Is "ctDNA as a Predictor of Surgical Margins and Recurrence in GI Cancers" a Good Meta-Analysis Topic?

1. What Literature Already Exists (The "Taken" Ground)

2. The Genuine Research Gaps

3. How to Sharpen the Topic

• Covers all GI cancer subtypes
• Primary outcome: recurrence-free survival (RFS) stratified by ctDNA positivity
• Secondary: sensitivity/specificity for detecting actual margin involvement
• Novelty: pan-GI scope + structured analysis of sampling time points
• Feasibility: High - individual study data exists across all subtypes

• Directly targets the surgical margin gap
• Novelty: Highest - no prior systematic review exists
• Feasibility: Harder, because most primary studies don't report margin status as a dedicated ctDNA endpoint; you'd be extracting it from study subgroups

4. Methodological Considerations

• Heterogeneity will be high - different assay platforms (ddPCR, NGS-based tumor-informed, methylation-based), different cancer subtypes, different sampling time points. Plan subgroup analyses by cancer type and assay method upfront.
• Clonal hematopoiesis (CHIP) confounding - the 2026 Safiejko review flags this as a major uncontrolled variable in current studies; address it in your methods.
• KRAS mutation as a shared marker - KRAS mutations are present in ~40% of CRC, ~90% of PDAC, and ~10% of gastric cancers, which may allow cross-cancer pooling for a mutation-specific subgroup.
• Registration: Pre-register on PROSPERO before starting. A quick search suggests no registered protocol exists for this exact topic (another signal the gap is real).

5. Summary Assessment

Complete Beginner's Guide to Conducting a Meta-Analysis

For Final-Year Medical Students | Topic: ctDNA in GI Cancer Surgery

Before You Start: What Is a Meta-Analysis, Really?

THE 10 STEPS

STEP 1: Frame Your Research Question Using PICO

"In adults with resectable GI cancers undergoing curative-intent surgery, does ctDNA positivity (pre- or postoperatively) predict surgical margin involvement and/or disease recurrence compared to ctDNA-negative patients?"

STEP 2: Register Your Protocol on PROSPERO

1. Go to prospero.ac.uk
2. Create a free account
3. Fill in the registration form - it asks for your PICO question, your search strategy, your inclusion/exclusion criteria, your planned outcomes, and your statistical methods
4. Submit - you'll receive a PROSPERO ID number (e.g., CRD42026XXXXXX)
5. Include this ID in your paper's methods section

STEP 3: Write Your Inclusion and Exclusion Criteria

1. Study design: original research (prospective or retrospective cohort, RCT, case-control) - no editorials, letters, or case reports
2. Population: adults (≥18 years) with histologically confirmed GI cancer (colorectal, gastric, pancreatic, esophageal, hepatocellular, cholangiocarcinoma)
3. Underwent surgical resection with curative intent
4. ctDNA measured from blood (plasma or serum) - not tissue-only studies
5. Reports at least one of your outcomes: RFS, DFS, OS, margin status correlation
6. Provides extractable data (HR, OR, RR with confidence intervals, or Kaplan-Meier data)
7. Published in English (or languages you can translate)

1. Metastatic/palliative setting only (no surgery)
2. ctDNA from stool or other non-blood samples
3. Duplicate publications (keep only the most recent/complete report)
4. Conference abstracts only (no full text available)
5. Animal or in vitro studies
6. Reviews, meta-analyses, editorials

STEP 4: Build and Run Your Search Strategy

• PubMed/MEDLINE
• Embase
• Cochrane Library
• Web of Science

("circulating tumor DNA" OR "circulating tumour DNA" OR "ctDNA" OR 
"cell-free DNA" OR "cfDNA" OR "liquid biopsy" OR "minimal residual disease")

("gastrointestinal cancer" OR "colorectal cancer" OR "colon cancer" OR 
"rectal cancer" OR "gastric cancer" OR "stomach cancer" OR "pancreatic cancer" OR 
"esophageal cancer" OR "oesophageal cancer" OR "hepatocellular carcinoma" OR 
"cholangiocarcinoma" OR "biliary cancer")

("surgical resection" OR "curative resection" OR "surgical margins" OR 
"resection margin" OR "R0" OR "R1" OR "recurrence" OR "disease-free survival" OR 
"recurrence-free survival" OR "postoperative")

