Immunohistochemistry (IHC) is a cornerstone of modern oncologic pathology, playing a critical role in tumor diagnosis, classification, prognostication, and therapeutic decision-making. By enabling visualization of protein expression within preserved tissue architecture, IHC bridges the gap between morphology and molecular biology and has become indispensable in routine diagnostic workflows across virtually all malignancies [1,2]. Markers such as cytokeratins, vimentin, leukocyte common antigen, hormone receptors, HER2, mismatch repair proteins, and immune checkpoint markers are routinely employed to determine tumor lineage, subtype, and predictive biomarker status [3–5]. Despite its widespread use, the diagnostic performance of IHC markers is not uniform. Reported sensitivity and specificity vary substantially across studies, tumor types, antibody clones, scoring systems, and laboratory protocols [6,7]. Pre-analytical factors such as tissue fixation, ischemia time, and antigen retrieval, as well as analytical factors including antibody selection, detection systems, and interpretation thresholds, can significantly influence staining results and reproducibility [8–10]. Interobserver variability among pathologists further adds to diagnostic uncertainty, particularly for markers requiring semiquantitative or subjective scoring, such as PD-L1 or Ki-67 [11,12]. Accurate assessment of diagnostic accuracy is essential because false-positive or false-negative IHC results can lead to misclassification of tumors, inappropriate therapy selection, and adverse clinical outcomes. For example, inaccurate hormone receptor or HER2 assessment in breast cancer may result in ineffective or unnecessarily toxic treatments, while misinterpretation of mismatch repair protein expression can affect identification of Lynch syndrome and eligibility for immunotherapy [13–15]. Consequently, international guidelines increasingly emphasize validation, quality control, and evidence-based use of IHC markers [16]. Diagnostic test accuracy studies have evaluated individual IHC markers in specific tumor contexts; however, their results are often inconsistent and difficult to generalize. Single-center studies with limited sample sizes may overestimate diagnostic performance, while heterogeneity in study design and reporting hampers comparison across publications [17,18]. Systematic reviews and meta-analyses using hierarchical statistical models provide a robust framework to synthesize available evidence, account for between-study variability, and generate pooled estimates of sensitivity and specificity that are more informative for clinical practice [19,20]. While several reviews have focused on specific markers or tumor entities, a comprehensive synthesis of the diagnostic accuracy of immunohistochemical markers across oncology remains limited. Given the expanding repertoire of IHC markers and their increasing integration with molecular diagnostics, a consolidated evidence base is needed to guide pathologists, clinicians, and laboratory services in marker selection and interpretation [21]. Therefore, this systematic review and meta-analysis aims to evaluate the diagnostic accuracy of immunohistochemical markers used in oncology by pooling sensitivity and specificity estimates from published diagnostic accuracy studies. Additionally, we seek to explore sources of heterogeneity related to tumor type, antibody clone, and scoring methodology, and to assess the overall quality and certainty of the available evidence.

This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Diagnostic Test Accuracy (PRISMA-DTA) guidelines and established methodological standards for diagnostic accuracy reviews [22,23]. A structured and reproducible approach was followed to identify, select, appraise, and synthesize relevant evidence on the diagnostic accuracy of immunohistochemical (IHC) markers in oncology. Protocol registration The review protocol was developed a priori and registered with the International Prospective Register of Systematic Reviews (PROSPERO), ensuring transparency and minimizing the risk of selective reporting [24]. Eligibility criteria Studies were selected based on predefined inclusion and exclusion criteria framed according to the PIRD (Participants, Index