific antigen (PSA) test. All tests were completed within a four‐week period from the diagnosis of VTE. Participants in the control group were investigated at the physician's discretion. If the investigations suggested the presence of a malignant process, further investigations were performed according to current standards. Participants were followed up at 3, 12 and 24 months following the diagnosis of VTE. The primary outcome was cancer‐related morbidity, defined as death due to a malignant disease itself, or death due to complications of diagnostic or surgical procedures performed to diagnose or treat cancer. A secondary outcome of this study consisted of the cluster of cancer‐related mortality and documented residual malignancy or recurrent malignancy at 24 months. The authors also measured the frequency of an underlying cancer diagnosis including type and stage as well as mean time to cancer diagnosis.

The second study was a randomised study in which 195 participants with a first episode of unprovoked VTE were randomised to extensive investigations (98 participants) or a discretionary diagnostic approach excluding CT scans (97 participants) (Prandoni 2016). Extensive investigations comprised a mandatory CT scan of the thorax, abdomen and pelvis together with faecal haemoccult testing or any test at physician's discretion according to good clinical practice. Participants allocated to the discretionary diagnostic approach or personalised strategy underwent additional testing based on physicians' judgements and participants' preferences, including a 'no‐further testing' option. Participants were followed at 3, 6, 12 and 24 months to document the incidence of newly discovered cancer and cancer‐related mortality. The primary outcomes were cancer‐related mortality (defined as death due to malignancy, or death due to the complications of the diagnostic or surgical procedures performed to diagnose or treat cancer) and incidence of newly discovered cancer. The secondary outcomes were cancer stage, using the tumours‐nodes‐metastases classification, at which tumours were diagnosed in the two study groups and the incidence of cancer‐related mortality in the two randomisation groups,

The third study was an open‐label randomised study in which 854 participants with a first episode of unprovoked VTE were randomised to limited occult‐cancer screening plus CT scanning of the abdomen and pelvis (423 participants) or limited occult‐cancer screening alone (431 participants) (Carrier 2015). The limited occult‐cancer screening comprised complete history and physical examination, measurement of complete blood counts and serum electrolyte and creatinine levels, liver‐function testing and chest radiography. Sex‐specific screening was conducted if it had not been performed in the previous year. A breast examination, mammography, or both were performed in women over 50 years of age and Pap testing and a pelvic examination were performed in women 18 to 70 years of age who had ever been sexually active. A prostate examination, PSA test, or both were performed in men over 40 years of age. The additional CT investigations comprised a virtual colonoscopy and gastroscopy, biphasic enhanced CT of the liver, parenchymal pancreatography and uniphasic enhanced CT of the distended bladder. Participants were followed up for one year to document the incidence of newly diagnosed cancer, type of cancer diagnosed, one‐year cancer‐related mortality, one‐year overall mortality, time to cancer diagnosis and incidence of recurrent VTE.

The fourth study was an open‐label randomised study in which 394 participants with a first episode of unprovoked VTE were randomised to a limited screening strategy (197 participants) or a screening strategy consisting of the limited strategy plus an 18‐fluorodeoxyglucose (¹⁸F‐FDG) PET/CT scan of the chest, abdomen and pelvis (197 participants) (Robin 2016). The limited screening comprised medical history taking, physical examination, routine laboratory tests (including complete blood count, erythrocyte sedimentation rate or C‐reactive protein, aminotransferases, alkaline phosphatase and calcium), chest radiograph, and recommended age‐specific and sex‐specific cancer screening tests (i.e. PSA in men older than 50 years, mammography in women older than 50 years and Pap smear in all women). Participants were followed up for two years to determine the proportion of people with a cancer diagnosis in each group after the initial screening assessment.

Excluded studies

We identified one new excluded study (Kraaijpoel 2018). See Characteristics of excluded studies.

Studies that were not randomised controlled trials were deemed not relevant and therefore not listed as an excluded study.

Ongoing studies

We identified two new ongoing studies (EUCTR2018‐003958‐25‐ES; EUCTR2020‐002210‐41‐FR). See Characteristics of ongoing studies.

Allocation

All four studies were randomised. Prandoni 2016 used envelopes. Piccioli 2004b used a Zelen approach (participants are randomised to either the treatment or control group before giving informed consent), Carrier 2015 used random number tables and Robin 2016 used a computer random number generator. Therefore, these four studies were judged at low risk of selection bias. In terms of concealing the allocation of treatment, Piccioli 2004b performed randomisation centrally and Carrier 2015 and Robin 2016 used a central web‐based randomisation system and, therefore, these three were judged at low risk of selection bias. The Prandoni 2016 study used serially numbered, opaque, sealed envelopes to conceal allocation and was, therefore, judged at low risk of selection bias too.