• Hand-search reference lists of all included studies ("snowballing")
• Search ClinicalTrials.gov for completed trials that may have published data
• Search Google Scholar for the first 5-10 pages

STEP 5: Screen Studies (Two Rounds)

• Import all results into a free tool like Rayyan (rayyan.ai) or Covidence (covidence.org)
• These tools de-duplicate results automatically (the same paper appears in multiple databases)
• Read only the title and abstract
• Mark each study: Include / Exclude / Unsure
• The rule: when in doubt at this stage, include - you'll look more carefully in Round 2

• Download the full PDF of every study that passed Round 1
• Now apply your inclusion/exclusion criteria strictly
• For each excluded study, record WHY you excluded it (e.g., "wrong population - metastatic only")

STEP 6: Extract Data From Included Studies

• If a study reports Kaplan-Meier curves but no HR, you can extract approximate HRs using free online tools like WebPlotDigitizer
• If results are only in a figure, note it and use digitizing software
• Do extraction twice (or have a second person verify) to catch errors

STEP 7: Assess Quality of Included Studies

• 4 domains: Patient Selection, Index Test, Reference Standard, Flow and Timing
• Rate each domain: Low risk / High risk / Unclear risk
• Free template at cochrane.org

• NOS is simpler: 9 stars maximum, 7+ stars = high quality
• Free access at ohri.ca/programs/clinical_epidemiology/oxford.asp

STEP 8: Statistical Analysis (The Meta-Analysis Part)

• RevMan 5 (Cochrane's software) - most widely used, free
• R with meta or metafor packages - more flexible, steeper learning curve
• OpenMeta[Analyst] - user-friendly GUI

The Key Concepts You Need to Understand:

• Fixed-effect: assumes all studies measure exactly the same true effect. Rarely appropriate.
• Random-effects (DerSimonian-Laird): assumes studies estimate similar but not identical effects. Use this. It's almost always the right choice in clinical meta-analyses because studies differ in patient populations, ctDNA assays, and follow-up.

• I² < 25%: low heterogeneity - studies are consistent
• I² 25-75%: moderate - investigate why
• I² > 75%: high - be cautious about pooling; results vary substantially

• Run subgroup analyses - split studies by cancer type (CRC vs. pancreatic vs. gastric), ctDNA assay type, or sampling time point (pre-op vs. post-op)
• Run meta-regression to see if a variable explains the variation

• Funnel plot (visual - asymmetric funnel suggests bias)
• Egger's test (statistical test for asymmetry)

What Your Main Analysis Will Produce:

Subgroup Analyses to Plan:

1. By cancer type (CRC / gastric / pancreatic / esophageal / hepatobiliary)
2. By ctDNA sampling time (preoperative / early postoperative / surveillance)
3. By assay type (tumor-informed ddPCR / tumor-agnostic NGS panel)
4. By study quality (high NOS score ≥7 vs. low)
5. By margin status (R0 vs. R1 subgroup, if enough studies report this)

STEP 9: Write the Paper

"Circulating tumor DNA as a predictor of recurrence and surgical margin status after curative-intent resection in gastrointestinal cancers: a systematic review and meta-analysis"

1. Clinical context: GI cancers, surgical recurrence, the problem of incomplete resection
2. What ctDNA is and its proposed role
3. What's missing: the gap in pan-GI evidence and margin-specific data
4. Your study aim and PICO question

• Protocol and registration (PROSPERO ID)
• Search strategy (paste exact search strings + databases + date)
• Inclusion/exclusion criteria (as a table)
• Data extraction process
• Quality assessment tools used
• Statistical methods (software, fixed vs. random effects, heterogeneity measures, subgroup analyses, publication bias tests)

• PRISMA flowchart (number of records identified → screened → assessed → included)
• Table of study characteristics
• Quality assessment table
• Forest plots for each outcome
• Subgroup analysis results
• Funnel plot and Egger's test result

1. Summary of main findings
2. How your results compare to existing cancer-specific meta-analyses
3. Biological plausibility of ctDNA as a margin/recurrence predictor
4. Limitations (heterogeneity, assay variability, publication bias)
5. Clinical implications and future research directions

