How to Read a Scientific Study
PICO, Study Design, and the Numbers That Matter
How to Read a Scientific Study
Most APPs read the abstract conclusion first and work backward. That is the wrong order. The conclusion is the authors’ opinion. The methods determine whether it is credible. This card gives you a system for reading any study in under five minutes.
Part 1 — The PICO Framework: Read the Abstract in 60 Seconds
Every study asks a question. PICO is the structure of that question. If you cannot find all four elements in the abstract, the study is not well reported.
| Letter | Element | What to Look For |
|---|---|---|
| P | Population | Who was enrolled? Age, comorbidities, exclusion criteria. Is this your patient? |
| I | Intervention | What was done, at what dose, for how long? |
| C | Comparison | What was the control? Placebo, usual care, active comparator, or nothing? |
| O | Outcome | What was measured? Primary endpoint, secondary endpoints. Is this a hard outcome (death, MI) or a surrogate (biomarker, imaging)? |
Part 2 — Study Design Hierarchy
Study design determines what a paper can and cannot prove. Hierarchy runs from weakest to strongest for establishing causation.
| Design | Can Prove | Cannot Prove |
|---|---|---|
| Case report / Case series | Signal detection, rare adverse events | Causation, generalizability |
| Cross-sectional | Prevalence, associations at one point in time | Cause and effect (no temporality) |
| Case-control | Association, odds ratios; good for rare diseases | Incidence; subject to recall bias |
| Cohort (observational) | Incidence, risk factors over time | Causation (confounding cannot be fully eliminated) |
| RCT | Causation (randomization controls confounding) | Efficacy in populations excluded from trial |
| Meta-analysis / Systematic review | Pooled effect size, consistency across trials | Quality above the included studies |
Practical rule: An observational study cannot establish that a treatment works. It can generate a hypothesis. Only a well-conducted RCT with hard outcomes can change practice.
Part 3 — Numbers That Matter
| Metric | What It Means | Red Flag |
|---|---|---|
| Effect size | Magnitude of the treatment effect (HR, RR, OR, mean difference) | Statistically significant but clinically trivial |
| Confidence interval | Range of plausible true effects at 95% confidence | Wide CI = imprecise estimate; crosses null = not significant |
| p-value | Probability of seeing this result if the null were true | p = 0.049 vs. 0.051 is not a meaningful distinction |
| NNT | Patients you must treat to prevent one event | NNT >100 over 5 years rarely justifies risk or cost |
| Follow-up duration | How long patients were observed | Short follow-up misses late harms and late benefits |
| Lost to follow-up | Patients who dropped out or were excluded from analysis | >10–20% attrition threatens validity |
Internal validity = was the study well conducted? (randomization, blinding, ITT analysis, low dropout)
External validity = does it apply to your patient? (age, comorbidities, excluded populations, real-world dose)
Read the Methods before the Conclusion. The conclusion is the authors’ opinion — the methods tell you whether to believe it. An RCT with 30% dropout reported per-protocol is not a clean RCT. A p < 0.05 in a composite endpoint driven by soft outcomes is not practice-changing.
This is one of 13 free reference sheets from the APP Cardiology Academy — no account required.