Blinding

As the study groups in all four trials were randomised to extensive screening or no further testing, it was impossible to blind participants and study personnel. However, we believe it was unlikely that the lack of blinding would have affected the outcome and, therefore, all studies were judged at low risk of performance bias.

In Piccioli 2004b, the physician at the follow‐up examination was unaware of the allocation of participants and, therefore, detection bias for outcome assessors was low. Similarly, for Carrier 2015 a blinded adjudication committee reviewed all suspected outcome events and, therefore, the risk of detection bias was low. Robin 2016 did not blind outcome assessors to treatment allocation and was, therefore, judged at high risk of detection bias. In Prandoni 2016, investigators performing the follow‐up visits were blinded to the participants' randomisation group and the study was, therefore, judged at low risk of detection bias.

Incomplete outcome data

The treatment groups in all four studies were well‐balanced with respect to baseline characteristics, completion of the study protocol and discontinuation of treatment. Furthermore, all missing data were accounted for and reported. Therefore, all four studies were judged at low risk of attrition bias (Carrier 2015; Piccioli 2004b; Prandoni 2016; Robin 2016).

Selective reporting

All four studies clearly prespecified all primary and secondary outcomes and data on all outcomes were reported (low risk of reporting bias) (Carrier 2015; Piccioli 2004b; Prandoni 2016; Robin 2016).

Other potential sources of bias

Two studies were deemed at low risk (Carrier 2015; Robin 2016), and two studies were deemed to be at high risk of other bias (Piccioli 2004b; Prandoni 2016). The study by Piccioli 2004b was terminated early after the inclusion of only 201 participants after five years for several reasons. First, only five of the more than 40 potential participating centres could contribute participants to the study. Second, some medical ethics committees rejected the protocol because of the absence of screening for occult cancer in the control group, other centres could not start because the proposed extensive screening was judged to be unethical. Finally, the identification of cancer at an apparent early stage in the extensive screening group led to an increasing tendency among physicians in the participating hospitals to initiate screening for cancer in the control participants. The study by Prandoni 2016 was judged at high risk of bias as results of an interim analysis, scheduled after the inclusion of approximately half of the planned sample size, showed no appreciable advantage of the CT‐based strategy over the control strategy for detection of occult cancers. In addition, there was a low recruitment rate, so the study promoters decided to terminate the study early.

Extensive tests versus tests at the physician's discretion

Two studies assessed the effect of testing for cancer versus clinically indicated tests only (Piccioli 2004b; Prandoni 2016).

Both studies measured the primary outcome cancer‐related mortality. In Piccioli 2004b, 2/99 participants in the extensive testing group died of cancer compared to 4/102 in the group who underwent tests at the physician's discretion (OR 0.51, 95% CI 0.09 to 2.82). In Prandoni 2016, 2/98 participants who underwent extensive testing and 4/97 participants who underwent tests at the physician's discretion died of cancer (OR 0.48, 95% CI 0.09 to 2.71). Meta‐analysis showed an OR of 0.49 (95% CI 0.15 to 1.67; low‐certainty evidence) in favour of extensive testing, which did not reach statistical significance (P = 0.26) (Analysis 1.1).

However, neither Piccioli 2004b nor Prandoni 2016 measured the review's other primary outcomes of all‐cause mortality and VTE‐related mortality, or the secondary outcomes VTE‐related morbidity, complications of anticoagulation, adverse effects of cancer tests, participant satisfaction and quality of life.

Piccioli 2004b looked at the location of the malignancy and found no clear difference in the incidence of any particular cancer between participants who underwent extensive tests and participants who were tested at the physician's discretion (lung: OR 2.08, 95% CI 0.19 to 23.34; bladder: OR 2.08, 95% CI 0.19 to 23.34; stomach: OR 1.03, 95% CI 0.06 to 16.71; kidney: OR 3.12, 95% CI 0.13 to 77.55; adrenal gland: OR 3.12, 95% CI 0.13 to 77.55; liver: OR 3.12, 95% CI 0.13 to 77.55; uterus: OR 3.12, 95% CI 0.13 to 77.55; breast: OR 1.03, 95% CI 0.06 to 16.71; ovary: OR 3.12, 95% CI 0.13 to 77.55; colon: OR 0.51, 95% CI 0.05 to 5.72; prostate: OR 0.51, 95% CI 0.05 to 5.72; pancreas: OR 0.20, 95% CI 0.01 to 4.26) (Analysis 1.2).

Piccioli 2004b compared the characteristics of the diagnosed cancer by assessing the proportion of early‐stage malignancies, defined as T1 or T2 without locoregional or distant metastases (N0 M0). Overall, malignancies were less advanced in participants who had undergone extensive testing. In total, 9/13 participants diagnosed with cancer in the tested group had a T1‐ or T2‐stage malignancy without locoregional or distant metastases compared to 2/10 participants diagnosed with cancer in the control group (OR 5.00, 95% CI 1.05 to 23.76; P = 0.04; low‐certainty evidence). There was no difference in detection of advanced stages between groups: one participant in the tested group had stage T3 compared with four participants in the control group (OR 0.25, 95% CI 0.03 to 2.28; P = 0.22; low‐certainty evidence) (Analysis 1.3).