STEP 10: Submit and Respond to Reviewers

• Gut (IF ~24) - top GI journal
• Annals of Surgery (IF ~13)
• European Journal of Surgical Oncology
• Surgical Oncology
• BMC Cancer - open access, good for first submissions
• PLOS ONE - open access, less selective, good if you need a guaranteed outlet

1. Format manuscript to journal's author guidelines (word limits, reference style, figure specs)
2. Write a cover letter explaining why your paper fits the journal and what it adds to the field
3. Submit through the journal's online portal
4. Expect 4-12 weeks for a first decision

YOUR TIMELINE (Realistic for a Medical Student)

FREE TOOLS SUMMARY

SUPERVISOR NOTE

• Strengthen your PROSPERO registration
• Help resolve disagreements during screening
• Add credibility to the submission (co-authorship on a meta-analysis requires meeting ICMJE criteria - substantial contribution to conception, data, or drafting)
• Connect you to institutional resources (database access, biostatistics support)

ONE LAST THING: The Most Common Mistakes First-Timers Make

1. Starting the search before registering on PROSPERO - always register first
2. Changing inclusion criteria mid-screening - decide before you start, stick to it
3. Only searching PubMed - you will miss studies; always use at least 3 databases
4. Pooling studies with I² > 75% without explanation - explain heterogeneity or don't pool
5. Forgetting to extract HR adjustment variables - unadjusted and adjusted HRs should not be pooled together
6. Not doing a sensitivity analysis - reviewers will ask for one if you don't include it
7. Submitting to a journal above your reach on first attempt - aim 1-2 tiers below your dream journal for the first submission

How to Do a Meta-Analysis: Simple Version

Your topic: Does ctDNA predict surgical margins & cancer recurrence in GI cancers?

What are you actually doing?

Analogy: Imagine 10 different people each taste-tested a new medicine on 50 patients. Each got slightly different results because their groups were small. You gather all 10 reports, combine all 500 patients' worth of data, and now get a much more trustworthy answer. That's a meta-analysis.

The 10 Steps — As Simply As Possible

STEP 1 - Decide exactly what you're asking

• Who are the patients? → Adults with GI cancer (colorectal, stomach, pancreatic, esophageal) who had surgery
• What are you testing? → ctDNA (a cancer DNA marker found in blood)
• Compared to what? → Patients whose ctDNA was negative
• What outcome? → Did the cancer come back? Were surgical margins clear or not?

"In GI cancer patients who had surgery, does a positive ctDNA blood test predict whether cancer will come back or whether the surgeon got clear margins?"

STEP 2 - Register your plan before you start

• What your question is
• Where you'll search for studies
• How you'll decide which studies to include

STEP 3 - Decide which studies you'll include and which you'll reject

• Are about real patients (not animals or lab experiments)
• Patients had GI cancer and had surgery to remove it
• A ctDNA blood test was done
• The study reports whether cancer came back, or survival data

• Only include patients with cancer that had already spread (no surgery done)
• Are just opinion pieces, letters, or conference summaries
• Don't give you any numbers you can use

STEP 4 - Search for studies

• PubMed (main medical database)
• Embase (broader coverage)
• Cochrane Library (reviews and trials)

("ctDNA" OR "circulating tumor DNA" OR "liquid biopsy")
AND
("colorectal" OR "gastric" OR "pancreatic" OR "esophageal")
AND
("recurrence" OR "surgical margin" OR "disease-free survival")

STEP 5 - Screen studies (filter down to the good ones)

STEP 6 - Extract the data

STEP 7 - Check the quality of each study

STEP 8 - Do the statistics (the actual "meta-analysis" part)

• Under 25% = they mostly agree ✅
• 25–75% = some disagreement, investigate why
• Over 75% = they disagree a lot — be careful pooling them ⚠️

STEP 9 - Write the paper

• How many studies you found in databases
• How many you excluded after title/abstract screening (and why)
• How many you excluded after reading the full text (and why)
• How many you finally included

STEP 10 - Submit to a journal

• BMC Cancer (open access, good reputation)
• Surgical Oncology
• European Journal of Surgical Oncology
• Aim lower first — getting published matters more than the journal name on your first paper

• Accepted (very rare on first submission)
• Major or minor revisions — this is normal. Reviewers ask questions. You answer each one, resubmit. Most papers get published after 1–2 revision rounds.
• Rejected — reformat and submit to the next journal on your list. Don't give up.

