Basic Statistics Every MSL Should Master

You do not need to be a statistician to be a great MSL. But you do need to read a results table and explain, in one clear sentence, what it means for a patient. Five concepts get you most of the way there.

KOLs will test your understanding of the evidence, and so will every MSL assessment panel. This guide covers the statistics you will meet in almost every clinical trial publication, with simple worked examples.

1. Endpoints and analysis populations

The primary endpoint is the outcome the trial was designed and powered to test. Secondary endpoints add context but are usually hypothesis-generating unless they were formally controlled for multiplicity.

Check which population was analysed. Intention-to-treat (ITT) analyses patients in the groups they were randomized to, preserving the benefits of randomization. Per-protocol analyses only those who followed the protocol, and is often used as a supportive or sensitivity analysis.

2. The p-value

A p-value is the probability of observing a result at least as extreme as the one seen, assuming there is truly no difference between treatments (the null hypothesis). By convention, p < 0.05 is called statistically significant.

3. Confidence intervals

A 95% confidence interval (CI) gives the range of values for the true effect that are compatible with the data. It tells you two things a p-value cannot:

• Size and precision: a narrow interval means a precise estimate, and a wide one means uncertainty.
• Significance: for a difference, an interval that crosses 0 is not statistically significant. For a ratio (such as a hazard ratio or relative risk), an interval that crosses 1 is not significant.

4. Relative versus absolute effects

Relative numbers sound bigger. Absolute numbers tell patients what changes for them. Always present both. A worked example, where 10% of control patients and 8% of treated patients have an event:

In plain words: you would need to treat 50 patients for one additional patient to avoid the event over the trial period.

5. Hazard ratios and Kaplan-Meier curves

Time-to-event endpoints, such as overall or progression-free survival, are usually analysed with a hazard ratio (HR). An HR of 0.70 means that, at any point during follow-up, patients on treatment had a 30% lower hazard (instantaneous risk) of the event than the control group. This assumes the hazards stay roughly proportional over time.

The Kaplan-Meier curve shows the proportion of patients event-free over time. Read the median from where each curve crosses 50%, check the number at risk below the curve and watch the tail, where few patients remain and estimates become less reliable. For a deeper dive, read our article on survival analysis and the Kaplan-Meier curve.

A 60-second checklist for any results table

1. What is the primary endpoint, and was it met?
2. What is the effect size, and what is its 95% CI?
3. What does it mean in absolute terms (ARR and NNT)?
4. Which population was analysed (ITT or per-protocol)?
5. Is the difference clinically meaningful, not just statistically significant?