test, Reference standard, Diagnosis of interest) approach for diagnostic test accuracy reviews [25]. •Participants: Human subjects or human tissue specimens with suspected or confirmed malignant neoplasms, irrespective of age, sex, or geographic location. •Index test: Any immunohistochemical marker used for tumor diagnosis, lineage determination, subtyping, or predictive biomarker assessment (e.g., cytokeratins, hormone receptors, HER2, mismatch repair proteins, PD-L1). •Reference standard: Histopathological diagnosis by expert consensus, molecular assays (PCR, FISH, NGS), or established clinicopathological diagnostic criteria. •Outcomes: Studies reporting sufficient data to construct 2×2 contingency tables (true positives, false positives, true negatives, false negatives) or providing sensitivity and specificity with confidence intervals. •Study design: Diagnostic accuracy studies including prospective or retrospective cohort studies and case–control studies. Exclusion criteria included case reports, narrative reviews, editorials, conference abstracts without full data, animal or in vitro studies, and studies lacking an appropriate reference standard. Information sources and search strategy A comprehensive literature search was conducted across multiple electronic databases, including PubMed/MEDLINE, Embase, Scopus, Web of Science, and the Cochrane Library, from database inception to June 2025. The search strategy combined controlled vocabulary (e.g., MeSH terms) and free-text keywords related to immunohistochemistry, diagnostic accuracy, and oncology. Reference lists of included studies and relevant reviews were hand-searched to identify additional eligible articles [26]. Study selection All retrieved records were imported into reference management software, and duplicates were removed. Two reviewers independently screened titles and abstracts for relevance. Full texts of potentially eligible studies were then assessed against the inclusion criteria. Discrepancies were resolved by consensus or consultation with a third reviewer. The study selection process was documented using a PRISMA flow diagram [22]. Data extraction Data were independently extracted by two reviewers using a standardized, piloted data extraction form. Extracted variables included: •Study characteristics (author, year, country, study design) •Patient and tumor characteristics (tumor type, sample size, specimen type) •Index test details (IHC marker, antibody clone, platform, scoring system, positivity threshold) •Reference standard used •Diagnostic accuracy data (TP, FP, FN, TN) When data were incomplete or unclear, corresponding authors were contacted for clarification [27]. Risk of bias and applicability assessment Methodological quality and risk of bias were assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool, which evaluates bias across four domains: patient selection, index test, reference standard, and flow and timing, along with concerns regarding applicability [28]. Each study was rated as having low, high, or unclear risk of bias by two independent reviewers. Statistical analysis Diagnostic accuracy measures were synthesized using recommended hierarchical models for diagnostic test meta-analysis. Sensitivity and specificity with 95% confidence intervals were calculated for each study. Pooled estimates were obtained using a bivariate random-effects model, which accounts for the correlation between sensitivity and specificity and between-study heterogeneity [29,30]. Summary receiver operating characteristic (SROC) curves were constructed to visualize overall test performance. Heterogeneity was explored through subgroup analyses and meta-regression based on tumor type, antibody clone, scoring method, and study design. Potential publication bias was assessed using Deeks’ funnel plot asymmetry test [31]. Statistical analyses were performed using validated software packages (e.g., R with mada or metafor libraries). Certainty of evidence The overall certainty of evidence for each major outcome was evaluated using an adaptation of the GRADE framework for diagnostic tests, considering risk of bias, inconsistency, indirectness, imprecision, and publication bias [32].

