Contents
On this page
- Quick answer
- Step 1: Frame a biomarker-defined PICO and register the protocol
- Step 2: Assess risk of bias with RoB 2 and ROBINS-I
- Step 3: Read PFS and OS with care
- Step 4: Report class toxicities from systematic evidence
- Step 5: Pool the trials and handle biomarker subgroups
- Step 6: Rate certainty with GRADE and report with PRISMA 2020
- What UAE-based review teams should consider
- Frequently asked questions
- Sources
AcademySystematic Review
Guide
Systematic Reviews of Targeted Cancer Therapies: How to Design Them and How to Read Them
A step-by-step guide to reviewing EGFR, HER2, PARP and CDK4/6 inhibitor evidence: biomarker-defined questions, risk of bias, surrogate end points, subgroup traps, GRADE and PRISMA 2020, with notes for UAE-based teams.
- Systematic review
- Oncology
- Targeted therapy
- UAE
- PFS as a surrogate
- 52% of trial-level correlations with OS were low (r ≤ 0.7)
- Risk of bias
- RoB 2 for RCTs, ROBINS-I for non-randomised studies
- Certainty
- GRADE, per outcome
- Reporting
- PRISMA 2020, 27 items; register in PROSPERO
At a glanceSystematic Review
Explore systematic review servicesDiscuss your project
Quick answer
- Define the population by biomarker and test method, and register the protocol in PROSPERO.
- Use RoB 2 for randomised trials and ROBINS-I for non-randomised studies, outcome by outcome.
- Treat PFS as a surrogate: many trial-level correlations with OS are weak, so report OS too.
- Test biomarker subgroups within trials, rate certainty with GRADE and report with PRISMA 2020.
Step 1: Frame a biomarker-defined PICO and register the protocol
A targeted drug is meant to work in patients whose tumours carry the target, so the review population must be defined by the biomarker, not only by cancer type. A Cochrane review of first-line EGFR-targeted therapy in non-small cell lung cancer (NSCLC) shows why this matters. Of its 19 included trials, 7 recruited only people with EGFR mutation-positive tumours; the rest recruited mixed populations and reported the mutation-positive patients as a subgroup1. Your PICO must say in advance whether subgroup data of this kind are eligible, and how the biomarker had to be tested.
Table 1. Illustrative example: a biomarker-defined PICO for a review of EGFR inhibitors
| Element | What to specify |
|---|---|
| Population | Adults with advanced non-squamous NSCLC and a confirmed activating EGFR mutation; the accepted test methods (tissue or plasma); whether mutation-positive subgroups from mixed trials count |
| Intervention | An EGFR tyrosine kinase inhibitor given first line, alone or in combination |
| Comparator | Platinum-based chemotherapy, or another EGFR inhibitor |
| Outcomes | Overall survival (OS), progression-free survival (PFS), objective response, grade 3 or higher adverse events, quality of life |
| Study designs | Randomised controlled trials (RCTs) for effects; non-randomised studies only if pre-specified, for example for rare harms |
Register the protocol in PROSPERO before screening starts. PROSPERO was set up to reduce unplanned duplication of reviews and to make the review process transparent, with the aim of minimising reporting bias; registration is free2. PRISMA 2020 asks authors to give the register name and registration number, or to state that the review was not registered3. Write the planned biomarker subgroup analyses into the protocol now. Subgroups added after the data are seen are much harder to defend.
Search tip
Search conference abstracts as well as databases. The Cochrane EGFR review searched CENTRAL, MEDLINE, Embase and Web of Science, plus ASCO and ESMO conference abstracts and NICE evidence review group submissions1.
Step 2: Assess risk of bias with RoB 2 and ROBINS-I
Use RoB 2 for randomised trials. It assesses five domains: the randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of the reported result4. Use ROBINS-I for non-randomised studies of interventions, such as registry cohorts. It covers bias due to confounding, selection of participants, classification of interventions, deviations from intended interventions, missing data, measurement of outcomes and selection of the reported result5. If you are unsure which tool fits a design, see our risk of bias tool decision tree.
