Venous thromboembolism (VTE), including deep vein thrombosis (DVT) and pulmonary embolism (PE), is a common complication in hospitalized and critically ill patients. Patients in intensive care units (ICUs) are at increased risk of VTE due to predisposing comorbidities, invasive procedures, immobilization, and use of sedative and paralytic medications.1 The incidence of DVT during critical illness varies based on study design with estimates of 5.1% to 5.8% shown in a large international randomized controlled trial (RCT) that utilized biweekly surveillance ultrasound to detect DVTs in medical-surgical ICUs.2 Except in COVID-19 patients, DVT is responsible for most cases of PE—a potentially life-threatening event that has an approximate 30% mortality rate in the ICU.3,4 For acutely ill hospitalized medical patients, the 9th American College of Chest Physicians (ACCP) Antithrombotic Therapy and Prevention of Thrombosis Guideline recommends thromboprophylaxis with either low-molecular-weight heparin (LMWH), low-dose unfractionated heparin (UFH) administered two or three times daily, or fondaparinux (Grade 1B).5,6 For critically ill patients, the ACCP guideline suggests LMWH or low-dose UFH thromboprophylaxis (Grade 2C), except for patients who are bleeding or are at high risk for major bleeding, in which case mechanical thromboprophylaxis with graduated compression stockings and/or intermittent pneumatic compression devices is recommended until the risk of bleeding is attenuated (Grade 2C).5 Meta-analyses conducted in hospitalized patients7 and critically ill patients8 have found that the risk of DVT is similar for patients treated with UFH and LMWH.

Although sex (biologic attributes) and gender (socially constructed roles) influence health and wellbeing in several ways,9 few studies have examined how sex and gender impact the treatment effect of various interventions on outcomes in the ICU.10 Despite well-documented gaps in the representation of women participants in clinical research and females in basic science research,9,11,12 most studies have failed to address whether inherent differences may exist in biology, dosing, metabolic pathways, and responses to treatments between male and female research participants. Sex- and gender-based analyses recognize that a standardized approach to treatment of males and females (alternatively, men and women) as research participants may have unintended consequences. For example, the pharmacokinetics and pharmacodynamics of medications may differ between sexes, resulting in differential treatment effects and adverse event profiles.13 Remarkably, eight of the ten prescription pharmaceuticals that were withdrawn from the USA market between 1997 and 2001 caused harm to women disproportionately compared with men.14

The 2016 Sex and Gender Equity in Research (SAGER) guidelines suggested a comprehensive procedure for reporting of sex and gender information in trials.15 Although most trials report the proportion of females and males enrolled, few trials report treatment effect by sex or gender.15 Not considering sex and gender in research assumes similar treatment effects for all participants and may underestimate the frequency and magnitude of beneficial effects or adverse events. In the Prophylaxis for Thromboembolism in Critical Care Trial (PROTECT) (ClinicalTrials.gov number, NCT00182143), we found that dalteparin was not superior to UFH in decreasing incident proximal DVT or any DVT. There was no significant difference in any VTE or VTE or death, but a significant difference in PE favouring dalteparin compared with UFH (24/1,873, 1.3% vs 43/1,873, 2.3%; HR, 0.51; 95% confidence interval [CI], 0.30 to 0.88; P = 0.01).2 We assessed for an effect modification of thromboprophylactic dose anticoagulation (thromboprophylactic dose dalteparin or UFHs) by sex on thromboembolic events (DVT, PE, VTE) and mortality in the ICU and hospital in a secondary analysis of the PROTECT trial.2

Methods

Ethics review was not required for this secondary analysis of data from the PROTECT trial.

