Strata Academy
GRADE Summary of Findings Tables Explained
How to read SoF tables in Cochrane reviews and guidelines — absolute vs relative effects, certainty ratings, and downgrade footnotes
Quick answer
A GRADE Summary of Findings (SoF) table presents one row per patient-important outcome: number of studies and participants, relative effect (e.g. RR), absolute effect (e.g. fewer events per 1,000), and certainty (high to very low) with footnotes explaining downgrades. Read the absolute effect and certainty before the p-value.
- Each outcome gets its own certainty rating — not one grade for the whole review.
- Absolute effects translate relative estimates into patient-meaningful numbers.
- Footnotes must name downgrade domains: bias, inconsistency, imprecision, indirectness, publication bias.
- SoF tables sit after meta-analysis and ROB judgements — they do not replace them.
1. What is a Summary of Findings table?
Summary of Findings (SoF) tables are the standard way systematic reviews and clinical guidelines present the bottom line for each outcome. Cochrane reviews, NICE technology appraisals, and WHO guidelines use SoF-style tables so readers can compare benefits, harms, and certainty without reading every forest plot.
An SoF table is not a decorative summary. It forces authors to state, per outcome: how many studies contributed, what the pooled effect is, what that means in absolute terms for a typical patient, and how confident we should be in the estimate.
Students often skip SoF tables and jump to the abstract conclusion. For coursework and journal club, reading the SoF table first prevents over-interpreting statistically significant but clinically trivial or imprecise effects.
- One row per patient-important outcome (mortality, quality of life, serious adverse events)
- Relative effect (RR, OR, HR, MD) paired with absolute effect where possible
- Certainty rating: high, moderate, low, or very low
- Footnotes document downgrade and upgrade decisions
2. Standard SoF columns and how to read them
Cochrane SoF tables typically include: outcome name, anticipated absolute effect in the comparator group (per 1,000 or per 100), relative effect with 95% CI, absolute effect in the intervention group, number of studies (participants), and certainty of evidence.
The comparator column anchors absolute effects. If the control risk is 50 deaths per 1,000, a relative risk reduction of 20% means roughly 10 fewer deaths per 1,000 — not a vague '20% improvement'.
When reviews cannot calculate absolute effects (heterogeneous baselines, continuous outcomes without meaningful thresholds), authors should explain why and still report relative effects with certainty.
| Column | What it tells you | Common student mistake |
|---|---|---|
| Outcome | Which endpoint this row addresses | Treating all rows as equally important clinically |
| Anticipated risk | Baseline event rate in control group | Ignoring that baseline risk changes absolute benefit |
| Relative effect | RR/OR/HR/MD with 95% CI | Reading RR alone without absolute translation |
| Absolute effect | Difference in events per 1,000 (or similar) | Assuming significant relative effect means large clinical benefit |
| Certainty | GRADE rating with symbols | Confusing statistical significance with high certainty |
| Footnotes | Why certainty was downgraded | Skipping footnotes entirely |
3. Relative vs absolute effects in SoF tables
Relative effects (risk ratios, odds ratios, hazard ratios) describe proportional change. Absolute effects describe how many additional or fewer events occur in a defined population over a defined time.
A relative risk of 0.80 sounds impressive, but if the baseline risk is 5 per 1,000, the absolute reduction is only 1 per 1,000. Guidelines and shared decision-making rely on absolute numbers.
For continuous outcomes (pain scores, HbA1c), SoF tables may report mean differences with minimally important difference thresholds. Ask whether the pooled MD crosses a clinically agreed threshold, not only whether p < 0.05.
Tip: When an SoF table omits absolute effects, calculate them from the control risk and relative effect — or note this as a reporting gap in your appraisal.
4. Certainty ratings and downgrade footnotes
GRADE certainty reflects confidence that the true effect is close to the estimate. RCT-dominated bodies start at high certainty; observational bodies start at low — then reviewers downgrade or upgrade.