Table 2. Where targeted-therapy studies often raise concerns
| Tool and domain | What to check |
|---|---|
| RoB 2: measurement of the outcome | Open-label trials with investigator-assessed PFS; check whether a blinded independent central review was done |
| RoB 2: deviations from intended interventions | Crossover to the targeted drug after progression, and other post-progression treatment that can affect OS comparisons |
| RoB 2: selection of the reported result | Biomarker or subgroup results reported without a pre-specified analysis plan |
| ROBINS-I: confounding | Patients who were tested and treated with a targeted drug may differ from others in stage, performance status or access to care |
Judge risk of bias for each outcome, not once per trial. The same trial can be at low risk of bias for OS and raise concerns for investigator-assessed PFS.
Step 3: Read PFS and OS with care
Most targeted-therapy trials report PFS because it matures earlier than OS. PFS is a useful decision outcome only if it predicts OS for the drug class and setting under review, and the evidence for that is often weak. A systematic review of trial-level meta-analyses identified 65 correlations between surrogate end points and survival in 36 articles. Of these, 52% were of low strength (r ≤ 0.7), 25% of medium strength and 23% were highly correlated with survival (r ≥ 0.85)6. The authors concluded that the evidence supporting surrogate end points in oncology is limited6.
The EGFR example shows the gap. Pooled PFS hazard ratios (HRs) favoured the inhibitors over chemotherapy: erlotinib HR 0.30 (95% CI 0.24 to 0.38), gefitinib HR 0.39 (0.32 to 0.48) and afatinib HR 0.42 (0.34 to 0.53). Yet the review found no increase in OS for the inhibitors compared with standard chemotherapy1. A review should report both outcomes, name the primary one, and not describe a PFS gain as a survival gain.
What we check for PFS and OS
- Is PFS assessed by investigators or by blinded central review?
- Is OS mature, and are crossover and post-progression treatments reported?
- Has the PFS–OS association been validated for this drug class and setting?
- Were hazard ratios with confidence intervals extracted, not medians alone?
Systematic review support for oncology evidence
Pubrica’s systematic review team supports protocol writing, PROSPERO registration, searching, risk of bias assessment, meta-analysis and GRADE. Authors remain accountable for the content and conclusions of their review, writing assistance is acknowledged in line with the ICMJE recommendations and GPP 2022, and Pubrica does not guarantee acceptance by any journal.
Step 4: Report class toxicities from systematic evidence
Safety needs the same rigour as efficacy. Extract grade 3 or higher adverse events, serious adverse events, dose reductions and discontinuations, and pre-specify the known class toxicities as harms of interest (Table 3).
Table 3. Class toxicities reported in systematic reviews and meta-analyses
| Drug class | Key toxicities | Evidence |
|---|---|---|
| EGFR tyrosine kinase inhibitors | Diarrhoea and rash | Meta-analysis of 28 RCTs against placebo (17,800 patients): all-grade diarrhoea 53.7% and rash 48.6%7 |
| CDK4/6 inhibitors | Neutropenia (palbociclib, ribociclib); diarrhoea (abemaciclib) | Systematic review and meta-analysis: palbociclib and ribociclib showed high rates of haematological toxicity, mainly neutropenia; abemaciclib mainly grade 1–2 diarrhoea8 |
| HER2-targeted therapy (trastuzumab) | Cardiotoxicity | Cochrane review of 8 trials (11,991 patients): congestive heart failure RR 5.11 (90% CI 3.00 to 8.72); LVEF decline RR 1.83 (90% CI 1.36 to 2.47)9 |
Put benefit and harm side by side. The same trastuzumab review found that trastuzumab-containing regimens improved OS (HR 0.66, 95% CI 0.57 to 0.77) while increasing cardiac risk9.
Step 5: Pool the trials and handle biomarker subgroups
Pool time-to-event outcomes as hazard ratios and quantify heterogeneity. The I² statistic describes the percentage of variability across studies that is due to heterogeneity rather than chance10. A random-effects model assumes that the true effect varies between trials. Its summary estimate is the average effect, and a prediction interval shows the range in which the effect in a new setting is likely to fall11. Report both. Our guide on how to read a forest plot shows where they appear.
Biomarker subgroup analysis is where many oncology reviews go wrong. Comparing pooled results across trials, such as trials in mutation-positive patients against trials in mixed populations, mixes the biomarker with every other difference between those trials. Fisher and colleagues recommend testing interactions using within-trial information alone12. Use ICEMAN to judge whether an apparent subgroup effect is credible; it has 5 core questions for RCTs and 8 for meta-analyses13. Few trials also limit what subgroups can show: the Cochrane trastuzumab review noted that its subgroup analyses were limited by the small number of studies9.
