Systematic Review

State the registration database, registration number, and date. Note any protocol amendments and their rationale.

Template sentences:

"This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines [REF]. The study protocol was prospectively registered with the International Prospective Register of Systematic Reviews (PROSPERO; registration number: CRD[XXXXXXXX], registered [DATE])."

"The study protocol was pre-registered on the Open Science Framework (OSF; https://osf.io/[ID], registered [DATE])."

Checklist:

• Registration number and database (PROSPERO, OSF, INPLASY)
• Protocol availability and access location
• Amendments from protocol with rationale

2.2 Eligibility Criteria

Define inclusion and exclusion criteria using PICO/PICOTS:

• P — Population (age range, clinical setting, diagnosis, geographical scope)
• I — Intervention or exposure (type, dose, timing)
• C — Comparator (unvaccinated, placebo, alternative)
• O — Outcomes (primary outcome defined precisely; secondary outcomes listed)
• T — Time (study period, follow-up duration)
• S — Study design (observational only? RCTs only? Both?)

Template sentence:

"Studies were eligible if they enrolled [POPULATION], assessed [INTERVENTION/EXPOSURE] compared with [COMPARATOR], and reported [OUTCOME] with a minimum follow-up of [DURATION]. Both [RCTs and observational studies / observational studies only] were included."

Checklist:

• All PICOTS elements stated
• Language restrictions stated (or absence thereof)
• Geographical restrictions stated (or absence thereof)
• Exclusion criteria enumerated

2.3 Information Sources and Search Strategy

State all databases searched, grey literature sources, date of last search, and search strategy availability.

Template sentence:

"A systematic literature search was conducted in MEDLINE via PubMed, Embase via Elsevier, the Cochrane Central Register of Controlled Trials (CENTRAL), and [OTHER] from inception to [DATE]. The full electronic search strategy for [DATABASE] is provided in Supplementary Material [X]. Reference lists of included studies and relevant reviews were hand-searched."

Checklist:

• All databases named with platform (e.g., "MEDLINE via PubMed")
• Grey literature sources specified
• Date of last search stated
• Full strategy in supplement for at least one database
• Hand-searching and citation chaining described

2.4 Study Selection

Describe the screening process, number of reviewers, software used, and conflict resolution.

Template sentence:

"Titles and abstracts were independently screened by two reviewers ([INITIALS] and [INITIALS]) using [Rayyan / Covidence / other]. Full-text articles were assessed against the eligibility criteria by the same reviewers independently. Discrepancies were resolved by consensus or, when necessary, by consultation with a third reviewer ([INITIALS])."

Checklist:

• Number of reviewers stated
• Screening software named
• Conflict resolution process described
• Pilot testing of screening criteria mentioned (if performed)

2.5 Data Extraction

Describe the data extraction form, pilot testing, duplicate extraction, and variables extracted.

Template sentence:

"Data were extracted independently by two reviewers using a standardised, pilot-tested data extraction form. Extracted variables included: first author, publication year, country, study design, population characteristics (N, age, sex), [exposure/intervention] definition, [comparator] definition, outcome definition, follow-up duration, effect estimate (with 95% CI), and confounders adjusted for."

Checklist:

• Form described (standardised, pilot-tested)
• Duplicate extraction stated
• All extracted variables enumerated
• Process for contacting study authors described (if applicable)

2.6 Risk of Bias Assessment

State the tool used, domains assessed, and process (independent, consensus).

Template sentence:

"Risk of bias was assessed independently by two reviewers using the [TOOL NAME]. For RCTs, the Cochrane Risk of Bias 2 (RoB 2) tool was used, covering five domains: randomisation process, deviations from intended interventions, missing outcome data, measurement of the outcome, and selection of reported results. For observational studies, the [NOS / ROBINS-I] was used. Disagreements were resolved by consensus. An overall risk-of-bias judgement (low, some concerns, high) was assigned to each study."

Checklist:

• Tool named for each study design included
• Domains listed
• Independent assessment stated
• Overall judgement categories defined

2.7 Effect Measures

Define the primary effect measure, any transformations, and the VE formula if applicable.

Template sentence:

"The primary effect measure was the [odds ratio (OR) / relative risk (RR) / hazard ratio (HR)], with 95% confidence intervals. Effect estimates were log-transformed for meta-analysis and back-transformed for presentation."

VE formula (for vaccine studies):

"Vaccine effectiveness was calculated as VE% = (1 - effect estimate) x 100%, where the effect estimate was the odds ratio, relative risk, or hazard ratio from each study, as reported."

Checklist:

• Primary effect measure defined
• Transformations stated (log-scale)
• VE formula stated (if applicable)
• Absolute vs relative VE distinguished (if applicable)

2.8 Synthesis Methods

This is the most complex subsection. Cover model type, estimation, heterogeneity, subgroup analyses, meta-regression, sensitivity analyses, and publication bias.

Meta-analytic model:

"A random-effects model with restricted maximum likelihood (REML) estimation was used to pool effect estimates. Standard errors were adjusted using the Knapp-Hartung (HKSJ) method to account for uncertainty in the between-study variance estimate."