Footnotes should name the domain: serious risk of bias in included trials, unexplained inconsistency, indirect PICO, wide confidence intervals or few events (imprecision), or suspected publication bias.
Symbols (⊕⊕⊕⊕ down to ⊕◯◯◯) are shorthand. The footnote text is what examiners and guideline panels need — 'moderate certainty' without rationale is incomplete appraisal.
- High — very confident true effect lies close to estimate
- Moderate — true effect probably close; further research may change estimate
- Low — true effect may differ substantially; further research likely important
- Very low — very uncertain; estimate unreliable for decision-making
Note: Do not treat 'moderate certainty' as a pass/fail grade for the whole review. Each outcome row has its own rating — mortality may be high certainty while quality of life is low.
5. Where SoF tables fit in the synthesis workflow
SoF tables are built after systematic review conduct: comprehensive search, screening, data extraction, risk-of-bias assessment (ROB 2, ROBINS-I), and synthesis (meta-analysis or narrative). GRADE certainty integrates those steps — it does not replace them.
Forest plots show study-level estimates and pooled CIs. SoF tables add certainty judgements and absolute effects for decision-makers. Read the forest plot for heterogeneity and study contributions; read the SoF for the clinical bottom line.
In dissertation write-ups, if you perform meta-analysis, include a simplified SoF table for primary outcomes even if your journal does not require full GRADE prose — supervisors expect outcome-level certainty reasoning.
6. Appraising SoF tables in published reviews
When appraising a systematic review, check whether SoF tables include all pre-specified primary outcomes from PROSPERO or the protocol. Outcome switching without SoF updates is a selective reporting red flag.
Cross-check certainty footnotes against ROB traffic-light plots and I² statistics. If authors rate high certainty despite serious ROB concerns or wide prediction intervals, document the inconsistency.
For guidelines citing SoF tables, trace whether recommendations are strong or conditional and whether certainty and effect magnitude support the wording.
- Primary outcomes from protocol all present in SoF?
- Downgrade footnotes match ROB and heterogeneity assessments?
- Absolute effects use plausible baseline risks?
- Harms and adverse events included, not only benefits?
7. Building a simplified SoF table for student projects
Full GRADE software (GRADEpro) is not always required for undergraduate dissertations, but a simplified table demonstrates synthesis literacy.
Include at minimum: outcome, studies (n), relative effect with CI, plain-language absolute effect, and a one-line certainty rationale citing bias, inconsistency, or imprecision.
Pair your SoF table with a forest plot for each pooled outcome and a ROB summary — examiners reward integrated reporting over isolated statistics.
Frequently asked questions
What is the difference between a forest plot and a Summary of Findings table?
A forest plot displays individual study estimates and a pooled effect with confidence intervals. An SoF table adds GRADE certainty ratings, absolute effects, and footnotes explaining downgrades — it is designed for clinical decision-making, not statistical detail alone.
Why do SoF tables show effects per 1,000 people?
Expressing absolute effects per 1,000 (or per 100) makes benefits and harms tangible for patients and policymakers. The denominator should match the review's population and follow-up duration where possible.
Can certainty be high if the pooled effect is not statistically significant?
Yes. High certainty with a confidence interval crossing the null means we are confident there is little or no meaningful effect — an important finding for ruling out benefit or harm.
Do I need GRADEpro for my dissertation SoF table?
Not always. Many UK medical schools accept a simplified SoF table with manual certainty judgements if you document downgrade reasons clearly. Check your supervisor's expectations and cite GRADE methods.
Why are different outcomes rated differently in the same review?
Each outcome has its own evidence base, heterogeneity, imprecision, and risk-of-bias profile. Mortality may rest on large low-bias trials (high certainty) while patient-reported outcomes may come from few small studies (low certainty).
Interactive walkthroughs and quizzes load when JavaScript is enabled — the checklist and tables above are fully readable without it.