Illustrative example: four trials each report HRs for patients with EGFR exon 19 deletions and with exon 21 L858R mutations. Estimate the ratio of the two HRs within each trial, then pool those four ratios. Do not compare the pooled exon 19 result with the pooled L858R result.
Subgroup pitfalls
- Subgroups not pre-specified in the protocol
- Many subgroups tested, so some look significant by chance
- Between-trial differences presented as biomarker effects
- A separate p value in each subgroup instead of a test for interaction
- Different biomarker assays across trials, not reported
Step 6: Rate certainty with GRADE and report with PRISMA 2020
GRADE rates the quality, or certainty, of evidence for each outcome as high, moderate, low or very low14. Evidence from randomised trials starts as high and can be rated down for study limitations, inconsistency, indirectness, imprecision and publication bias14. In targeted-therapy reviews, indirectness is common when a biomarker subgroup from a mixed trial, or a surrogate such as PFS, stands in for the population or outcome of interest. In the Cochrane EGFR review, the quality of evidence was rated high for erlotinib, gefitinib and afatinib each compared with chemotherapy1.
Report the review with PRISMA 2020, which provides a 27-item checklist, an expanded checklist, an abstract checklist and revised flow diagrams3. Oncology trials are often published several times: a primary analysis, an updated OS analysis and separate biomarker papers. Link all reports to one study so that patients are not counted twice, and show this in the PRISMA 2020 flow diagram.
What UAE-based review teams should consider
We found no UAE-specific guidance on systematic reviews of targeted cancer therapies. The points below draw on a small number of regional studies and should be read as prompts, not rules.
- Biomarker prevalence may differ. A retrospective study from centres in Saudi Arabia, the UAE and Qatar found EGFR mutations in 66 of 230 patients with NSCLC (28.7%). This was higher than reported in Western populations and lower than in Far Eastern populations15. By contrast, 1,700 of the 2,317 mutation-positive participants in the Cochrane EGFR review were of Asian origin1. Consider whether the trial populations are indirect for your patients when you apply GRADE.
- Testing access shapes the evidence. In 451 tumour samples from patients treated in the UAE and Thailand that underwent comprehensive molecular profiling, 75% had an alteration with an associated therapy. The authors note that such profiling is not widely available across the globe16. Record how biomarker status was established in any local study you include.
- Regional evidence is sparse. Report how many UAE or Gulf studies met your eligibility criteria, and say so plainly if none did. Many patients worldwide do not yet have access to precision medicines17, so access to testing and drugs is a fair point for the discussion section.
Pubrica team verdict
A review of targeted cancer therapies is only as good as its biomarker definition, its handling of PFS and its subgroup analysis. Pre-specify all three, and let GRADE show where the evidence is indirect.
Frequently asked questions
RoB 2 for randomised trials and ROBINS-I for non-randomised studies of interventions, such as registry cohorts. Judge each outcome separately.
Not by default. A review of trial-level meta-analyses found that more than half of reported surrogate–survival correlations in oncology were of low strength, so report OS alongside PFS.
Pre-specify them in the protocol, estimate interactions within trials before pooling, and judge credibility with ICEMAN rather than relying on separate subgroup p values.
Registration is free and strongly advised. PRISMA 2020 asks you to report the register and number, or to state that the review was not registered.
Some regional studies exist, such as a Gulf study of EGFR mutation prevalence, but they are few. Search for them, report what you find, and say so if none met your criteria.
Sources
Sources were checked on 19 November 2026. Requirements and guidance may change after that date; confirm the current version with the issuing body.
