Stefano Garzon

Research · Oct 23, 2026

Reporting and interpreting results: CONSORT, plausibility, absolute effect and applicability

Concept

Publication in a peer-reviewed journal is not an endorsement of the results. Investigators know the limitations of the data better than reviewers and readers and have the obligation to report them. Friedman organizes interpretation around three questions: did the trial work as planned, how do the findings compare with other studies, and what are the clinical implications.

Reporting guidelines

CONSORT gathers the essential items of a randomized trial report: objectives, design, participants, interventions, primary and secondary outcomes, sample size, randomization, blinding, statistical analysis, baseline data, outcomes, harms, limitations and generalizability. The 2001 revised version had 22 items and a flow diagram with the number of participants per group at each stage, which lets the reader check whether the analysis was by intention to treat.

Friedman points to two loosely used terms. “Intention to treat” does not apply when randomized participants were excluded from the analysis, and “modified ITT” is a contradiction. “Per protocol” suggests the analysis preferred by the protocol; Friedman prefers “on treatment” (see intention to treat).

The p value, point estimate and confidence interval should be reported together for the main results.

Publication and reporting bias

Negative trials are published less often. Turner and colleagues compared 74 antidepressant studies registered with the FDA. 31% were not published. In the published literature, 94% of trials appeared positive; in the FDA analysis, 51%.

Chan and colleagues compared protocols and publications of 102 trials approved in Copenhagen. Half of the efficacy outcomes and 65% of the harm outcomes per trial were incompletely reported. According to Friedman, 62% of trials had a discrepancy between the primary outcome in the protocol and the one published.

Did the trial work as planned?

Plausibility and consistency

The result should be compatible with the known mechanism of action and with other studies. Credibility increases with the proportion of good-quality independent studies reaching the same conclusion. Confidence intervals allow readers to check whether results of different trials are compatible.

Absolute and relative effect

Presentation changes decisions. Forrow and colleagues presented the same results to 235 physicians as a relative change and as an absolute change. 108 (46%) responded differently, and 90% of these were more inclined to treat after reading the relative change. Naylor and colleagues observed the lowest effectiveness ratings when the result was presented as a number needed to treat. Friedman recommends reporting both absolute and relative effects (see RRR, ARR and NNT).

Applicability

Generalization to the eligible population is direct. Extrapolation to patients who would not meet the inclusion criteria requires judgment. Rothwell lists the determinants of external validity: setting, patient selection, characteristics of the intervention and control, outcomes and adverse events. Further questions include dose-response, class effect and newer versions of devices.

Common errors when reading a paper

What to check when reading a trial

  1. Trial registration and agreement between the registered and published primary outcome.
  2. Flow diagram with numbers by group at each stage.
  3. Estimate, confidence interval and absolute effect for the primary outcome.
  4. Adherence, losses to follow-up and sensitivity analysis for missing data.
  5. Total number of comparisons made.
  6. Comparison with other trials and biological plausibility.
  7. Similarity between the trial population and the patient at hand.

References