Your Timeline at a Glance

Month 1   → Write question, register on PROSPERO, plan searches
Month 2   → Run searches, screen titles/abstracts
Month 3   → Read full texts, extract data, check quality
Month 4   → Run statistics in RevMan, make forest plots
Month 5   → Write the paper
Month 6   → Submit to journal

The Only Tools You Need

The 5 Biggest Mistakes to Avoid

1. Starting searches before registering on PROSPERO — always register first
2. Only searching PubMed — you'll miss studies; use at least 3 databases
3. Changing your inclusion rules halfway through — decide before you start and stick to it
4. Pooling results when I² is very high without explaining why — reviewers will catch this
5. Giving up after one rejection — almost every paper gets rejected at least once

Your PROSPERO Registration — Field by Field

Before You Open the Form

1. A free account at prospero.ac.uk — sign up with your university email
2. An ORCID number — get one free at orcid.org (takes 2 minutes). It's a unique researcher ID. Not mandatory but strongly recommended.

• Submitting in English? → Yes
• Is this a scoping/mapping review? → No
• Outcome related to human health? → Yes
• Is this a Cochrane review? → No
• Is this for a training course? → No
• Have you searched for similar existing reviews? → Yes
• Do you have a protocol? → Yes
• More than one person involved? → Yes (even if unsure, say yes — you may get a supervisor)
• Do you intend to publish? → Yes
• Preliminary searches started? → Started
• Screening started? → Not started

THE FULL FORM — Every Field Explained

1. Review Title

Circulating tumor DNA (ctDNA) as a predictor of surgical margin status and disease recurrence following curative-intent resection in gastrointestinal cancers: a systematic review and meta-analysis

2. Original Language Title

3. Anticipated Start Date

4. Anticipated Completion Date

Example: If today is June 2026, write: December 2026

5. Stage of the Review at Time of Registration

6. Review Question

In adults with histologically confirmed gastrointestinal cancers (including colorectal, gastric, pancreatic, esophageal, hepatocellular carcinoma, and cholangiocarcinoma) undergoing curative-intent surgical resection, does the detection of circulating tumor DNA (ctDNA) in peripheral blood — measured preoperatively, perioperatively, or postoperatively — predict pathological surgical margin status (R0 vs. R1/R2) and/or disease recurrence (recurrence-free survival, disease-free survival, overall survival) compared to ctDNA-negative patients?

7. Searches

The following electronic databases will be searched from inception to the date of search with no language restrictions applied initially: MEDLINE via PubMed, Embase via Ovid, Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science. Search terms will be constructed using three concept blocks combined with the Boolean operator AND: (1) ctDNA/liquid biopsy terms; (2) gastrointestinal cancer type terms; (3) surgical resection/margin/recurrence terms. MeSH headings and free-text terms will be used. Reference lists of all included studies will be hand-searched to identify additional eligible studies. ClinicalTrials.gov will be searched for completed but unpublished trial data.

8. Condition or Domain Being Studied

Gastrointestinal cancers including colorectal cancer, gastric cancer, pancreatic ductal adenocarcinoma, esophageal cancer, hepatocellular carcinoma, and cholangiocarcinoma, in the context of curative-intent surgical resection.

9. Population

Adults (aged 18 years or older) with histologically confirmed gastrointestinal cancers who underwent surgical resection with curative intent. Both resectable and borderline-resectable disease will be included. Studies enrolling only metastatic or palliative patients will be excluded.

10. Intervention(s), Exposure(s)

Detection of circulating tumor DNA (ctDNA) in peripheral blood (plasma or serum) using any validated assay platform, including but not limited to digital droplet PCR (ddPCR), next-generation sequencing (NGS)-based tumor-informed panels, and methylation-based cell-free DNA assays. ctDNA measured at any time point (preoperative, intraoperative, early postoperative, or during surveillance) will be included.

11. Comparator(s) / Control(s)

Patients who tested ctDNA-negative at the corresponding time point, or comparison against standard clinicopathological assessment alone (e.g., margin status by histopathology, CEA levels, conventional imaging).

