NEJM perspective-what next is the primary endpoint of clinical trials "negative"?

Article source: Med

Author: Ala Lei

Editor's note

There is a common psychology-"reporting the good but not the worry"

"Negative" results are also of great significance, and the interpretation of the results should be based on the overall rather than individual indicators

The credibility of a well-designed clinical trial comes from including pre-set a priori assumptions that help authors avoid potential false positive results when conducting exploratory analysis of data

The P value should be regarded as a continuum.

12 main questions to think about when the primary end point is "negative"

1.

2.

3.

4.

5.

6.

7.

8.

9.

10.

11.

12.

Are there any signs of potential benefits?

When P>0.
05, whether there is a signal of treatment benefit (or: trend) needs to be carefully evaluated
.
For example, in the PERFORM study, the efficacy of terutroban and aspirin in ischemic stroke was compared
.
The results showed that there was no statistical difference between the two groups in the composite primary endpoint (ischemic stroke, myocardial infarction, and other vascular-related deaths); terutroban also showed no advantages in terms of safety
.
Therefore, this experiment is "negative"
.

In contrast, in the TORCH study, comparing salmeterol-fluticasone and placebo in chronic obstructive pulmonary disease (COPD), the P value of the primary endpoint (death from any cause) was 0.
052; however, "at the other endpoints" Show benefits” (e.
g.
COPD worsening, health status)
.
Therefore, in addition to "negative" results, more in-depth evaluation is needed
.

Is the test efficacy insufficient?

If the sample size of the trial is too small, it may increase the risk of type II error (false negative), that is, it does not show a significant therapeutic benefit
.

Is the primary endpoint reasonable?

Using a composite endpoint can increase the number of major events, but it does not necessarily increase statistical power
.
The success of the trial may depend on the definition of the endpoint and the method of evaluation
.

Is the included patient population reasonable?

When a new treatment plan fails, it is necessary to evaluate whether the included patient group is reasonable
.

Is the treatment plan reasonable?

In a key clinical trial of a new drug, the formulation of the drug dose is challenging
.

Are there any shortcomings in the execution of the test?

If the execution of the research plan is poor, the true curative effect may be weakened or even disappear
.

Is it a non-inferiority test?

When a new treatment regimen does not show an advantage compared with the positive control, can it be evaluated from its non-inferiority? If the new treatment plan has other advantages (for example: less damage, fewer adverse reactions), non-inferiority assessment can be considered
.
However, in most cases, this assessment can only be carried out if non-inferiority assumptions have been set in advance
.

Is there a positive result in the subgroup analysis?

It is reasonable to conduct subgroup analysis in any clinical trial
.
However, for trials where the primary endpoint is neutral or negative, subgroup analysis is often misleading
.
The credibility of this qualitative interaction is low (unless supported by a strong mechanism of action), and statistical analysis is usually not corrected for multiple comparisons
.
Even if the statistical result of the interaction is positive, it is only useful for making hypotheses
.
In fact, in a trial with a negative primary endpoint, subgroup analysis showed positive and confirmed in subsequent trials, such cases are very rare
.

Is there a positive result for the secondary study endpoint?

If the primary end point is negative, a positive secondary result can often also help formulate the hypothesis
.
Of course, regulators are unlikely to approve new drugs based on secondary endpoints
.
However, in some cases, the results of secondary endpoints can also influence guidelines and clinical practice
.

Will changing the analysis method help?

1.
Using baseline variables closely related to the primary end point as covariates for correction analysis can improve statistical power
.
2.
The subject's poor compliance or cross-treatment may affect the true treatment effect, resulting in the intent-to-treat (ITT) analysis that cannot find statistical differences between groups
.
Some researchers believe that it is more reasonable to use treatment or PP analysis in this case
.
3.
In the research of chronic diseases, the composite end point analysis usually only focuses on the time when the event occurs for the first time, and ignores any subsequent repetitive events.
This method will lead to the loss of statistical power and the underestimation of the efficacy
.

Is there external evidence of a positive result?

If a clinical trial with sufficient statistical power, but the primary endpoint is negative, the strength and quality of the previous "positive" related studies need to be evaluated
.
If the previous trial is a non-randomized control, or with a surrogate endpoint as the main outcome, the reliability of the results will be reduced
.
Similar clinical trials or meta-analysis in terms of patient characteristics, treatment plan, endpoint setting, etc.
have more reference value
.

Is there a strong biological theory to support this treatment?

Almost all new treatment options in phase III clinical trials have a large number of supporting data from animal studies and early clinical trials
.
Despite this, there have been many large key clinical trials in history that did not show treatment benefit (or show unexpected safety issues)
.

references

N Engl J Med.
2016; 375:861-70.