decorative image
Official websites use .gov
A .gov website belongs to an official government organization in the United States.
decorative image
Secure .gov websites use HTTPS
A lock ( ) or https:// means you've safely connected to the .gov website. Share sensitive information only on official, secure websites.

n/a Warning! This browser is not supported - Some features might not work. Try using a different browser such as Chrome, Edge, Firefox, or Safari.

Analyses by Sex or Gender, Race and Ethnicity for NIH-defined Phase III Clinical Trials (Valid Analysis)


When an NIH-defined Phase III clinical trial is proposed, evidence must be reviewed to show whether or not clinically important sex/gender and race/ethnicity differences in the intervention effect are to be expected. The application or proposal must address plans for the analysis of intervention effect differences on the basis of sex/gender, race, and ethnicity unless there is clear evidence that such differences are unlikely to be seen.

Definition of Valid Analysis

Valid analysis means an unbiased assessment. Such an assessment will, on average, yield the correct estimate of the difference in outcomes between two groups of subjects. Valid analysis can and should be conducted for both small and large studies. A valid analysis does not need to have a high statistical power for detecting a stated effect. 

The principal requirements for ensuring a valid analysis of the question of interest are: 

  • allocation of study participants of both sexes/genders (males and females) and from different racial and/or ethnic groups to the intervention and control groups by an unbiased process such as randomization; 
  • unbiased evaluation of the outcome(s) of study participants; and 
  • use of unbiased statistical analyses and proper methods of inference to estimate and compare the intervention effects by sex/gender, race, and/or ethnicity.

Reducing Bias and Examining Intervention Effects

Bias can be reduced using several methods. For example, bias in the evaluation can be reduced by using objective measures and staff who are blind to treatment assignment. Bias in the statistical analysis can be reduced by adjusting for potential confounders. Comparison of intervention effects can be achieved by reporting intervention effects and their confidence intervals separately for each sex/gender and for each race/ethnicity group. Generally, it is not sufficient to adjust the primary analysis for sex/gender or race/ethnicity.

Applicant/Recipient Instructions and Guidance

Other Resources

Policy Notices