Three-level meta-analysis (v1.0.0):

"A three-level random-effects meta-analytic model was fitted using the rma.mv function in the metafor package (version [X.X]) in R (version [X.X]), with effect sizes (level 2) nested within studies (level 3). This structure accounts for within-study correlation between repeated estimates from the same cohort."

Robust variance estimation (v1.0.0):

"Robust variance estimation (RVE) with small-sample corrections was applied to account for dependent effect sizes within studies, using the clubSandwich package in R."

Heterogeneity:

"Heterogeneity was assessed using Cochran's Q statistic (with p < 0.10 indicating significant heterogeneity), the I-squared statistic (interpreted as low < 25%, moderate 25-75%, high > 75%; Higgins et al.), the between-study variance tau-squared, and the 95% prediction interval."

Subgroup analyses:

"Pre-specified subgroup analyses were conducted by [VARIABLES]. Subgroup differences were assessed using a chi-squared test of interaction."

Meta-regression:

"Meta-regression was conducted using REML estimation with HKSJ-corrected standard errors to evaluate [COVARIATES] as potential sources of heterogeneity."

Publication bias:

"Publication bias was assessed visually using funnel plots and statistically using Egger's weighted regression test for funnel plot asymmetry (where >= 10 studies were available). The trim-and-fill method was applied to estimate the number and direction of potentially missing studies."

Sensitivity analyses:

"Pre-specified sensitivity analyses included: (1) exclusion of studies rated as high risk of bias; (2) exclusion of [SPECIFIC CRITERIA]; (3) use of fixed-effects model; (4) [OTHER]. Each sensitivity analysis was compared with the base-case estimate to assess robustness."

Dose-response cross-reference (v1.0.0): For dose-response meta-analysis methods, see skills/study-types/dose-response/SKILL.md.

Checklist:

• Model type stated (random-effects, fixed-effects, multilevel)
• Estimation method stated (REML, DL, HKSJ)
• Heterogeneity statistics listed (Q, I-squared, tau-squared, 95% PI)
• Subgroup variables listed and pre-specified vs post-hoc stated
• Meta-regression covariates and estimation described
• Publication bias method stated (Egger's, funnel, trim-and-fill)
• Each sensitivity analysis listed with rationale
• Software and version cited

Section 3 — Results Conventions

Subsection Order

1. Study Selection — PRISMA flow diagram
2. Study Characteristics — summary table
3. Risk of Bias — domain-level and overall summary
4. Primary Meta-Analysis — pooled estimate with full heterogeneity reporting
5. Sensitivity Analyses — alternative estimates
6. Subgroup Analyses — stratified estimates with interaction tests
7. Secondary Outcomes — additional meta-analyses

Study Selection

Report total records identified (by database), records after deduplication, records screened (title/abstract), full-texts assessed, studies included, and reasons for exclusion. Reference the PRISMA flow diagram.

Study Characteristics Table

Required columns: First author, Year, Country, Design, Population (N), Exposure/Intervention, Comparator, Outcome definition, Follow-up duration, Confounders adjusted.

Risk of Bias

Report domain-level judgements and overall judgement for each study. Summarise the distribution of RoB ratings across included studies.

Primary Pooled Estimate

Report all of the following for every meta-analysis:

1. Number of studies (k) and total participants (N)
2. Model type and estimator (e.g., random-effects, REML, HKSJ)
3. Pooled point estimate with 95% CI
4. Cochran's Q statistic (df, p-value)
5. I-squared with 95% CI if available
6. Tau-squared with 95% CI via profile likelihood if available
7. 95% Prediction interval (PI)
8. Cross-reference to forest plot: (Figure X)

Three-level model additional reporting:

• Sigma-squared-between (between-cluster variance) with I-squared-between
• Sigma-squared-within (within-cluster variance) with I-squared-within
• Profile likelihood CIs for variance components
• Justification for three-level structure

Interpreting heterogeneity:

• I-squared < 25%: low heterogeneity
• I-squared 25-75%: moderate heterogeneity
• I-squared > 75%: high heterogeneity
• Always pair I-squared with the 95% PI; I-squared can be low even when PI is wide (small k problem)

Vaccine Effectiveness Reporting

VE% = (1 - theta) x 100, where theta is the OR, RR, or HR. Back-transform from log-scale before reporting. Explicitly distinguish absolute VE (vs unvaccinated) from relative VE (vs alternative vaccine) throughout.

Waning Analysis

Report slope (log-odds per year), p-value, between-study variance (sigma-squared-between, I-squared-between), within-study variance (sigma-squared-within, I-squared-within). Report model-predicted VE at key time points. State clearly what the prediction interval implies at each time horizon. Distinguish interpolated (within data range) from extrapolated (beyond data range) predictions.

Sensitivity Analyses

Report each alternative estimate with its own CI and I-squared. State the direction of change relative to the base case. Do not over-interpret small numerical differences.

Subgroup Analyses

Always report the test for subgroup difference (chi-squared statistic, degrees of freedom, p-value). Do not claim a subgroup effect without a statistically significant interaction test. Pre-specified subgroups carry higher evidential weight than post-hoc.