- 1Greenhalgh J, Dwan K, Boland A, et al. First-line treatment of advanced epidermal growth factor receptor (EGFR) mutation positive non-squamous non-small cell lung cancer. Cochrane Database Syst Rev. 2016;(5):CD010383. https://doi.org/10.1002/14651858.CD010383.pub2
- 2Booth A, Clarke M, Dooley G, et al. The nuts and bolts of PROSPERO: an international prospective register of systematic reviews. Syst Rev. 2012;1:2. https://doi.org/10.1186/2046-4053-1-2
- 3Page MJ, McKenzie JE, Bossuyt PM, et al. The PRISMA 2020 statement: an updated guideline for reporting systematic reviews. BMJ. 2021;372:n71. https://doi.org/10.1136/bmj.n71
- 4Sterne JAC, Savović J, Page MJ, et al. RoB 2: a revised tool for assessing risk of bias in randomised trials. BMJ. 2019;366:l4898. https://doi.org/10.1136/bmj.l4898
- 5Sterne JA, Hernán MA, Reeves BC, et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016;355:i4919. https://doi.org/10.1136/bmj.i4919
- 6Prasad V, Kim C, Burotto M, Vandross A. The strength of association between surrogate end points and survival in oncology: a systematic review of trial-level meta-analyses. JAMA Intern Med. 2015;175(8):1389-98. https://doi.org/10.1001/jamainternmed.2015.2829
- 7Yin X, Zhao Z, Yin Y, et al. Adverse event profiles of epidermal growth factor receptor-tyrosine kinase inhibitors in cancer patients: a systematic review and meta-analysis. Clin Transl Sci. 2021;14(3):919-933. https://doi.org/10.1111/cts.12957
- 8Onesti CE, Jerusalem G. CDK4/6 inhibitors in breast cancer: differences in toxicity profiles and impact on agent choice. A systematic review and meta-analysis. Expert Rev Anticancer Ther. 2021;21(3):283-298. https://doi.org/10.1080/14737140.2021.1852934
- 9Moja L, Tagliabue L, Balduzzi S, et al. Trastuzumab containing regimens for early breast cancer. Cochrane Database Syst Rev. 2012;2012(4):CD006243. https://doi.org/10.1002/14651858.CD006243.pub2
- 10Higgins JP, Thompson SG, Deeks JJ, Altman DG. Measuring inconsistency in meta-analyses. BMJ. 2003;327(7414):557-60. https://doi.org/10.1136/bmj.327.7414.557
- 11Riley RD, Higgins JP, Deeks JJ. Interpretation of random effects meta-analyses. BMJ. 2011;342:d549. https://doi.org/10.1136/bmj.d549
- 12Fisher DJ, Carpenter JR, Morris TP, et al. Meta-analytical methods to identify who benefits most from treatments: daft, deluded, or deft approach? BMJ. 2017;356:j573. https://doi.org/10.1136/bmj.j573
- 13Schandelmaier S, Briel M, Varadhan R, et al. Development of the Instrument to assess the Credibility of Effect Modification Analyses (ICEMAN) in randomized controlled trials and meta-analyses. CMAJ. 2020;192(32):E901-E906. https://doi.org/10.1503/cmaj.200077
- 14Guyatt GH, Oxman AD, Vist GE, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336(7650):924-6. https://doi.org/10.1136/bmj.39489.470347.AD
- 15Jazieh AR, Jaafar H, Jaloudi M, et al. Patterns of epidermal growth factor receptor mutation in non-small-cell lung cancers in the Gulf region. Mol Clin Oncol. 2015;3(6):1371-1374. https://doi.org/10.3892/mco.2015.644
- 16Dawood S, Natarajan V, Danchaivijitr P. Comprehensive molecular profiling identifies actionable biomarkers for patients from Thailand and the United Arab Emirates with advanced malignancies. Front Oncol. 2024;14:1374087. https://doi.org/10.3389/fonc.2024.1374087
- 17Rulten SL, Grose RP, Gatz SA, Jones JL, Cameron AJM. The future of precision oncology. Int J Mol Sci. 2023;24(16):12613. https://doi.org/10.3390/ijms241612613
targeted cancer therapiessystematic reviews in oncologyprecision oncologycancer drug effectivenesstargeted therapy safetyclinical outcomes
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation, so check its guide if it has one.
Dr. Charlie. (2026, November 19). Systematic Reviews of Targeted Cancer Therapies: How to Design Them and How to Read Them. Pubrica. https://pubrica.com/academy/systematic-review/systematic-reviews-targeted-cancer-therapies/
This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Next steps
Systematic review support for oncology evidence
Pubrica’s systematic review team supports protocol writing, PROSPERO registration, searching, risk of bias assessment, meta-analysis and GRADE. Authors remain accountable for the content and conclusions of their review, writing assistance is acknowledged in line with the ICMJE recommendations and GPP 2022, and Pubrica does not guarantee acceptance by any journal.
This article is general information for researchers, not legal, regulatory or funding advice. Check the current requirements of your journal, funder, ethics committee or regulator before you act on it.
More in Systematic Review