Sex and Gender Subgroup Analyses in Randomized Cardiovascular Trials:

1
Sex and Gender Subgroup Analyses in Randomized
Cardiovascular Trials Are these differences
real? Amandev Aulakh, BHSc and Sonia S Anand,
MD, PhD, FRCPC The CARING Network and McMaster
University, Hamilton, Ontario, Canada
Objectives
Discussion
Methods

• Subgroup analyses are increasingly being
  performed in randomized control trials
• The methodological inadequacy of many sex and
  gender-based subgroup analyses has been
  highlighted by this review
• Because subgroup analyses by sex and gender are
  frequently performed, it is important that
  readers of research exercise caution when
  interpreting results of these analyses
• Previously published guidelines can aid
  researchers in the conduct of subgroup analyses
  by sex and gender and perform research that
  advances womens health

• To assess the frequency of subgroup analyses by
  sex and gender in randomized control trials
  (RCTs) of cardiovascular disease (CVD)
• To determine the comprehensiveness of subgroup
  analyses using specified criteria
• To determine the proportion of subgroup analyses
  considered proper using predefined criteria

• Data extraction was performed by two raters and
  the level of agreement was assessed
• The criteria used to score subgroup analyses for
  comprehensiveness were
• The subgroup analysis was pre-stated or planned
  prior to study commencement
• A hypothesis or rationale regarding the subgroup
  analysis was provided
• A statistical test for interaction was performed
  between the subgroups
• The overall treatment results were emphasized
  more than the subgroup results
• A proper subgroup analysis was defined as one
  that included
• A statistical test for interaction to test
  subgroup differences
• Conclusions emphasizing the overall results of
  the trial and not those of the subgroup analysis
• To be classified as a proper analysis, both
  criteria had to be met

Figure Strategy Search used for Article
Selection
Cardiovascular disease, randomized controlled
trial and gender or sex

• Databases
• Medline
• EMBASE
• Cochrane Central Register of Controlled Trials
  (CCTR)

N 2666
Excluded review articles and systematic reviews
N 1538
Background
Restricted time period to 1990-2006, included
only English language articles and removed
duplicates

• Subgroup analyses in randomized control trials
  examine differences in baseline characteristics,
  exposure-disease associations and treatment
  effects in a subset of study participants
• Subgroup analyses by sex and gender are often
  performed because of growing interest in sex and
  gender differences
• Responses to drug treatment or surgical
  interventions may be heterogeneous between men
  and women
• Since 1994, the National Institutes of Health
  requires that at least half of all clinical trial
  participants enrolled are women
• Although sex and gender-based subgroup analyses
  are more prevalent, their rigour has been
  relatively left unchecked
• The use of improperly conducted subgroup analyses
  can yield misleading or incorrect conclusions,
  which may negatively impact womens health
  improvements
• Examples
• The 1978 Canadian Cooperative Study Group Trial
  The trial found that aspirin reduced the risk of
  stroke in men by 48 but increased the risk in
  women by 42. Women composed less than one third
  of the study population.
• The 1995 Asymptomatic Carotid Atherosclerosis
  Study Women experienced less benefit from
  carotid endarterectomy than men in reducing risk
  of transient ischemic attack, stroke and death.
  The trial was not powered to detect sex
  differences.

N 1236
Sorted manually for i) men and
women ii) RCTs
iii) cardiovascular disease
Guidelines
Articles not including a subgroup analysis N 80
N 169

• State the subgroup analysis a priori
• Provide a rationale for performing the subgroup
  analysis
• Offer a hypothesis regarding the outcome of the
  subgroup analysis
• Perform a statistical test for interaction
  between the sex or gender-based subgroups
• Adjust p-values for the number of comparisons
  being made
• Emphasize overall findings instead of subgroup
  analysis findings

Articles including a subgroup analysis N 89
Results

• 89 of the total 169 studies (53) meeting the
  selection criteria performed a subgroup analysis
  by sex or gender
• Cohens kappa coefficient was 0.814 for the level
  of agreement between the comprehensive scores of
  the two data extractors
• Proper Subgroup Analysis
• 35 of the articles (n 31) performed a proper
  subgroup analysis
• Of the articles in which a proper SGA was
  performed, 9.7 (n 3) met all of the criteria,
  and 58 (n 18) met 3 or more of the 4 specified
  criteria

• Comprehensiveness of Subgroup Analyses
• 57 (n 51) of the articles indicated intent for
  the subgroup analysis by gender/sex a priori
• Only 27 (n 24) of the articles offered a
  hypothesis pertaining to the findings of the
  subgroup analysis or a rationale for the subgroup
  analysis
• 62 (n 55) performed a statistical test for
  interaction to determine a difference between
  male and female subgroups
• 73 (n 65) of the studies emphasize the overall
  findings of the study as opposed to the findings
  of the subgroup analysis

Limitations

• Relevant articles may not have been selected by
  the search strategy used if they did not include
  an exhaustive list of MeSH keywords
• Studies publishing the results of a sex or
  gender-based subgroup analysis in a separate
  publication may not have been identified by the
  search strategy used
• Publication bias is a concern because studies
  with a statistically significant subgroup result
  are more likely to report their subgroup findings
  than studies not finding a significant difference

Graph 1. Distribution of sex and gender subgroup
analyses from 1990 - 2005 by year
Graph 2. Frequency count of comprehensive
subgroup analysis scores.
Keywords
Conclusions
Sex a classification of male and female on the
basis of biology Gender a classification of
male or female that is socially or culturally
oriented Subgroup Analysis analysis of the
treatment effect within a specified subgroup that
is formed on the basis of a baseline covariate,
or a covariate that is not changed by the
treatment Statistical Test for Interaction a
comparison that tests for a statistically
significant difference in the treatment effects
between subgroups

• Conducting sex and gender subgroup analyses is
  increasingly prevalent in cardiovascular
  randomized control trials
• The rigour with which subgroup analyses are
  conducted should be improved
• Investigators must be aware of the importance of
  conducting proper subgroup analyses to avoid
  misleading interpretation of their results
• Guidelines can serve as an aid to investigators
  in performing and interpreting subgroup analyses
  by sex and gender