This trial evaluated the superiority of a LMWH, dalteparin, over UFH2 among patients recruited in 67 ICUs in Canada, Australia, Brazil, Saudi Arabia, the UK, and the USA. Patients enrolled in PROTECT were at least 18 yr of age, weighed at least 45 kg, and were expected to remain in the ICU for at least three days. Exclusion criteria in PROTECT included major trauma, neurosurgery, orthopedic surgery, need for therapeutic anticoagulation, heparin administration in the ICU for at least three days, contraindication to heparin or blood products, pregnancy, life-support limitation, or enrollment in a related trial. Investigators randomized 3,764 patients to receive either subcutaneous dalteparin 5,000 IU once daily plus placebo (administered daily) or UFH (5,000 IU twice daily) for the duration of their ICU stay. Randomization was stratified by centre and admission diagnostic category (medical vs surgical) but not by sex. Research pharmacists at participating sites prepared identical syringes of either dalteparin once daily plus placebo once daily (for parallel group twice-daily injections) or of UFH twice daily for subcutaneous injection. Patients, family members, all clinicians, research personnel, ultrasonographers, and outcome adjudicators were blinded to treatment assignment.

The primary outcome was new proximal leg DVT defined as a lack of compressibility using ultrasound evaluation at 1-cm intervals at any of six sites: common femoral vein, proximal femoral vein, middle femoral vein, distal superficial femoral vein, popliteal veins, or the venous trifurcation. Proximal DVT was considered incident or “new” if detected three or more days after randomization.16 Conversely, DVTs that were diagnosed on the first screening ultrasonography were classified as prevalent DVTs. Patients with prevalent DVTs were included in the main analysis but did not contribute to the primary outcome of incident DVT in PROTECT.16 Surveillance compression ultrasound examinations were performed within two days after ICU admission, twice weekly during the ICU stay, and as clinically indicated. Patients were followed until the time of hospital discharge or death.16 Secondary outcomes included any DVT, PE, VTE (combination of DVT and PE), death, a composite of either VTE or death, major bleeding, and heparin-induced thrombocytopenia. The authors classified PE as definite (characteristic intraluminal filling defect on computed tomography of the chest, a high-probability ventilation–perfusion scan, or autopsy finding), probable (high clinical suspicion and either no test results or nondiagnostic results on noninvasive testing), possible (clinical suspicion and nondiagnostic results on noninvasive testing), or absent (negative or normal test results without reference to pretest probability).16 Outcomes were reviewed independently by two blinded adjudicators, except for PE, which was reviewed by four blinded adjudicators.16

Statistical analysis

We report continuous variables as mean and standard deviation (SD) and binary variables as frequency and percentage. In PROTECT, we collected information regarding participant apparent sex (male or female) as opposed to gender (man, woman, or other). Consequently, we analyzed thromboembolic and mortality outcomes based on sex and treatment (dalteparin, UFH).

To examine the effect of sex on outcome, we performed Cox proportional hazards analysis with sex and treatment as the independent variables and including centre and admission diagnostic category (medical vs surgical) as stratification variables as per PROTECT design. To assess for effect modification of treatment (dalteparin, UFH prophylaxis) by sex on thrombosis and mortality outcomes, we conducted Cox proportional hazards analysis, stratified by centre and admission diagnostic category including sex, treatment, and the interaction between them as independent variables. We report summary estimates of effect using the hazard ratio (HR) with 95% CIs.

Subsequently, we performed adjusted analyses wherein we added independent variables including personal history of VTE, family history of VTE, end-stage renal disease, baseline inotrope/vasopressor use, and APACHE II17 score to the Cox proportional hazards analysis described above. In an additional analysis, we examined the effect of weight as a covariate on thrombotic events. We conducted Cox proportional hazards analysis, stratified by centre and admission diagnostic category and included sex, treatment, weight (< 90 kg vs ≥ 90 kg) and the interaction between treatment assignment and sex as independent variables. All analyses were performed using SAS® 9.4 (SAS Institute Inc., Cary, NC, USA). We used the Instrument for Assessing the Credibility of Effect Modification Analyses (ICEMAN) criteria for RCTs (five core questions) to assess the credibility of our sex subgroup analyses.18 With this instrument, each core question contains four response options (definitely reduced credibility, probably reduced credibility, probably increased credibility, and definitely increased credibility) and “probably no” or “unclear” if insufficient information exists to accurately appraise and an optional question to assess additional considerations (e.g., sensitivity analysis, or dose response) that can reduce or increase credibility. The instrument provides an overall credibility assessment on a visual analogue scale (very low, low, moderate, high) credibility that roughly correspond to probabilities of < 25%, 25–50%, 51–75%, and > 75%, respectively, that an effect modification truly exists.18

Results

Baseline characteristics

Of the 3,746 patients included in the PROTECT analysis, we included 3,727 (1,614 females and 2,113 males) with a documented sex. Sex was not collected for 19 trial patients who did not fully meet eligibility criteria but were included in the intention-to-treat analysis. For this analysis, we excluded these patients.