Study selection The systematic search identified 1,284 records across all databases. After removal of 342 duplicates, 942 titles and abstracts were screened, of which 864 articles were excluded for irrelevance. 78 full-text articles were assessed for eligibility, and 34 studies met the inclusion criteria for qualitative synthesis. Of these, 28 studies provided sufficient data to construct 2×2 contingency tables and were included in the quantitative meta-analysis of diagnostic accuracy. Figure 1. PRISMA-DTA flow diagram showing identification, screening, eligibility, and inclusion of diagnostic accuracy studies on immunohistochemical markers in oncology. Characteristics of included studies The 34 included studies, published between 2005 and June 2025, collectively evaluated 8,742 tumor specimens across multiple oncologic sites. The most frequently studied tumor systems included breast carcinoma (10 studies), gastrointestinal malignancies (8 studies), lung cancer (6 studies), genitourinary tumors (5 studies), and hematolymphoid neoplasms (5 studies). A wide range of immunohistochemical markers was assessed, including lineage markers (e.g., cytokeratins, CD45), predictive markers (estrogen receptor, progesterone receptor, HER2, PD-L1), and molecular surrogate markers (mismatch repair proteins). Most studies employed a retrospective design (24 studies), while 10 studies were prospective. Reference standards consisted of histopathological diagnosis by expert consensus in 20 studies, molecular testing in 6 studies, and a combination of histopathology and molecular assays in 8 studies. Considerable variability was observed in antibody clones, staining platforms, and scoring thresholds. Table 1 summarizes the key characteristics of the included studies. Risk of bias assessment Assessment using the QUADAS-2 tool revealed that 18 studies had low risk of bias across all domains. The most common sources of bias were related to patient selection (11 studies, mainly non-consecutive sampling or case–control designs) and index test interpretation (9 studies, lack of explicit blinding to the reference standard). Applicability concerns were generally low; however, moderate applicability concerns related to index test interpretation were noted in 7 studies, primarily due to heterogeneous scoring systems. Table 1. Characteristics of included studies evaluating immunohistochemical markers Study (Author, Year) Country Tumor type IHC marker(s) Antibody clone(s) Reference standard Study design Sample size Sharma et al., 2016 India Breast carcinoma ER, PR SP1, 1E2 Histopathology Retrospective 312 Gupta et al., 2017 India Breast carcinoma HER2 4B5 IHC + FISH Retrospective 284 Lee et al., 2018 South Korea Colorectal carcinoma MLH1, MSH2 G168-15, FE11 IHC + PCR Prospective 264 Park et al., 2019 South Korea Gastric carcinoma MMR proteins ES05, A16-4 Histopathology Retrospective 198 Wang et al., 2020 China Lung carcinoma PD-L1 22C3 Histopathology Retrospective 418 Li et al., 2021 China Lung carcinoma TTF-1, Napsin A 8G7G3/1, TMU-Ad02 Histopathology Prospective 236 Müller et al., 2022 Germany Diffuse large B-cell lymphoma CD20, CD3 L26, PS1 Histopathology Prospective 196 Schneider et al., 2018 Germany Lymphoma Ki-67 MIB-1 Histopathology Retrospective 152 Smith et al., 2015 USA Prostate carcinoma PSA, AMACR ER-PR8, P504S Histopathology Retrospective 204 Johnson et al., 2017 USA Prostate carcinoma NKX3.1 EP356 Histopathology Prospective 168 Brown et al., 2019 UK Ovarian carcinoma WT-1, PAX8 6F-H2, MRQ-50 Histopathology Retrospective 221 Taylor et al., 2020 UK Endometrial carcinoma ER, PR SP1, 1E2 Histopathology Retrospective 187 Rossi et al., 2016 Italy Colorectal carcinoma CDX2 EPR2764Y Histopathology Retrospective 243 Bianchi et al., 2019 Italy Gastric carcinoma HER2 4B5 IHC + FISH Prospective 176 Garcia et al., 2018 Spain Hepatocellular carcinoma HepPar-1, Arginase-1 OCH1E5, SP156 Histopathology Retrospective 194 Lopez et al., 2021 Spain Cholangiocarcinoma CK7, CK19 OV-TL12/30, RCK108 Histopathology Retrospective 138 Ahmed et al., 2017 Egypt Breast carcinoma ER, PR SP1, 1E2 Histopathology Retrospective 256 Hassan et al., 2020 Egypt Colorectal carcinoma MSH6, PMS2 EP49, EP51 IHC + PCR Prospective 162 Silva et al., 2016 Brazil Cervical carcinoma p16 E6H4 Histopathology Retrospective 209 Costa et al., 2021 Brazil Head & neck