12. Types of Study to Be Included

Prospective and retrospective cohort studies, case-control studies, and randomized controlled trials reporting ctDNA detection in relation to surgical margins or recurrence outcomes in GI cancer patients undergoing curative-intent resection. Conference abstracts without full-text data, review articles, editorials, letters, case reports, and studies conducted in animal or in vitro models will be excluded.

13. Context

Studies conducted in any country and any healthcare setting (academic, community, tertiary referral centers) will be eligible. Both open and minimally invasive (laparoscopic, robotic) surgical approaches will be included.

14. Primary Outcome(s)

(1) Recurrence-free survival (RFS) or disease-free survival (DFS), reported as hazard ratio (HR) with 95% confidence interval, comparing ctDNA-positive versus ctDNA-negative patients following curative-intent resection. (2) Correlation between ctDNA positivity and pathological surgical margin status (R0 versus R1/R2 resection), reported as odds ratio (OR) or risk ratio (RR).

15. Secondary Outcome(s)

(1) Overall survival (OS) stratified by ctDNA status. (2) Sensitivity and specificity of ctDNA for detecting positive surgical margins. (3) Lead time of ctDNA detection of recurrence compared to conventional imaging or tumor markers (e.g., CEA, CA19-9). (4) Subgroup analyses by: cancer type (colorectal, gastric, pancreatic, esophageal, hepatobiliary); ctDNA sampling time point (preoperative vs. postoperative); and assay platform (tumor-informed ddPCR vs. tumor-agnostic NGS).

16. Data Extraction (Selection and Coding)

Two independent reviewers will screen titles and abstracts against pre-specified eligibility criteria. Full-text articles will then be assessed independently by the same two reviewers. Disagreements will be resolved by discussion and, if necessary, arbitration by a third reviewer. A standardized data extraction form (Microsoft Excel) will be used to collect: first author, year, country, study design, cancer type, sample size, ctDNA assay method and platform, timing of ctDNA sampling, reported outcomes with effect estimates (HR, OR, RR) and 95% confidence intervals, adjustment covariates, and follow-up duration. Authors will be contacted by email for missing or unclear data.

17. Risk of Bias (Quality) Assessment

Risk of bias for prognostic studies will be assessed using the Quality in Prognosis Studies (QUIPS) tool, evaluating six domains: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting. For studies reporting diagnostic accuracy of ctDNA against histopathological margin assessment, the Quality Assessment of Diagnostic Accuracy Studies tool (QUADAS-2) will be applied. Assessment will be performed independently by two reviewers with discordances resolved by consensus.

18. Strategy for Data Synthesis

Where two or more studies report the same outcome with extractable effect estimates, data will be pooled using a random-effects model (DerSimonian and Laird method) to account for expected clinical and methodological heterogeneity. Pooled HRs will be calculated for survival outcomes; pooled ORs or RRs for margin status. Statistical heterogeneity will be quantified using the I² statistic (low <25%, moderate 25–75%, high >75%) and Cochrane Q test. If I² exceeds 75%, results will be reported narratively rather than pooled. Pre-specified subgroup analyses will be conducted by cancer type, sampling time point, and assay method. Publication bias will be assessed using funnel plots and Egger's test when ≥10 studies are available for an outcome. A sensitivity analysis will be performed by sequentially excluding one study at a time and by excluding studies with high risk of bias. All analyses will be performed in RevMan 5.4 (Cochrane) and R (metafor package).

19. Analysis of Subgroups or Subsets

Pre-specified subgroup analyses include: (1) cancer type (colorectal vs. gastric vs. pancreatic vs. esophageal vs. hepatobiliary); (2) ctDNA sampling time point (preoperative vs. early postoperative [within 4 weeks] vs. surveillance); (3) ctDNA assay platform (tumor-informed vs. tumor-agnostic); (4) study quality (high-quality studies scoring ≥7 on Newcastle-Ottawa Scale vs. lower quality). Subgroup differences will be tested using the chi-squared test for interaction.

20. Type and Method of Review

21. Language

22. Country

23. Other Registration Details

24. Reference for Protocol

25. Dissemination Plans

The completed systematic review and meta-analysis will be submitted for publication in a peer-reviewed international journal. Results will be reported in accordance with the PRISMA 2020 statement.