Aligned with the SAGER recommendations, we present the demographic characteristics of participants by sex and treatment arm in Table 1. Female (vs male) trial participants had a slightly higher mean (SD) body mass index (BMI; 28.8 [8.6] kg·m–2 vs 27.9 [6.9] kg·m–2; P < 0.001). Females were more likely to be admitted with a medical diagnosis, have a personal or familial history of DVT, and have sepsis. Compared with females, male participants were more likely to have a history of malignancy and a cardiovascular or gastrointestinal admitting diagnosis. Although similar in frequency overall, on study day one, females were more likely to receive vasopressors and males were more likely to have end-stage renal disease and receive dialysis.

Table 1 Baseline characteristics overall and by participant sex
Full size table

Overall thrombosis and mortality outcomes by sex

Disregarding treatment assignment, female and male PROTECT participants experienced similar rates of any DVT (129/1,614, 8.0% vs 170/2,113, 8.0%; P = 0.38) and proximal DVT (96/1,614, 5.9% vs 109/2,113, 5.2%; P = 0.91). Additionally, females had similar rates of PE (possible, probable, or definite) (22/1,614, 1.4% vs 45/2,113, 2.1%; P = 0.07) or definite/probable PE (21/1,614, 1.3% vs 41/2,113, 1.9%; P = 0.15) compared with males. Females also had similar VTE (143/1,614, 8.9% vs 197/2,113, 9.3%; P = 0.16) and ICU mortality (269/1,614, 16.7% vs 318/2,113, 15.0%; P = 0.80) and hospital mortality rates (374/1,614, 23.2% vs 497/2,113, 23.5%; P = 0.24) compared with males (Table 2). Disregarding treatment assignment, adjusted analyses supported similar thromboembolic and mortality outcome event rates between female and male participants.

Table 2 Thrombotic and mortality outcomes by sex without considering treatment assignment
Full size table

Analyses assessing for treatment by sex interaction

Unadjusted analyses did not support an interaction of treatment by sex for proximal leg DVT, any DVT, and any PE with dalteparin compared with UFH. An interaction between treatment effect and sex was statistically significant for any VTE (males: HR, 0.71; 95% CI, 0.52 to 0.96 vs females: HR, 1.16; 95% CI, 0.81 to 1.68; P = 0.04). (Table 3, Figure) We did not identify an effect modification of treatment by sex on either ICU or hospital mortality.

Table 3 Unadjusted subgroup analyses assessing for treatment effect by sex
Full size table
Figure
figure 1

Forest plot of hazard ratios for thromboembolic and mortality outcomes in PROTECT.

DVT = deep venous thrombosis; ICU = intensive care unit; PE = pulmonary embolism; UFH = unfractionated heparin; VTE = venous thromboembolism. The hazard ratios depicted are from unadjusted analyses

Full size image

Similarly, adjusted analyses did not support an interaction of treatment by sex for proximal DVT (males: HR, 0.74; 95% CI, 0.49 to 1.11 vs females: HR, 1.19; 95% CI, 0.76 to 1.85; P = 0.12 for the interaction), any DVT (males: HR, 0.79; 95% CI, 0.57 to 1.10 vs females: HR, 1.13; 95% CI, 0.77 to 1.66; P = 0.17 for the interaction) or any PE (males: HR, 0.39; 95% CI, 0.20 to 0.77 vs females: HR, 0.87; 95% CI, 0.33 to 2.30; P = 0.195 for the interaction). Nevertheless, adjusted analyses supported a statistically significant interaction of treatment by sex, with a significantly lower HR for any VTE in male compared with female trial participants (males: HR, 0.70; 95% CI, 0.52 to 0.96 vs females: HR, 1.17; 95% CI, 0.81 to 1.68; P = 0.04).