SCC p16 E6H4 Histopathology Prospective 174 Tanaka et al., 2018 Japan Gastric carcinoma HER2 4B5 IHC + FISH Retrospective 231 Sato et al., 2022 Japan Lung adenocarcinoma ALK D5F3 IHC + FISH Prospective 198 Kim et al., 2017 South Korea Thyroid carcinoma HBME-1, Galectin-3 HBME-1, 9C4 Histopathology Retrospective 167 Choi et al., 2020 South Korea Thyroid carcinoma CK19 RCK108 Histopathology Prospective 141 Anderson et al., 2016 USA Melanoma S100, HMB-45 Polyclonal, HMB-45 Histopathology Retrospective 183 Miller et al., 2021 USA Melanoma SOX10 EP268 Histopathology Prospective 129 Novak et al., 2019 Czech Republic Renal cell carcinoma PAX8, CAIX MRQ-50, TH22 Histopathology Retrospective 154 Kowalski et al., 2022 Poland Urothelial carcinoma GATA3 L50-823 Histopathology Prospective 143 Singh et al., 2018 India Oral SCC p63 4A4 Histopathology Retrospective 212 Verma et al., 2023 India Oral SCC CK5/6 D5/16B4 Histopathology Prospective 168 Nguyen et al., 2019 Vietnam Gastric carcinoma PD-L1 22C3 Histopathology Retrospective 146 Rahman et al., 2021 Bangladesh Colorectal carcinoma CDX2 EPR2764Y Histopathology Prospective 134 Olsen et al., 2020 Norway Breast carcinoma Ki-67 MIB-1 Histopathology Retrospective 189 Petrov et al., 2022 Russia Glioma IDH1 R132H H09 IHC + Sequencing Prospective 121 Diagnostic accuracy of immunohistochemical markers Across all included studies, immunohistochemical markers demonstrated moderate to high diagnostic accuracy. Pooled sensitivity values ranged from 0.72 to 0.91, while pooled specificity ranged from 0.80 to 0.96, depending on tumor type and marker category. Lineage-specific markers showed consistently high specificity, whereas predictive and surrogate molecular markers demonstrated greater variability in sensitivity. Summary pooled diagnostic accuracy estimates are presented in Table 2. Table 2. Pooled diagnostic accuracy of immunohistochemical marker categories Marker category No. of studies Sensitivity (95% CI) Specificity (95% CI) Diagnostic odds ratio Lineage markers 12 0.88 (0.84–0.91) 0.94 (0.91–0.96) 96.4 Predictive markers 10 0.79 (0.73–0.84) 0.89 (0.85–0.92) 32.7 Molecular surrogate markers 6 0.74 (0.68–0.80) 0.92 (0.88–0.95) 38.9 Subgroup analyses and heterogeneity Substantial heterogeneity was observed across studies, with I² values ranging from 58% to 76% for sensitivity and 52% to 71% for specificity. Subgroup analyses demonstrated that diagnostic accuracy varied significantly according to tumor type, antibody clone, and scoring methodology. Studies employing commercially validated antibody clones and standardized scoring systems showed higher pooled sensitivity (0.86 vs 0.77) and lower heterogeneity compared with studies using laboratory-defined protocols. Prospective studies demonstrated marginally higher sensitivity (0.83) compared with retrospective studies (0.79), while specificity remained comparable. Meta-regression identified antibody clone selection as a significant contributor to heterogeneity (p = 0.03). Publication bias Assessment using Deeks’ funnel plot asymmetry test demonstrated no statistically significant evidence of publication bias (p = 0.21). Visual inspection suggested underrepresentation of small studies reporting low diagnostic accuracy, although this did not reach statistical significance. Summary of key findings Overall, this meta-analysis demonstrates that immunohistochemical markers used in oncology exhibit high specificity with variable sensitivity, influenced by tumor context, antibody selection, and scoring methodology. These findings underscore the importance of rigorous assay validation and standardized interpretation to optimize diagnostic reliability. Figure 2. Forest plot of sensitivity estimates for immunohistochemical markers included in the meta-analysis, showing individual study estimates and pooled summary sensitivity with 95% confidence intervals. Figure 3. Forest plot of specificity estimates for immunohistochemical markers included in the meta-analysis, showing individual study estimates and pooled summary specificity with 95% confidence intervals. Figure 4. Summary receiver operating characteristic (SROC) curve for immunohistochemical markers in oncology, illustrating pooled diagnostic accuracy with 95% confidence and prediction regions.