One study measured time to cancer diagnosis (measured from the time of diagnosis of VTE) (Piccioli 2004b), reported as a mean of one month in tested participants compared to 11.6 months in participants who were tested at the physician's discretion (P < 0.001). Standard deviations for the means were not given. We attempted to contact the author for these data but received no response.

Both studies measured the frequency of an underlying cancer diagnosis. Piccioli 2004b detected underlying cancer in 13/99 participants who underwent extensive testing, whereas it became symptomatic in 10/102 control participants (OR 1.39, 95% CI 0.58 to 3.34). Prandoni 2016 detected cancer in 2/98 participants who had further tests and it became apparent in 2/97 participants who were tested at the physician's discretion (OR 0.99, 95% CI 0.14 to 7.17). The combined incidence of an underlying cancer diagnosis was 15/197 in the tested group and 12/199 in the control group (OR 1.32, 95% CI 0.59 to 2.93; low‐certainty evidence) (Analysis 1.4). Therefore, after 24 months of follow‐up, the incidence of cancer was no different in the tested and control groups.

Standard testing plus PET/CT scanning versus standard testing alone

Two studies assessed the effect of standard testing plus PET/CT scanning versus standard testing alone (Carrier 2015; Robin 2016).

Both studies measured the primary outcome all‐cause mortality. In the standard testing plus CT scanning group, 11/620 participants died during follow‐up compared to 9/628 participants who received standard testing alone (OR 1.22, 95% CI 0.49 to 3.04; moderate‐certainty evidence) (Analysis 2.1).

Carrier 2015 and Robin 2016 also measured cancer‐related mortality and reported an incidence of 6/620 participants with standard testing plus CT scanning compared to 11/628 participants who received standard testing alone (OR 0.55, 95% CI 0.20 to 1.52; moderate‐certainty evidence) (Analysis 2.2).

The study by Carrier 2015 measured VTE‐related morbidity. The incidence of recurrent VTE was 14/423 participants who underwent standard testing plus CT scanning compared to 14/431 participants who had standard testing alone (OR 1.02, 95% CI 0.48 to 2.17; moderate‐certainty evidence) (Analysis 2.3).

Carrier 2015 and Robin 2016 also looked at the location of the malignancy and found no clear difference in the incidence of any particular cancer between the two groups (acute leukaemia: OR 1.62, 95% CI 0.20 to 13.22; gynaecological: OR 2.39, 95% CI 0.43 to 13.36; melanoma: OR 1.02, 95% CI 0.06 to 16.34; colorectal: OR 0.43, 95% CI 0.08 to 2.40; prostate: OR 2.52, 95% CI 0.48 to 13.12; pancreatic: OR 4.81, 95% CI 0.55 to 42.48; cholangiocarcinoma: OR 0.51, 95% CI 0.05 to 5.63; lymphoma: OR 0.74, 95% CI 0.09 to 5.83; breast: OR 0.20, 95% CI 0.01 to 4.24; urological: OR 0.62, 95% CI 0.03 to 12.32; liver: OR 0.33, 95% CI 0.01 to 8.19; head and neck: OR 3.02, 95% CI 0.12 to 74.47; lung: OR 3.02, 95% CI 0.12 to 74.47; unknown primary origin: OR 0.34, 95% CI 0.01 to 8.34) (Analysis 2.4).

Robin 2016 also measured the stage of cancer. Early‐stage cancer was detected in 7/197 participants who underwent standard testing plus CT scanning compared to 4/197 participants who underwent standard testing alone (OR 1.78, 95% CI 0.51 to 6.17; low‐certainty evidence), while advanced‐stage cancer was detected in two participants in each group (OR 1.00, 95% CI 0.14 to 7.17; low‐certainty evidence) (Analysis 2.5).

Time to cancer diagnosis was 4.0 months in the standard testing plus PET/CT group and 4.2 months in the standard testing group in one study (P = 0.88) (Carrier 2015). However, standard deviations for these means were not given. We attempted to contact the author for these data but received no response. Robin 2016 did not measure time to cancer diagnosis.

Carrier 2015 and Robin 2016 measured the frequency of underlying cancer diagnosis. Underlying cancer was detected in 30/620 participants who underwent standard testing plus CT scanning compared to 18/628 participants who underwent standard testing alone (OR 1.71, 95% CI 0.91 to 3.20; moderate‐certainty evidence) (Analysis 2.6).

The studies by Carrier 2015 and Robin 2016 did not measure the other review outcomes of VTE‐related mortality, complications of anticoagulation, adverse effects of cancer tests, participant satisfaction or quality of life.