Types of studies

We will include prospective or retrospective consecutive series studies reporting on the diagnostic accuracy of SLNB. We will exclude diagnostic case‐control studies. We will also exclude studies reporting insufficient data for accurate identification of the target population or for the construction of a 2 x 2 table.

Participants

Patients meeting the following criteria will be considered for this review:

1. squamous cell carcinoma of the oral cavity and/or oropharynx;

2. size of primary tumour < 4 cm;

3. no palpable neck nodes (clinically node‐negative);

4. radiologically node‐negative (if any pre‐surgical imaging has been done).

Index tests

SLNB will be the index test for this study. A sentinel node(s) is defined as a lymph node(s) sending a radioactive signal with an activity more than 10‐fold above background radiation level or a node(s) that appears blue intra‐operatively. The sentinel node(s) identified will be removed and submitted for standard histological examination with H&E staining. If the sentinel node(s) is found to be malignant by histological examination, it is defined as a positive sentinel node. If a sentinel node(s) is detected but histologically does not show any malignancy, it is defined as a negative node(s). If the sentinel node(s) is negative on H&E staining but positive on immunohistochemistry or PCR, it is considered positive. If the sentinel node(s) cannot be identified by tracer substance or blue dye, then it is defined as 'failure to detect sentinel node(s)'.

Comparator tests

No comparator test will be considered in this review.

Target conditions

Lymph node involvement in early‐stage (cT1‐T2), clinically and/or radiologically node‐negative (cN0) oral and/or oropharyngeal cancer.

Reference standards

Completion neck dissection (either selective or comprehensive neck dissection) followed by standard histological assessment of the surgical specimen will be the only reference standard. Selective neck dissection means the removal of selected neck node levels (e.g. supraomohyoid neck dissection wherein level I‐III lymph nodes are dissected) whereas comprehensive neck dissection refers to the removal of neck nodes at all levels (e.g. modified neck dissection wherein levels I‐V are dissected). The removed surgical pathology specimen is submitted for standard histological examination (H&E staining). If any of the removed nodes (either at SLNB or completion neck dissection) show metastasis, the reference standard will be considered positive.

Electronic searches

We will identify published, unpublished and ongoing studies by searching the following databases from their inception: the Cochrane Register of Diagnostic Accuracy Studies; MEDION; ARIF (1996 onwards); PubMed (1940 onwards); EMBASE, via Ovid (1947 onwards); CINAHL, via EBSCO; LILACS; KoreaMed (1997 onwards); IndMed (1985 onwards); PakMediNet; CAB Abstracts; Web of Science; BIOSIS Previews; Google Scholar and Google. We will model subject strategies for databases on the search strategy designed for PubMed (see Appendix 1).

Searching other resources

We will examine the reference lists of the retrieved articles. We will search The Cochrane Library, TRIP database and the National Guidelines Clearing House to retrieve evidence‐based resources relevant to this systematic review, so that we can scan their reference lists for additional studies.

Selection of studies

We will download all titles and abstracts retrieved by electronic searching to the reference management database Endnote. We will remove duplicates and two review authors (SK, TG) will examine the remaining references independently. We will exclude those studies which clearly do not meet the inclusion criteria, and obtain copies of the full text of potentially relevant references. Two review authors (SN, TG) will assess the eligibility of retrieved papers independently. We will resolve disagreements if possible between SN and TG and, if necessary, involve a third review author (PC or SK). We will document the reasons for exclusion.

Data extraction and management

We will extract data on the following items.

• Author, year of publication and journal (including language)

• Country

• Settings

• Inclusion and exclusion criteria

• Study design

• Study population

• Number of patients

• Imaging prior to the index test, if any

• Staging details of oral/oropharyngeal cancer

• Biopsy/histopathology from primary site

• Reference standard and performance of the reference standard

  • Type of neck dissection

  • Lymph node(s) yield (number and levels)

  • Histological assessment methods

• Performance of the index test (SLNB)

  • Radioactive tracer or blue dye, or both

  • Technique of injection of radioactive substance and blue dye

  • Timing of injection

  • Amount of tracer used and location of injection

  • Exact technique of detection of tracer substance

  • Histological assessment methods

  • Failed test if any

• Reporting of results

• QUADAS items

• Data for 2 x 2 table (true positive, true negative, false positive, false negative)

Two review authors (TG, SN) will abstract data independently onto a data abstraction spreadsheet specially designed for the review. We will resolve differences between review authors by discussion or by appeal to a third review author (SK or PC) if necessary. We will use two primary diagnostic studies to pilot the data abstraction spreadsheet (including the quality assessment).

Assessment of methodological quality

We will use the QUADAS‐2 quality assessment tool (Whiting 2011) to assess quality. Two authors (TG and SK) will independently assess the quality of each study. Any disagreement will be resolved by discussion and consensus with all review authors. We will report the results in a tabular and graphical form. We will carry out quality assessment of the studies according to the phases/domains outlined in Appendix 2.

Statistical analysis and data synthesis

We will input data from 2 x 2 tables and calculate sensitivity, specificity, positive predictive value and negative predictive value with 95% confidence intervals (CI) for all studies. We will use RevMan 5.1 to create coupled forest plots to evaluate the variation in the estimates of sensitivity and specificity between studies (RevMan 2011). We will plot the results of individual studies of diagnostic test accuracy in a summary receiver operator characteristic (SROC) space. As there is no threshold, we will use the bivariate method to provide summary estimates of sensitivity and specificity (Reitsma 2005). We will carry out analysis using the Metandi module of STATA v11.

Investigations of heterogeneity

We will investigate heterogeneity by visual examination of forest plots and ROC plots of sensitivities and specificities. We will analyse the sources of heterogeneity as covariates in meta‐regression analysis. The categorical covariates that we will analyse include:

• year of publication (before or after 31 December 2005);

• geographical location (North America, Europe, Asia Pacific);

• sentinel node procedure (use of radioactive tracer, blue dye or a combination of the two);

• type of neck dissection (selective, comprehensive);

• histological methods of assessment (H&E, immunohistochemistry, PCR).

Sensitivity analyses

We will carry out sensitivity analysis by including studies with a rating of low risk of bias for patient selection, index test, reference standard, and flow and timing. The studies with high or unclear risk of bias will be omitted.

Assessment of reporting bias

As stated previously we will try to account for withdrawals and drop‐outs from individual studies (if not described) by contacting the corresponding author of the respective study (to minimise reporting bias).

To minimise publication bias, we will try to retrieve unpublished data on the accuracy of SLNB in OOSCC by contacting leading experts in the field.

We will produce an effective sample size funnel plot and associated regression test of asymmetry to ascertain publication bias (Deeks 2005) within the limitations of such assessment.