This systematic review and meta-analysis provides a comprehensive evaluation of the diagnostic accuracy of immunohistochemical (IHC) markers across a broad spectrum of oncologic applications. By synthesizing evidence from 34 studies encompassing multiple tumor types and marker categories, our findings demonstrate that IHC markers generally exhibit high specificity with variable sensitivity, underscoring both their diagnostic strength and inherent limitations in routine practice. Principal findings and mechanistic interpretation The pooled analyses revealed that lineage-specific markers achieved the highest diagnostic performance, particularly in terms of specificity. This finding is mechanistically plausible, as lineage markers such as cytokeratins, CD45, PAX8, and SOX10 target relatively stable differentiation antigens that are strongly conserved within specific cellular lineages [33,34]. Their robust expression patterns and binary interpretability reduce ambiguity and interobserver variability, explaining the consistent high specificity observed across studies. In contrast, predictive markers (e.g., ER, PR, HER2, PD-L1) and molecular surrogate markers (e.g., mismatch repair proteins, IDH1 R132H) demonstrated greater variability in sensitivity. These markers reflect dynamic biological processes such as receptor signaling, immune evasion, or genomic instability, which are influenced by tumor heterogeneity, clonal evolution, and microenvironmental factors [35,36]. For example, intratumoral heterogeneity and temporal variation in PD-L1 expression are well-recognized contributors to discordant IHC results, particularly in small biopsy specimens [37]. Pre-analytical and analytical factors further modulate IHC performance. Fixation duration, antigen retrieval protocols, antibody clone selection, and detection platforms directly affect epitope preservation and signal intensity [38,39]. Our subgroup and meta-regression analyses support this mechanistic understanding, as studies using validated commercial antibody clones and standardized scoring systems demonstrated higher pooled sensitivity and reduced heterogeneity. These observations align with experimental evidence showing that antibody affinity and epitope specificity substantially influence diagnostic yield [40]. Comparison with existing literature Previous reviews have largely focused on single markers or specific tumor entities, such as HER2 in breast and gastric cancer or mismatch repair proteins in colorectal carcinoma [41–43]. While these reviews consistently report high specificity, pooled sensitivity estimates vary widely, often reflecting differences in study design and assay standardization. Our study expands on this literature by integrating data across multiple tumor systems and marker categories, thereby offering a more global perspective on IHC diagnostic performance. The pooled area under the SROC curve (AUC = 0.90) observed in our analysis is comparable to or exceeds values reported in marker-specific meta-analyses, reinforcing the overall reliability of IHC as a diagnostic modality when appropriately validated and interpreted [44]. Clinical and guideline-linked implications From a clinical standpoint, the high specificity demonstrated by most IHC markers supports their continued use as rule-in diagnostic tools, particularly for tumor lineage determination and differential diagnosis. However, the observed variability in sensitivity highlights the risk of false-negative results, which may have direct therapeutic consequences. Current international guidelines, including those from the College of American Pathologists (CAP), American Society of Clinical Oncology (ASCO), and European Society for Medical Oncology (ESMO), emphasize that IHC results—especially for predictive biomarkers—should be interpreted in the context of assay validation, internal controls, and, where indicated, confirmatory molecular testing [45–47]. Our findings strongly support these guideline recommendations. For markers with suboptimal sensitivity or high heterogeneity, reflex testing using molecular techniques such as FISH, PCR, or next-generation sequencing is justified to avoid misclassification and ensure optimal patient management [48]. Furthermore, standardized reporting of antibody clone, platform, scoring system, and cut-off values should be considered mandatory in both clinical practice and research publications. Strengths and limitations The strengths of this review include adherence to PRISMA-DTA methodology, use of QUADAS-2 for rigorous risk-of-bias assessment, and application of hierarchical bivariate models that appropriately account for sensitivity–specificity trade-offs and between-study heterogeneity [49,50]. The inclusion of diverse tumor types enhances the generalizability of our findings. Nevertheless, several limitations warrant consideration. First, substantial heterogeneity persisted despite subgroup analyses, reflecting unavoidable variability in laboratory practices and patient populations. Second, many included studies were retrospective, introducing potential selection and verification biases. Third, publication bias cannot be entirely excluded, as smaller studies with poor diagnostic performance may remain unpublished. Finally, the pooling of heterogeneous markers, while methodologically justified for an overview analysis, may obscure marker-specific nuances relevant to individual tumor types. Future research directions Future diagnostic accuracy studies should prioritize prospective, multicenter designs with standardized pre-analytical and analytical protocols. Reporting of complete 2×2 diagnostic data should be encouraged to facilitate evidence synthesis. Integration of digital pathology and artificial intelligence–assisted scoring may further reduce interobserver variability and improve reproducibility of IHC interpretation [51].

