What does primary, secondary and safety outcome of a drug study mean?

Question

I was reading a research article and came across this:

We studied 55 outcomes of interest, including both effectiveness and safety endpoints. We divided effectiveness outcomes into three primary endpoints (ie, acute myocardial infarction, hospitalization for heart failure, and stroke, on the basis of their use in the 2017 AHA/ACC Guidelines systematic review), and six secondary effectiveness outcomes that major hypertension treatment RCTs have considered.3, 9, 10 The 46 safety outcomes were antihypertensive drug side-effects, including angioedema, cough, electrolyte imbalance, gout, diarrhea, and kidney disease.

My questions

1. I understand drug is said to be effective if it meets the purpose for which it was designed (ex: taking the drug cures the disease). What does primary endpoints mean in this context?

2. What does secondary effectiveness outcomes mean?

3. What are safety outcomes? Why are drug side-effects considered as safety outcomes?

Answer 1

The primary endpoint is the thing you design your study around. You choose a sample size based on a statistical power analysis, making sure you have enough participants that you will be able to detect a statistically significant effect if there is a large enough actual effect. Your p-value for your primary end point should be a good p-value if you've pre-planned everything in your statistical approach.

Secondary endpoints are anything else that you plan to measure. There can be lots of them, and often they aren't corrected for multiple comparisons so readers of the study should interpret the results cautiously. There may not be enough subjects to detect a real effect, and significant results may be false-positives due to multiple endpoints.

Safety endpoints can also be secondary or even primary endpoints, but often they are kept separate because no statistical analysis is planned. Serious safety concerns are usually rare, but can still be an issue even if not "statistically significant". Our threshold for determining something is "unsafe" is much lower than for determining something is "effective", for good reason.

Primary endpoints aren't always the true endpoint of most interest, it's just what happens to be measured for that specific study. For example, if you are giving a drug to reduce blood pressure, what you really want to do long term is to make people live longer or live without cardiovascular consequences like heart attacks and strokes. However, these are long term goals, so your primary endpoint for a specific study you hope to complete within a year may be "change in systolic blood pressure" or "fewer patients categorized as having hypertension". For a cancer drug, you might look at "reduced tumor growth rate" rather than "survival at 5 years".