The statistically significant interaction of treatment by sex, with a lower HR for any VTE in male vs female PROTECT participants remained robust in an additional analysis that also included weight (HR, 0.70; 95% CI, 0.52 to 0.96 vs HR, 1.20; 95% CI, 0.83 to 1.73; P = 0.03).

Credibility of effect modification analyses

We evaluated our analyses to have moderate credibility (likely effect modification) using ICEMAN criteria.18 Consequently, we summarize our findings using separate effects for subgroups and note remaining uncertainty.

Discussion

In this secondary analysis of a multinational thromboprophylaxis trial,2 we found that, although critically ill female and male trial participants experienced similar rates of DVT, proximal DVT, PE, VTE, and ICU and hospital mortality, there were no significant differences in treatment effect by sex on outcomes (proximal leg DVT, any DVT, and PE). We found an interaction between treatment effect and sex that was statistically significant for any VTE. As such, we did not identify a significant interaction of sex by treatment for the different subtypes of VTE, only for the composite classification of any VTE as a whole. The significantly lower HR for any VTE favouring male participants treated with dalteparin was supported by adjusted and additional analyses. We did not identify an effect modification of treatment by sex on either ICU or hospital mortality. These findings are hypothesis-generating and highlight the need for sex-based analyses, and gender-based analyses where data permit, in research.

The impact of risk factors, including sex, on incident VTE is not fully understood in the outpatient setting. The COMMAND VTE Registry enrolled 3,027 patients (29 centres) with symptomatic VTE and compared clinical characteristics and outcomes between men (1,169/3,207; 39%) and women (1,858/3,207; 61%). Compared with women, men were significantly younger, more often had had prior VTE, and less often had transient VTE risk factors but had similar three-year incidence risks for recurrent VTE and major bleeding.19 In an analysis of three large studies, Scheres et al. found sex-specific differences across databases from three large studies in the distribution of the presenting VTE location with PE being more often the location of presentation for women compared with men.20 The relationship between sex and incident VTE has not been clearly characterized and may vary based on whether sex-specific risk factors are considered during analysis.21,22,23,24 Male sex is associated with an increased risk of recurrent VTE.25 A meta-analysis of seven outpatient studies (n = 20,534) found that men were more likely to develop proximal DVT (popliteal, femoral, or iliac veins) and women were more likely to develop distal DVT (calf veins).26 The latter finding may relate to the higher prevalence of lower leg varicose veins in women compared with men. The differences in the risk of recurrent thrombosis in males compared with females has led to the development and validation of a widely used scoring system that using sex-specific criteria for assigning risk of recurrent thrombosis, and, subsequently, for determining the duration of anticoagulation therapy.27,28

In a national sample of hospitalized patients (n = 107,896) receiving treatment for VTE, Marshall and coworkers identified a significantly higher incidence of proximal DVT in men but no sex-based differences in PE, hospital mortality, or length of stay.29 Nevertheless, after controlling for age, race, insurance status, hospital division, and medical comorbidities, these findings were no longer statistically significant, and the authors concluded that VTE-related outcomes did not differ significantly based on sex. The risk of recurrent VTE has been reported to be two to threefold higher in men compared with women in observational studies,30,31 but in RCTs the magnitude of risk is less pronounced with relative risk estimates of approximately 1.5.32,33,34,35,36,37 Given the small magnitude of increased recurrence risk in men in RCTs, and the fact that many confidence intervals crossed unity,25 it is not surprising that VTE-related outcomes did not differ significantly based on sex. Although male sex has been widely associated with an increased risk of recurrent VTE, the risk for provoked VTE after ICU admission may depend upon the population, study design, and analytic approach and the risk of recurrent VTE is lower.