Meta-Regression

Report: moderator variable(s) and coding, slope coefficient with SE, 95% CI, and p-value, intercept with SE, 95% CI, and p-value, R-squared analogue (% variance explained), residual heterogeneity after moderator.

Prose Style

• Write in full paragraphs, never bullet points in the final manuscript
• Past tense throughout
• Cross-reference all figures and tables inline: (Figure 1), (Table 2)
• Do not cite external references in the Results section
• UK English spelling

Section 4 — Discussion Considerations

Main Findings — Opening Paragraph Template

"This systematic review and meta-analysis of [N] studies -- [N] [design] and [N] [design], published between [years] across [countries] -- found that [intervention] was associated with [primary outcome] of [point estimate] (95% CI: [lower], [upper]; I-squared = [X]%; 95% prediction interval, [lower] to [upper]). [Secondary finding 1, with statistics]. [Null result or non-significant trend, phrased proportionately]. [Waning / subgroup result, with appropriate caveat if from single study]."

Literature Comparison Framing

Consistent:

"The pooled estimate of [X]% observed in the present review is broadly consistent with prior evidence. [Author] and colleagues, in a [SRMA / cohort study] of [scope], reported [comparable estimate] (95% CI: [lower], [upper])..."

Divergent:

"In contrast to [Author] et al., who reported [estimate], the present review found [different estimate]. This divergence may reflect [methodological difference / population difference / time period / comparator heterogeneity]..."

Extending:

"The present review extends the work of [Author] et al. by [methodological contribution -- e.g., separating absolute from relative VE / applying three-level meta-regression / restricting to a specific age group]..."

Typical Strengths

• Pre-specified registered protocol reduces post-hoc bias
• Comprehensive multi-database search reduces publication bias
• Three-level meta-regression accounts for within-study correlations between repeated estimates
• Explicit separation of absolute vs relative VE (if applicable)
• HKSJ adjustment provides appropriate interval widths under small-study conditions
• Rigorous heterogeneity assessment (tau-squared, I-squared, PI) informs generalisability
• Multiple pre-specified sensitivity analyses test robustness
• Design-appropriate risk of bias tools (RoB 2, ROBINS-I, NOS)
• PRISMA 2020 adherence ensures transparent reporting

Template:

"This review was conducted against a pre-specified, registered protocol, reducing the risk of post-hoc analytical decisions inflating spurious positive findings."

"The Hartung-Knapp-Sidik-Jonkman adjustment was applied to all confidence intervals, providing more appropriate interval widths under the small-study conditions (k < 10) prevalent in this literature."

Typical Limitations

For each limitation, state: (1) the limitation precisely, (2) likely direction of bias, (3) impact on conclusions.

• Small k / limited power: "The base-case meta-analysis included only [k] studies, and the waning meta-regression was based on [N] paired estimates from [N] studies, affording limited statistical power. Wide confidence and prediction intervals throughout reflect this imprecision."
• Temporal heterogeneity: "The pooled estimate represents an average across studies with markedly different post-vaccination follow-up periods, ranging from [X] to [X] years."
• Observational study bias: "All [N] observational studies were at moderate to serious risk of bias from unmeasured confounding. Healthy vaccinee bias may have inflated effectiveness estimates to an unknown degree."
• Comparator heterogeneity: "Differences in comparator vaccines across studies precluded direct pooling of relative and absolute VE estimates."
• Single-study subgroup: "Age-stratified data were available from a single study, precluding replication of the age-differential finding."
• Geographical restriction: "Evidence was restricted to [N] high-income countries; generalisability to LMIC settings is uncertain."
• Midpoint assumption: "Time since vaccination was derived using the midpoint of study periods; misclassification may have attenuated the slope estimate."

Section 5 — Software Citations

R:

R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. [Year]. Available from: https://www.R-project.org/

metafor (R package):

Viechtbauer W. Conducting meta-analyses in R with the metafor package. J Stat Softw. 2010;36(3):1-48. doi:10.18637/jss.v036.i03

Rayyan (screening):

Ouzzani M, Hammady H, Fedorowicz Z, Elmagarmid A. Rayyan -- a web and mobile app for systematic reviews. Syst Rev. 2016;5:210. doi:10.1186/s13643-016-0384-4

Covidence (screening):

Covidence systematic review software, Veritas Health Innovation, Melbourne, Australia. Available from: www.covidence.org

RevMan (meta-analysis):

Review Manager (RevMan) [Computer program]. Version [X.X]. The Cochrane Collaboration. [Year].

PROSPERO registration template:

"This systematic review and meta-analysis was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA 2020) guidelines [REF]. The study protocol was prospectively registered with PROSPERO (registration number: CRD[XXXXXXXX], registered [DATE])."

OSF registration template:

"The study protocol was pre-registered on the Open Science Framework (OSF; https://osf.io/[ID], registered [DATE])."

clubSandwich (robust variance estimation):

Pustejovsky JE. clubSandwich: Cluster-Robust (Sandwich) Variance Estimators with Small-Sample Corrections. R package. Available from: https://CRAN.R-project.org/package=clubSandwich