Sex representation within intensive care trials in Australia and New Zealand

Dear Editor,

Recruiting a representative study population helps to ensure the generalizability of a trial’s findings across demographic groups. It also promotes trust and fair access to trial participation among the target population of affected patients [1]. Nonetheless, women remain under-represented in some fields including oncology, cardiology and infectious diseases trials [2, 3]. Regarding critical care trials, Kristensen and colleagues found that a minority of participants were women, however they did not compare sex balance between study populations and their respective target populations [4].

We examined sex-based representativeness in randomized controlled trials (RCTs) endorsed by the Australia and New Zealand Intensive Care Society (ANZICS) Clinical Trials Group and completed 2014–2023 (supplement Fig. 1). The ANZICS Adult Patient Database (APD) was taken to represent the general population of intensive care unit (ICU) patients, as it captures 90% of ICU admissions in Australia and New Zealand [5]. The database records patient sex—the focus of our study—defined as the biological distinction between female and male [5].

For each trial, the study population was compared with a matched target population identified from the APD using the trial’s own eligibility criteria (supplement Table 1). Quality of matching was assessed using pre-specified criteria; target populations were considered well-matched if matched in at least 4 of 5 domains (supplement Table 2). We calculated the arithmetic difference between the percentage of women in each study population minus the percentage women in each target population, with estimated 95% confidence interval. We used random effects meta-analysis to calculate pooled differences, using STATA 17 BE (Statacorp, Texas USA) for all analyses.

There were 24 eligible trials with 68,968 participants in total (36.3% women). The sex balance ranged from 16.7 to 47.8% women (Fig. 1a). Each trial recorded patient sex but not gender and no trial included a third sex category, e.g., intersex. Therefore, sex and gender-based minorities were not represented.

Fig. 1.

a Percentage of women in study population compared to target populations. Lines connecting triangles and circles represent the arithmetic difference in the percentage of women in the study population and target population. Pink lines indicate more women in study than target population; blue lines indicate less women in study than target population (adapted from Steinberg et al. [2]). References 6–29 are provided in the electronic supplementary materials. b. Forest plot of sex balance in study population compared to target populations

There were 2,017,197 admissions (42.9% women) recorded in the APD during the trials’ cumulative recruitment period October 2008 to November 2021. There were substantially less women in the pooled study population than the general ICU population represented by the APD (arithmetic difference − 6.6%, 95% confidence interval [CI] − 69 to − 6.2%).

When compared to matched target populations, there was no significant disparity in sex balance between the study populations and target populations overall (pooled arithmetic difference − 0.5%, 95% CI − 2.2% to 1.2%; Fig. 1b). This finding was robust to sensitivity analysis including only trials with well-matched target populations (supplement Fig. 2).

On diagnostic subgroup analysis, women were under-represented in cardiac arrest trials [arithmetic difference − 5.5% (95% CI − 8.6 to − 2.4%)] and sepsis trials [arithmetic difference − 2.9% (95% CI − 5.1% to − 0.7%)], compared to the target populations of ICU patients with these diagnoses. There was no sex-based disparity in trauma trials or trials including multiple diagnostic categories (supplement Fig. 3).

We found critical care trials in Australia and New Zealand recruited representative proportions of women and men overall, when compared to target populations defined by their own eligibility criteria, suggesting that screening and enrolment of participants was not affected by patient sex. However, the pooled study population had substantially less women than the general population of ICU patients in Australia and New Zealand. This suggests the trials examined relatively male-dominated groups of ICU patients by design, related to study question, participating sites or eligibility criteria. Many trials examined conditions predominantly affecting men, including trauma or cardiac arrest, rather than female-dominated conditions such as subarachnoid haemorrhage or asthma [5].

Concordant with previous research, women were under-represented in cardiac arrest and sepsis trials [2]. This may relate to the decision-making of surrogate decision-makers or clinicians in high-risk situations. Further research is needed to identify effective strategies for improving sex-based representation in these trials.

Supplementary Information

Below is the link to the electronic supplementary material.

Declarations

Conflicts of interest

KT is supported by a National Health and Medical Research Council of Australia (NHMRC) Investigator Grant (APP1194058). MW is supported by a similar NHMRC grant (APP1174120). LM is a member of the Women in Intensive Care Medicine Network of the Australia and New Zealand Intensive Care Society (WIN-ANZICS).

Ethics statement

Ethics approval was obtained from Alfred Health Human Research Ethics Committee.