Cumulative survival curves for female and male sex according to length of intensive care unit (ICU) stay (n = 595 patients).
Romo H, Amaral ACK, Vincent J. Effect of Patient Sex on Intensive Care Unit Survival. Arch Intern Med. 2004;164(1):61–65. doi:10.1001/archinte.164.1.61
Human observations have shown different mortality rates between men and women with various pathological conditions, but this issue has not been widely studied in a heterogeneous population of critically ill patients.
Retrospective analysis of all patients admitted to a mixed medical-surgical, 31-bed intensive care unit (ICU) during 2 different years (1983 and 1995) to evaluate possible differences in mortality between male and female patients and between medical and surgical admissions and variations in these differences over time.
From a total of 4420 admissions (1587 women, 2833 men), women showed a higher mortality, with an odds ratio (OR) of 1.18 (95% confidence interval [CI], 1.02-1.38). This pattern was the same for the 2 periods, and all patient data were therefore analyzed together. After age stratification, the differences were significant for female patients older than 50 years (OR, 1.33; 95% CI, 1.12-1.58) but not in the younger age group. The subgroup of medical admissions had a higher mortality (24.4% vs 7.4%, P<.001) and a higher female proportion (37.9% vs 34.2%, P = .01) than surgical admissions. In multivariate analysis, female sex remained an important predictor of mortality (OR, 1.54; 95% CI, 1.25-1.89). Women had a higher mortality than men in the subgroup of cardiovascular diseases. The highest mortality in female patients was present in the first days after admission and decreased over time, showing a covariance of time and sex.
In a mixed medical-surgical ICU, older women have a higher mortality rate than men. This difference is not apparent for patients staying longer in the ICU.
Recently, there has been increased interest in the possible influences of sex on disease development and intensive care unit (ICU) survival. Women have a lower incidence of coronary heart disease (CHD), an effect that is blunted with aging.1 After an acute myocardial infarction (AMI), women have a higher rate of early mortality,2- 5 although long-term mortality after AMI is not different between the sexes.5 Higher mortality rates in men have been observed in some situations, including congestive heart failure6,7 and trauma,8,9 but in other areas, including sepsis,10- 16 mechanical ventilation,17,18 and burns,19- 21 the literature is conflicting.
The reasons behind possible sex differences have been explored in several experimental models, mainly focusing on the possible immunological effects of sex hormones. Numerous studies have shown that the immune response, including that associated with sepsis, may be modulated by sex hormones, with women and men treated with estrogens developing a more marked proinflammatory response22- 24 and having improved cardiac and hepatic function compared with men.25 These changes are influenced by the female reproductive cycle,26 and the cytokine response is reversed when older animals with decreased hormonal levels are compared.27 Indeed, higher male mortality rates have been observed,28,29 and treatment of septic male animals with dehydroepiandrosterone resulted in improved outcome.30 Whether these experimental findings can be translated to the clinical arena, however, remains unclear. The objective of this study was to observe if there were any differences in mortality rates and length of ICU stay between men and women in a mixed medical-surgical ICU population.
Records for all consecutive admissions to the ICU of Erasme University Hospital, Brussels, Belgium, during 2 different years, 1983 and 1995, were retrieved by searching the unit databases. These years were chosen because they were readily available from our databases and allowed a comparison of mortality rates over time. Erasme University Hospital is a 900-bed, academic, adult hospital with a multidisciplinary, mixed medical and surgical, 31-bed intensive care department. Patients with acute, uncomplicated coronary disease are admitted to a separate coronary care unit. Patients admitted several times during these years were included as new admissions. Demographic data, including sex, admission diagnosis, medical or surgical status, age, length of ICU stay, and ICU outcome, were collected. Univariate and multivariate analyses were performed to determine predictors of mortality in this population using statistical software (SPSS version 10.0.5; SPSS Inc, Chicago, Ill). Categorical variables were assessed with the χ2 test. The 2-tailed t test and Mann-Whitney U test were used as appropriate for continuous variables. Stratified analysis was performed using the Mantel-Haenszel test. General linear model and time-dependent Cox regression model were also applied. The actuarial method was used to assess cumulative differences in survival. P<.05 was considered statistically significant unless a Bonferroni correction was applied to evaluate multiple comparisons. Data are presented as mean ± SD, excepted where stated otherwise.
Demographic data for the 2 years are summarized in Table 1. There were 4420 admissions (1587 women, 2833 men), 1950 admitted in 1983 and 2470 in 1995. The proportion of female patients increased throughout the 12-year period. There were no significant differences in the mean ICU length of stay when both years were compared. Since mortality rates were similar in the 2 different periods, data from the 2 years were analyzed collectively.
From the total number of patients, 595 (13.5%) died during their ICU stay. Survivors were younger than nonsurvivors (55 ± 18 years vs 61 ± 18 years, P = .006), and they had a significantly shorter length of ICU stay (3.5 ± 5.7 days vs 5.4 ± 7.6 days, P<.001). Mortality increased with age from 9.9% in patients younger than 45 years, 11.8% in patients aged 45 to 65 years, and 18.1% in patients older than 65 years (χ2 trend = 40.5, P<.001). Women had a higher mortality rate than men (15% vs 13%; odds ratio [OR], 1.18; P<.05). When the sample was stratified according to age, women older than 50 years showed the highest risk of mortality (OR, 1.33; P<.001). The crude OR and the Mantel-Haenszel adjusted OR were not significantly different, showing no evidence of confounding. However, the homogeneity test across strata confirmed the presence of a significant interaction between sex and mortality when age was controlled (P<.001).
