Biostatistics Cheat Sheet for Busy Professionals

The most important skill to begin with is to know how to read clinical trials. Once you know what to look for, you can identify key information without having to read multiple pages. For this aim, you just need to follow these steps:

1. Identify the primary endpoint and the control group — this is crucial to understand exactly what is being tested and against what baseline standard.
2. Look at the p-value — this is important to assess whether findings are statistically significant (typically p < 0.05).
3. Contextualize this with the confidence interval (CI) — this is essential to determine the actual range of the treatment’s effect in real-life applications.
   • Narrow CI — equals a reliable estimate.
   • Wide CI — indicates highly unpredictable patient outcomes.
4. Compare the relative risk to the absolute risk reduction — this is the definitive measure of whether the study’s success actually translates into a meaningful benefit for the patient.

Once a trial is completed, researchers present one of their most well-known results: the p-value. It’s often treated as the ultimate proof: $p < 0.05$ means that something works and $p > 0.05$ means that it doesn’t. Though the p-value is an indicator, it actually shows something different.

A better way to think about the p-value is as a measure of how surprising the results are. It tells you how likely it is to see these findings if the drug actually had no effect at all. For example, a p-value of $0.01$ means that if the drug truly did nothing, there would only be a 1% chance of observing results this strong just by chance. That’s why it is considered statistically significant.

But this is where you need to be careful — statistical significance is not the same as clinical relevance. With a large enough study, e.g., 50,000 patients, even very small differences can produce extremely low p-values.

For example, a study might show that a weight-loss drug leads to an average loss of $0.4\ \rm lbs$ over six months, with a p-value of $0.001$. Statistically, the effect is significant. Clinically, it’s meaningless, as such a small change is unlikely to matter for your patient.

Once you’ve had a look at the p-value, it is time to focus on the confidence interval (CI). It provides additional information: the p-value focuses on how likely we would be to obtain results at least as extreme as those observed, assuming there is actually no true effect in the population, whereas the CI tells you how much the result actually matters.

Since it is not possible to test each and every individual, studies use a sample to estimate the situation in the general population. Thus, the 95% CI extends beyond statistical significance to show the range of plausible values for the true effect size, helping you judge its practical importance.

For example, if you find a narrow interval (e.g., 8 to 12) in your paper, this means that the effect of the drug is likely similar across patients. A wide interval (e.g., 1 to 19), however, is a red flag, as it shows that the effects of the drug vary largely amongst study participants.

The most critical quick check for a clinician is to look for the line of no effect. Here, you also need to be aware of the kind of data you’re dealing with, as there are different criteria for measured differences and ratios.

• In studies measuring a difference, that line is $0$; and
• In studies measuring a ratio, that line is $1.0$.

If the CI is $-2$ to $+5$ for the differences (e.g., measuring the drop in blood pressure), it means the drug might help, but it might also make things worse or do nothing at all.

If the CI for a new treatment’s risk ratio (e.g., relative risk) is $0.5$ to $1.2$, the result is technically not significant, because $1.0$ (no change) is a possibility.

Having the confidence interval explained, there is another critical value in clinical studies. Relative risk is one of the most commonly misunderstood statistics. It is the number that generates headlines like “Bacon increases cancer risk by 20%”.

What relative risk actually tells you is how much more or less likely an event is in one group compared to another. For example, a health condition occurs in 12 patients receiving a new treatment versus 10 patients receiving conventional treatment. The relative risk is $1.2$, and thus a 20% increase. But as clinicians, absolute risk is what matters.

If the condition normally affects $1$ in 1,000 patients, that 20% increase only raises the risk to $1.2$ in 1,000. That 0.2-in-1,000 absolute risk increase means you’d need to treat 5,000 patients to see one additional case. This figure is the Number Needed to Harm (NNH), calculated as 1 divided by the absolute risk increase, and puts relative changes into context.

The next time you’re handed a paper, don’t look for the formulas. Follow this biostatistics cheat sheet to know how to read clinical trial results:

• Study design — What is the primary endpoint? Is the control group a fair comparison?
• p-value — Look for statistical significance (p < 0.05).
• Confidence interval — How wide is it? Does the interval cross the line of no effect (0 or 1.0)?
• Relative risk — What is the absolute risk change behind that relative figure, and how many patients actually got better?

Helping tools for statistics:

• p-value calculator;
• Confidence interval calculator;
• 95% confidence interval calculator; and
• Relative risk calculator.

Articles on basic statistics:

• A p-value Less Than 0.05 — What Does it Mean?;
• p-value for the Null Hypothesis: When to Reject the Null Hypothesis;
• How to Interpret Confidence Intervals: A Complete Guide.

Articles related to your field:

• Decision Making in Health and Medicine​: Understanding Risk and Uncertainty in Health Studies; and
• Statistics for Experiments.

---

This article was written by Julia Kopczyńska and reviewed by Steven Wooding.