26. Conflict of Interest

The review authors declare no conflicts of interest. No external funding has been obtained for this review.

27. Review Team Members

• Your name
• Your institution (your medical school or hospital)
• Your email
• Your role: "Review conception, protocol design, data screening, extraction, analysis, manuscript writing"

After You Submit

• You'll receive a PROSPERO ID number by email (usually within 2-5 working days)
• Save it — it looks like: CRD42026XXXXXX
• Put it in your paper's Methods section like this:

"This systematic review was prospectively registered on PROSPERO (registration number CRD42026XXXXXX)."

Quick Checklist Before You Hit Submit

• Title includes "systematic review and meta-analysis"
• PICO question is clearly written out
• All 4 databases listed in the Searches field
• Both primary outcomes stated clearly
• Risk of bias tools named (QUIPS and QUADAS-2)
• Statistical method stated (random-effects, RevMan)
• Subgroup analyses listed
• Dissemination plan filled in
• Your name and institution added as team member

Updated PROSPERO Registration — Plan B

Topic: Postoperative ctDNA for MRD Detection and Recurrence Prediction After Curative Resection Across GI Cancers

What Changes From the Previous Version

• When: Postoperative ctDNA only (after surgery, not intraoperative or pre-op as primary)
• What it detects: Minimal Residual Disease (MRD) — tiny amounts of cancer left behind that imaging can't see
• What it predicts: Cancer coming back (recurrence), recurrence-free survival, overall survival
• Who: All GI cancer types together in one pan-GI analysis

Pre-Screening Questions (Same as Before)

• Submitting in English? → Yes
• Scoping/mapping review? → No
• Outcome related to human health? → Yes
• Cochrane review? → No
• Training course? → No
• Searched for similar existing reviews? → Yes
• Have a protocol? → Yes
• More than one person involved? → Yes
• Intend to publish? → Yes
• Preliminary searches started? → Started
• Screening started? → Not started

THE FULL FORM — Updated for Plan B

1. Review Title

Postoperative circulating tumor DNA (ctDNA) for minimal residual disease detection and recurrence prediction following curative-intent resection in gastrointestinal cancers: a systematic review and meta-analysis

2. Original Language Title

3. Anticipated Start Date

4. Anticipated Completion Date

5. Stage of Review at Time of Registration

6. Review Question ⭐ Most Important Field

In adults with histologically confirmed gastrointestinal cancers (colorectal, gastric, pancreatic, esophageal, hepatocellular carcinoma, and cholangiocarcinoma) who have undergone curative-intent surgical resection, does postoperative detection of circulating tumor DNA (ctDNA) in peripheral blood identify minimal residual disease (MRD) and predict disease recurrence — measured as recurrence-free survival (RFS), disease-free survival (DFS), and overall survival (OS) — compared to patients with undetectable postoperative ctDNA?

7. Searches

The following electronic databases will be searched from inception to the date of search with no date restrictions: MEDLINE via PubMed, Embase via Ovid, Cochrane Central Register of Controlled Trials (CENTRAL), and Web of Science Core Collection. The search strategy will use three concept blocks joined by AND: (1) ctDNA/liquid biopsy/MRD terms; (2) gastrointestinal cancer type terms (colorectal, gastric, pancreatic, esophageal, hepatocellular, cholangiocarcinoma); (3) postoperative/resection/recurrence terms. Both MeSH headings and free-text synonyms will be applied in each database. Reference lists of all included studies will be hand-searched. ClinicalTrials.gov and the WHO International Clinical Trials Registry Platform will be searched for completed but unpublished studies. Grey literature will be reviewed via Google Scholar (first 200 results).

8. Condition or Domain Being Studied

Gastrointestinal malignancies including colorectal cancer (colon and rectal), gastric cancer, pancreatic ductal adenocarcinoma, esophageal cancer (adenocarcinoma and squamous cell carcinoma), hepatocellular carcinoma, and cholangiocarcinoma (intrahepatic and extrahepatic), in patients who have undergone surgical resection with curative intent.

9. Population

Adults aged 18 years or older with histologically or cytologically confirmed gastrointestinal cancer who underwent curative-intent surgical resection (R0 resection or attempted curative resection). Studies must include postoperative blood sampling for ctDNA. Studies limited to metastatic, unresectable, or palliative patients will be excluded. Pediatric studies will be excluded.

