Describe the treatment effect
Choose group counts or a published estimate. Use Worked Examples for sourced or hypothetical walkthroughs.
Comparator can mean an active treatment, placebo, or usual care. Use a rate relevant to the same outcome and time frame. Applying a pooled or adjusted ratio produces a modeled scenario; it need not reproduce observed study percentages.
No SD is needed for the distribution illustration. To express an SMD in scale points, use an SD from an appropriate population on the specific scale. A meta-analysis across different scales has no single scale SD; do not average those SDs.
Enter signed SMD as treatment mean minus control mean, divided by SD. Positive means a higher score, not necessarily benefit. The selected binary outcome direction does not determine the meaning of a score.
N is recorded but does not determine missing event rates, standard deviations, or confidence intervals. Enter the study’s reported CI when available.
Group-count calculations assume independent parallel groups and one binary outcome per person at the same follow-up. Published-effect conversions preserve the supplied estimate; they do not reconstruct counts or remove confounding.
Relative and absolute effects
Put this SMD in context: common medications
Historical examples from psychiatry and other fields
These are drug-versus-placebo estimates reported in Figure 1 of Leucht and colleagues’ How effective are common medications? (2015; literature search through 5 August 2014). They provide a sense of scale, not current treatment recommendations or a ranking of therapies.
Compare the outcome as well as the number. This overview reports some effects from continuous scores and others from binary outcomes on a standardized scale. A symptom-score SMD is not directly interchangeable with a standardized effect derived from response, remission, or mortality. Populations, follow-up, instruments, and study methods differ.
| Treatment | Outcome measured | SMD / 95% CI |
|---|
How to use this context
- A modest SMD can matter clinically. The familiar 0.2 / 0.5 / 0.8 descriptors are conventions, not universal thresholds for clinical importance.
- A large change in a surrogate does not ensure a large patient benefit. The metformin entries illustrate how glucose and mortality can tell different stories; the mortality interval includes zero.
- Effect magnitude does not summarize adverse effects, burden, durability, baseline risk, or patient preferences. These are comparisons across different studies, not head-to-head comparisons between the listed medications.
- The same disease can have different effects for symptom relief, response, relapse prevention, and mortality. Choose context that matches your question.
Source: Leucht S, Helfer B, Gartlehner G, Davis JM. BMC Medicine 2015;13:253, Figure 1 and accompanying text. Read the open-access paper ↗. Adapted under CC BY 4.0. Numerical SMDs and confidence limits are transcribed from the figure; wording, grouping, and presentation are adapted. This table is a historical reference, not a new evidence review.
Methods & limits
Risk difference = treatment risk − control risk. Its 95% CI uses Newcombe's unpaired score method (Wilson intervals; no continuity correction). RR = treatment risk / control risk. OR compares event odds. Ratio CIs use a log-normal approximation when all four cells are positive; otherwise no ratio CI is reported and no 0.5 correction applied.
NNTB / NNTH = 1 / absolute risk difference, rounded up, interpreted using outcome direction and follow-up. The interval inverts the risk-difference CI. If the CI includes zero, a single finite NNT interval is inappropriate.
Intervals reflect sampling uncertainty under the model, not bias, missing data, applicability, or confounding. Sparse-data ratio intervals may be unreliable. Cochrane guidance · Newcombe (1998).
Describe the test
Defaults are a hypothetical teaching example, not a named test.
Reference-standard disease status. Enter people, not repeated tests.
Target pretest probability stays independent of the study's disease fraction, which may reflect selected sampling.
Finite, nonnegative ratios. This mode updates probabilities without inferring sensitivity, specificity, or test-result frequencies.
Update the probability
Show the calculation & assumptions
Sensitivity = TP / (TP + FN). Specificity = TN / (TN + FP). LR+ = sensitivity / (1 − specificity). LR− = (1 − sensitivity) / specificity.
Point estimates only: diagnostic confidence intervals and uncertainty in pretest probability are not estimated in v1. The explorer holds performance fixed; actual performance may differ by population, setting, threshold, and reference standard. Do not multiply LRs from related findings without accounting for dependence.
Explore alternative base rates
Change starting probability while holding test sensitivity/specificity or likelihood ratios fixed.
Horizontal: pretest probability. Vertical: disease probability after the result. Marker line: current pretest probability. NPV is the complement of the negative-result curve.
Base-rate search helper
Find a population estimate, then judge whether it fits the clinical question. Population prevalence is not automatically an individual's pretest probability.
Separate required concepts with commas (AND). Use | for alternatives within a concept (OR), e.g., fatigue | exhaustion. Recognized variants of female, male, and fatigue expand automatically; other concepts stay as entered. These are search alternatives, not strict eligibility filters.
Refine PubMed search (optional)
Enter a condition phrase and separate population concepts with commas. Each population concept becomes an AND group, with alternatives joined by OR. Multiword concepts stay together. These tagged terms do not rely on PubMed automatic term mapping; the synonym list is limited. An optional confirmed condition MeSH heading adds indexed coverage. Geography and setting can still make searches too narrow.
Inspect generated PubMed query
PubMed search guidance ↗Generated searches only; no results or estimates are automatically retrieved or verified. Search terms leave this page only when you open a search link.