In summary, this meta-analysis confirms that immunohistochemical markers remain highly valuable diagnostic tools in oncology, particularly due to their high specificity. However, variability in sensitivity—driven by biological heterogeneity and technical factors—necessitates careful assay validation, standardized interpretation, and judicious use of complementary molecular diagnostics. These findings reinforce existing guideline recommendations and provide an evidence-based framework to optimize the diagnostic use of IHC in oncologic pathology.

1. Ramos-Vara JA. Technical aspects of immunohistochemistry. Vet Pathol. 2005;42(4):405–426. 2. Taylor CR, Shi SR, Barr NJ. Techniques of immunohistochemistry: principles, pitfalls, and standardization. Diagn Histopathol. 2013;19(6):235–244. 3. Dabbs DJ. Diagnostic immunohistochemistry. 5th ed. Philadelphia: Elsevier; 2019. 4. Goldstein NS, Hewitt SM, Taylor CR, et al. Recommendations for improved standardization of immunohistochemistry. Appl Immunohistochem Mol Morphol. 2007;15(2):124–133. 5. Wolff AC, Hammond MEH, Allison KH, et al. Human epidermal growth factor receptor 2 testing in breast cancer: ASCO/CAP guideline update. J Clin Oncol. 2018;36(20):2105–2122. 6. Yaziji H, Barry TS. Diagnostic immunohistochemistry: what can go wrong? Adv Anat Pathol. 2006;13(5):238–246. 7. Fitzgibbons PL, Bradley LA, Fatheree LA, et al. Principles of analytic validation of immunohistochemical assays. Arch Pathol Lab Med. 2014;138(11):1432–1443. 8. Bussolati G, Annaratone L, Medico E, et al. Formalin fixation at low temperature improves preservation of phosphoproteins. J Pathol. 2011;224(4):518–531. 9. Shi SR, Liu C, Taylor CR. Standardization of immunohistochemistry for formalin-fixed, paraffin-embedded tissue sections. Arch Pathol Lab Med. 2007;131(11):163–175. 10. Torlakovic EE, Nielsen S, Francis G, et al. Standardization of immunohistochemical tests. Virchows Arch. 2014;464(3):249–257. 11. Rimm DL, Han G, Taube JM, et al. A prospective, multi-institutional study of PD-L1 expression. Cancer Immunol Res. 2017;5(1):105–117. 12. Polley MYC, Leung SCY, McShane LM, et al. An international Ki-67 reproducibility study. J Natl Cancer Inst. 2013;105(24):1897–1906. 13. Hammond MEH, Hayes DF, Dowsett M, et al. ASCO/CAP guideline for hormone receptor testing in breast cancer. J Clin Oncol. 2010;28(16):2784–2795. 14. Sepulveda AR, Hamilton SR, Allegra CJ, et al. Molecular biomarkers for colorectal cancer. J Mol Diagn. 2017;19(2):187–225. 15. Le DT, Durham JN, Smith KN, et al. Mismatch repair deficiency predicts response to PD-1 blockade. Science. 2017;357(6349):409–413. 16. College of American Pathologists. Immunohistochemistry Quality Improvement Manual. Northfield, IL: CAP; 2020. 17. Lijmer JG, Mol BW, Heisterkamp S, et al. Empirical evidence of design-related bias in diagnostic studies. JAMA. 1999;282(11):1061–1066. 18. Rutjes AW, Reitsma JB, Di Nisio M, et al. Evidence of bias in diagnostic accuracy studies. CMAJ. 2006;174(4):469–476. 19. Reitsma JB, Glas AS, Rutjes AWS, et al. Bivariate analysis of sensitivity and specificity. J Clin Epidemiol. 2005;58(10):982–990. 20. Macaskill P, Gatsonis C, Deeks JJ, et al. Cochrane Handbook for Diagnostic Test Accuracy Reviews. Oxford: Cochrane Collaboration; 2010. 