Our findings contrast with sex- and gender-based analyses of cardiology treatment trials. Although no thromboprophylaxis trials have examined modification of treatment effects by sex, several cardiology trials have noted greater benefit of therapeutic LMWH vs UFH in women. A meta-analysis of two large trials that compared enoxaparin vs UFH (Efficacy and Safety of Subcutaneous Enoxaparin in Non–Q-Wave Coronary Events [ESSENCE] and Thrombolysis in Myocardial Infarction [TIMI] II-B)38,39 for unstable angina/non–Q-wave myocardial infarction (MI) found that enoxaparin was more effective than iv dose-adjusted UFH in reducing the risk of death, MI, or revascularization, and that the benefit on the composite outcome was greater in women.40 Similarly, in the Fragmin During Instability in Coronary Artery Disease (FRISC) trial, dalteparin (Fragmin) was shown to reduce the risk of death and MI compared with placebo in patients with acute coronary syndrome (ACS) with women experiencing both a larger absolute and relative reduction of the composite primary endpoint, death and nonfatal MI, compared with men.41 Lastly, in the Enoxaparin and Thrombolysis Reperfusion for Acute Myocardial Infarction Treatment-TIMI 25 (ExTRACT-TIMI 25) trial, which randomized patients with ST-segment elevation MI and planned fibrinolysis to either enoxaparin or UFH, women had a similar relative benefit and greater absolute benefit when treated with enoxaparin vs UFH compared with men, despite women presenting with higher baseline risk and increased short-term mortality.42,43 We postulate that the sex differences noted in our study are not likely solely a medication effect. A post hoc analysis of the TIMI II-A trial44 showed that the anti-Xa pharmacokinetic and pharmacodynamic profiles of enoxaparin-treated men and women who presented with non-ST segment elevation ACS were not affected by age, sex, or BMI; however, both profiles were affected by renal insufficiency, especially with creatinine clearance ≤ 40 mL·min–1.44,45 In our study, slightly more males than females had a renal condition at baseline (2.0% vs 1.4%); in these patients, dalteparin clearance could have been lower in male participants resulting in higher plasma concentrations and fewer VTEs. It is also conceivable that a drug-host interaction exists and that dalteparin may be a more effective thromboprophylaxis agent in critically ill males or alternatively UFH may be less effective in males. Aligned with this, women, who typically have higher average body fat and lower weight, plasma volume, and blood flow than men,46 treated with therapeutic dose intravenous UFH typically achieve a higher maximum concentration (Cmax), are predisposed to a higher risk of bleeding, and have lower average heparin requirements to reach target therapeutic activated partial thromboplastin clotting times.47,48 Similarly, women have higher anti-Xa activity during the acute phase of treatment and experience more minor bleeding events with LMWH.43 Little is known about the pharmacokinetics and pharmacodynamics of fixed dose, prophylactic LMWH, and UFH in females. This may be an important consideration with fixed dose prophylaxis as female patients typically have a higher average BMI as was observed in our study. Finally, differences in the patient populations and illness severity may also explain the differences between our findings and those of cardiology trials. Baseline risk factors (e.g., sex, BMI) may be less important determinants of VTE in heterogeneous critically ill patients with concomitant organ failure, infection, or inflammation compared with cardiology outpatients presenting with ACS.

Our study has several strengths. First, to our knowledge, it is the first study of critically ill patients to assess whether a treatment effect by sex in thromboprophylaxis effectiveness exists as opposed to a difference in baseline risk of VTE between sexes. Second, it directly addresses the call by the National Institutes of Health Research, the Canadian Institutes of Health Research, and others to report sex-disaggregated outcomes data.49,50,51,52 Sex- and gender-based analyses enhance scientific rigor, examine reproducibility, and directly evaluate the generalizability of research findings to all study participants that can ultimately inform patient care and health policy.53 Third, we conducted an additional analysis using weight as a covariate to reflect the clinical practice of weight-based dosing. Our study also has weaknesses. We conducted apparent sex-based analyses as data regarding participant gender was not collected. As an analysis of a single large trial, our database may be underpowered to identify differences in thromboembolic outcomes that occur at low frequency (incident DVT, PE, and mortality) and consequently we were only able to discern differences for the outcome of any VTE. We cannot exclude that the significantly lower HR for any VTE favouring dalteparin in males may be spurious or represent a statistically significant but not clinically significant finding. Conversely, our findings may reflect the critically ill population studied. We did not adjust P values for multiple testing. Although, robustly conducted primary subgroup sex-based analysis can provide high quality data to assess for sex differences in outcomes, these findings are hypothesis-generating and require confirmation or refutation in other thromboprophylaxis trials of critically ill patients or a meta-regression. They also highlight the importance of and need for sex- and gender-based analyses in acute care research.