10. Intervention / Exposure ⭐

Postoperative detection of circulating tumor DNA (ctDNA) in peripheral blood (plasma or serum), measured at any time point after curative-intent surgical resection. This includes early postoperative sampling (within 4–8 weeks of surgery), landmark time point sampling (e.g., 3, 6, 12 months), and serial/longitudinal ctDNA surveillance. Any validated assay platform will be eligible, including tumor-informed digital droplet PCR (ddPCR), personalized NGS-based panels (e.g., Signatera, FoundationOne Tracker), and tumor-agnostic methylation-based cfDNA assays. Studies using preoperative ctDNA only, without a postoperative measurement, will be excluded.

11. Comparator(s) / Control(s)

Patients with undetectable (negative) postoperative ctDNA at the corresponding time point following curative-intent resection of the same GI cancer type.

12. Types of Study to Be Included

Prospective cohort studies, retrospective cohort studies, case-control studies, and randomized controlled trials that report postoperative ctDNA measurement with at least one eligible outcome (RFS, DFS, or OS) in GI cancer patients after curative resection. Minimum sample size of 20 patients. Studies must provide extractable effect estimates (HR, OR, or RR with 95% confidence intervals) or Kaplan-Meier survival data sufficient for estimate derivation. Excluded: conference abstracts without full-text data, review articles, editorials, letters, case reports, animal studies, in vitro studies.

13. Context

Studies from any country and healthcare setting are eligible. Both open and minimally invasive surgical approaches (laparoscopic, robotic) will be included. Studies incorporating ctDNA alongside neoadjuvant or adjuvant chemotherapy are eligible provided postoperative ctDNA data is reported separately and extractable.

14. Primary Outcome(s) ⭐

(1) Recurrence-free survival (RFS) or disease-free survival (DFS) stratified by postoperative ctDNA status (positive vs. negative), reported as hazard ratio (HR) with 95% confidence interval.

(2) Detection of minimal residual disease (MRD) by postoperative ctDNA and its association with clinical recurrence — specifically, the sensitivity and specificity of postoperative ctDNA positivity for predicting eventual disease recurrence.

15. Secondary Outcome(s)

(1) Overall survival (OS) stratified by postoperative ctDNA status (HR with 95% CI).

(2) Lead time advantage of ctDNA-detected recurrence versus standard surveillance methods (imaging, CEA, CA19-9) — reported as months of earlier detection.

(3) ctDNA clearance dynamics: association between ctDNA becoming undetectable after adjuvant therapy and improved survival outcomes.

(4) Subgroup analyses:

• By GI cancer type (colorectal / gastric / pancreatic / esophageal / hepatocellular carcinoma / cholangiocarcinoma)
• By postoperative ctDNA sampling time point (early [≤8 weeks post-surgery] vs. late [>8 weeks])
• By ctDNA assay platform (tumor-informed ddPCR vs. tumor-agnostic NGS vs. methylation-based)
• By cancer stage at surgery (stage I-II vs. stage III)
• By receipt of adjuvant chemotherapy (yes vs. no)

16. Data Extraction (Selection and Coding)

Two reviewers will independently screen all titles and abstracts identified by the search against pre-specified eligibility criteria using Rayyan (rayyan.ai). Full-text articles of all potentially eligible records will be independently assessed by both reviewers. Disagreements at either stage will be resolved by discussion, with arbitration by a third reviewer if consensus is not reached.

Data will be extracted using a standardized Microsoft Excel form capturing: first author, publication year, country, study design, cancer type and stage, total sample size, number of ctDNA-positive and -negative patients, ctDNA assay platform and technology, timing of postoperative ctDNA sampling, follow-up duration, reported effect estimates (HR, OR, RR) with 95% CI, variables adjusted for in multivariable analyses, and recurrence rates. If Kaplan-Meier curves are the only reported data, HRs will be estimated using the method described by Tierney et al. (2007). Corresponding authors will be contacted by email for missing or unclear data, with up to two contact attempts.

17. Risk of Bias (Quality) Assessment

Risk of bias will be assessed independently by two reviewers.