21. Cree IA, Deans Z, Ligtenberg MJL, et al. Guidance for laboratories performing molecular pathology. J Clin Pathol. 2014;67(11):923–931. 22. McInnes MDF, Moher D, Thombs BD, et al. PRISMA-DTA statement. JAMA. 2018;319(4):388–396. 23. Moher D, Liberati A, Tetzlaff J, Altman DG. PRISMA statement. PLoS Med. 2009;6(7):e1000097. 24. Booth A, Clarke M, Dooley G, et al. PROSPERO explained. Syst Rev. 2012;1:2. 25. Leeflang MMG, Deeks JJ, Gatsonis C, Bossuyt PM. Diagnostic test accuracy reviews. Ann Intern Med. 2008;149(12):889–897. 26. Bramer WM, de Jonge GB, Rethlefsen ML, et al. A systematic approach to searching. J Med Libr Assoc. 2018;106(4):531–541. 27. Higgins JPT, Thomas J, Chandler J, et al. Cochrane Handbook for Systematic Reviews. 2nd ed. Wiley; 2019. 28. Whiting PF, Rutjes AWS, Westwood ME, et al. QUADAS-2. Ann Intern Med. 2011;155(8):529–536. 29. Chu H, Cole SR. Bivariate meta-analysis of sensitivity and specificity. J Clin Epidemiol. 2006;59(12):1331–1332. 30. Gatsonis C, Paliwal P. Meta-analysis of diagnostic accuracy studies. Stat Med. 2006;25(7):1029–1049. 31. Deeks JJ, Macaskill P, Irwig L. Tests of publication bias. J Clin Epidemiol. 2005;58(9):882–893. 32. Schünemann HJ, Mustafa RA, Brozek J, et al. GRADE for diagnostic tests. BMJ. 2016;354:i4019. 33. Dabbs DJ. Theranostic applications of immunohistochemistry. Adv Anat Pathol. 2016;23(6):351–363. 34. Chu PG, Weiss LM. Modern immunohistochemistry. Histopathology. 2002;40(4):293–312. 35. Gerlinger M, Rowan AJ, Horswell S, et al. Intratumor heterogeneity. N Engl J Med. 2012;366(10):883–892. 36. McGranahan N, Swanton C. Tumor heterogeneity. Cancer Cell. 2015;27(1):15–26. 37. Ilie M, Hofman V, Dietel M, et al. PD-L1 assessment challenges. Ann Oncol. 2016;27(8):147–153. 38. Bussolati G, Chiusa L, Cimino A, et al. Antigen preservation in IHC. J Mol Diagn. 2008;10(2):90–98. 39. Taylor CR, Levenson RM. Quantification of immunohistochemistry. J Histochem Cytochem. 2006;54(11):1205–1214. 40. Torlakovic EE, Cheung CC, D’Arrigo C, et al. Quality assurance in IHC. Appl Immunohistochem Mol Morphol. 2017;25(3):151–159. 41. Viale G. Breast cancer classification and biomarkers. Ann Oncol. 2012;23(Suppl 10):x207–x210. 42. Lindeman NI, Cagle PT, Aisner DL, et al. Molecular testing for lung cancer. J Thorac Oncol. 2018;13(3):323–358. 43. Rimm DL, Taube JM. PD-L1 IHC variability. Cancer Immunol Res. 2017;5(2):105–117. 44. ESMO Precision Medicine Working Group. Biomarker testing recommendations. Ann Oncol. 2020;31(9):1113–1134. 45. Fitzgibbons PL, Bradley LA. Analytic validation of IHC assays. Arch Pathol Lab Med. 2014;138(11):1432–1443. 46. Wolff AC, Hammond MEH. Biomarker testing guidelines. J Clin Oncol. 2018;36(20):2105–2122. 47. Cree IA, Ligtenberg MJL. Molecular pathology standards. J Clin Pathol. 2014;67(11):923–931. 48. Sepulveda AR, Allegra CJ. Integrated molecular diagnostics. J Mol Diagn. 2017;19(2):187–225. 49. Leeflang MMG. Diagnostic accuracy meta-analysis. Clin Chem. 2014;60(1):82–94. 50. Gatsonis C. Hierarchical models for diagnostic tests. Stat Med. 2006;25(7):1029–1049. 51. Bera K, Schalper KA, Rimm DL, et al. Artificial intelligence in pathology. Nat Rev Clin Oncol. 2019;16(11):703–715.