For prognostic factor studies (the majority of included studies): the Quality in Prognosis Studies (QUIPS) tool will be used, evaluating six domains — study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, and statistical analysis and reporting.

For studies reporting ctDNA as a diagnostic test for MRD detection: the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool will be applied across four domains — patient selection, index test, reference standard, and flow and timing.

Overall certainty of evidence for each outcome will be assessed using the GRADE (Grading of Recommendations, Assessment, Development and Evaluations) framework, producing a summary of findings table.

18. Strategy for Data Synthesis ⭐

A random-effects model (DerSimonian and Laird method) will be used for all pooled analyses to account for anticipated clinical and methodological heterogeneity across GI cancer types, assay platforms, and sampling time points.

For survival outcomes (RFS/DFS and OS), log-transformed HRs and their standard errors will be pooled. For binary outcomes (MRD detection rates), pooled ORs or RRs will be calculated. Statistical heterogeneity will be quantified using the I² statistic and Cochrane Q test, interpreted as: I² <25% low, 25–75% moderate, >75% high. Where I² exceeds 75% and a clear clinical explanation cannot be identified through subgroup analysis, results will be presented narratively without statistical pooling.

Publication bias will be assessed using funnel plot visual inspection and Egger's weighted regression test, applicable when ≥10 studies contribute to a single analysis.

Sensitivity analyses will include: (1) sequential leave-one-out analysis to assess influence of individual studies; (2) exclusion of studies with high risk of bias (QUIPS overall rating); (3) restriction to studies using only tumor-informed assay platforms; (4) restriction to prospective studies only.

All statistical analyses will be performed using RevMan 5.4 (Cochrane Collaboration) and R (version 4.x, metafor and meta packages).

19. Analysis of Subgroups or Subsets

The following subgroup analyses are pre-specified:

1. GI cancer subtype: colorectal cancer vs. gastric cancer vs. pancreatic ductal adenocarcinoma vs. esophageal cancer vs. hepatocellular carcinoma vs. cholangiocarcinoma
2. Postoperative sampling time point: early (≤8 weeks post-surgery) vs. late (>8 weeks)
3. ctDNA assay platform: tumor-informed (ddPCR or personalized NGS) vs. tumor-agnostic (methylation-based or generic panel)
4. Disease stage: early stage (I–II) vs. locally advanced (III)
5. Adjuvant chemotherapy: received vs. not received
6. Study quality: high quality (QUIPS low risk in ≥4 of 6 domains) vs. lower quality

Subgroup differences will be evaluated using the chi-squared test for interaction (p < 0.10 considered significant). Subgroup analyses will only be performed where ≥3 studies contribute per subgroup.

20. Type and Method of Review

• Type: Systematic review
• Method: Meta-analysis
• Health area: Cancer

21. Language

22. Country

23. Other Registration Details

24. Reference for Protocol

25. Dissemination Plans

The completed systematic review and meta-analysis will be submitted to a peer-reviewed international oncology or surgical journal. Reporting will adhere to the PRISMA 2020 guidelines. Results will be presented at regional or national surgical/oncology conferences where possible.

26. Conflict of Interest

The review authors declare no conflicts of interest. No external funding has been received or sought for this review.

27. Review Team Members

• Name: Your full name
• Institution: Your medical school / hospital
• Email: Your university email
• Role: "Review conception, protocol design, literature search, title/abstract screening, data extraction, quality assessment, statistical analysis, manuscript preparation"

Your Final Pre-Submit Checklist

• Title says "systematic review and meta-analysis"
• Review question is one clear sentence with all PICO elements
• "Postoperative" is specified — preoperative-only studies excluded
• All 6 GI cancer types named (colorectal, gastric, pancreatic, esophageal, HCC, cholangiocarcinoma)
• Both primary outcomes written (RFS/DFS as HR, and MRD sensitivity/specificity)
• All 5 secondary outcomes and subgroup analyses listed
• QUIPS and QUADAS-2 both named as quality tools
• GRADE mentioned for certainty of evidence
• Random-effects model specified (DerSimonian-Laird)
• RevMan 5.4 and R named as analysis software
• Sensitivity analyses described (leave-one-out, exclude high-risk-of-bias)
• Dissemination plan filled in
• Your name and institution added

One Thing to Note About "MRD" in Your Title