Medical patients represented 35.9% of the population; differences between the medical and surgical groups are depicted in Table 2. There were no differences in age between the medical and surgical groups as a whole or when stratified according to diagnosis or sex. There were more female than male medical patients (37.9% vs 34.2%, P = .01). Surgical patients had a shorter mean length of stay than medical patients (3.6 ± 6.1 days vs 4.1 ± 5.9 days, P = .04). The mortality rate was higher in medical vs surgical patients (24.4% vs 7.4%, P<.001). Cardiovascular problems were the most frequent source of admission to the ICU for the medical group, accounting for 29% of admissions. The mortality rate of female patients with cardiovascular disease was higher than that of their male counterparts (23% vs 12%; OR, 2.07; 95% CI, 1.50-2.87; P<.001) (Table 3).
Using the actuarial method, life tables were used to assess the median length of stay in the ICU until death. The median length of stay was 26 days for surgical patients and 14 days for medical patients (P<.001). Differences in mortality were observed from days 3 through 6, favoring men (Table 4), but after this initial period, the mortality rate was lower for women, although not significantly, suggesting that survival for men and women is not constant over time. The plot of survival curves (Figure 1) shows the cumulative proportion of patients staying in the ICU until death. The curves eventually cross each other due to interaction of sex with length of ICU stay; therefore, a time-dependent Cox regression model was used to assess the joint effects of prognostic factors, such as age, sex, and medical-surgical status, on length of stay in the ICU until death. The model showed mortality for women to be associated with an OR of 1.54 (95% CI, 1.25-1.89) overall and an OR of 0.95 (95% CI, 0.92-0.98) for each day of increase in length of stay, suggesting a reduction in the OR for longer stays (Table 5). This confirms the significant interaction of sex and time that was already found with the stratified analysis.
More male patients are admitted to the ICU than female patients, with men accounting for approximately two thirds of admissions.13,31 Our study confirms this proportion, although suggests a slight increase in the proportion of female admissions in more recent years.
In experimental studies,28,29,32 mortality rates have consistently been shown to be higher in male than female animals in various disease models, yet our study showed no significant differences in mortality rates between sexes, except in women older than 50 years, who had a higher mortality rate. So why do our results and those from other clinical studies14,19 differ from what we would expect from these experimental observations? First, experimental data are generally based on young women in their proestrus state. Clinical studies usually involve populations of mixed ages, with some women in the proestrus state and others in the postmenopausal state; thus, the hormonal milieu will be different from the controlled animal experiment.26 The difference in ages may not only be important because of decreased estrogen levels in older women, but, as shown by Kahlke et al27 in mice, the cytokine response is reversed so that older male mice show a better proinflammatory response than female mice. This factor could explain why, in our results as well as others,14 differences were noted only in older patients and why women fared worse. Second, it is possible that a female sex advantage may prevent women from developing a more severe state to the same insult. Wichmann et al13 studied the development of severe sepsis and septic shock in the postoperative state. These authors observed that the mortality from severe sepsis and septic shock was similar for men and women, but more men needed ICU care, and in this subgroup, more men developed severe sepsis and septic shock. Male sex has also been observed as a risk factor for sepsis and multiple organ dysfunction after burns33 and trauma,34 as well as for infections after surgery.35 This fact may limit the detection of any direct sex-related differences in mortality if patients are studied only in the ICU.
The greatest paradox is seen in CHD, because female sex protects against its development but still carries a greater mortality risk for acute events. Several factors can account for this finding. First, the most important difference is observed for the younger groups3; thus, younger women may have a more severe form of CHD, as manifested by a greater tendency to show eroded plaques,36 higher platelet reactivity,37 fewer collateral vessels,38 and a higher incidence of AMI with normal angiogram.39 Second, women have been shown to have more complications, such as arrhythmia, acute pulmonary edema, and cardiogenic shock,3- 5,39 as well as more comorbid disease processes, such as congestive heart failure, arterial hypertension, and type 2 diabetes mellitus (DM). There may be a relationship among sex, CHD, and DM. Diabetic women seem to have a higher mortality from AMI than women without DM, whereas in men such a difference is not noted.40 Most interestingly, the risk of fatal AMI in women without DM may be lower than for men without DM.41 Third, a possible survival advantage of women may allow more of them to reach the hospital alive, giving a biased population of more severely ill female patients. These data are suggested by 2 studies42,43 that confirm a higher AMI prehospital mortality for men, higher hospital mortality for women, and, subsequently, the same overall mortality.
Our study is, to our knowledge, the first report of differences in mortality in a large population derived from a mixed medical-surgical ICU rather than in specific patient groups. We observed a significantly higher mortality in women older than 50 years and in the subgroup admitted for cardiovascular diseases. Although more women were admitted for medical reasons, and this subgroup had a higher mortality, in multivariate analysis female sex remained an important predictor of mortality. Importantly, patients with uncomplicated acute coronary syndromes were not included but have relatively low mortality rates; thus, for differences in mortality to be shown, large patient populations may be required. The ICU mortality rates are more valuable when controlled for severity of illness; unfortunately, one of the limitations of our study was the lack of a severity score, not widely used in 1983, to further explore our data and perhaps provide some explanations for the observed differences in mortality.
Survival analysis showed that women's survival chances improve with longer length of stay, as shown by a decreasing hazard function for female patients, which crossed over the male hazard function. This finding is in parallel with studies that compare mortality in men and women after coronary events, which showed an increase in early mortality in women.2,44 This result could also be due to an enhanced initial proinflammatory reaction in women, as demonstrated in animal experiments,22- 25 followed by a more balanced reaction later, as indicated by a higher interleukin 10 level in women after 10 days.10 Hence, the timing of observations may be important when studying sex-related differences in mortality.
In conclusion, ICU survival depends on the effects of several predictors interacting with each other and with time. Clinical trials should consider that men and women respond differently to noxious stimuli, although the possible female advantages described in animals have not yet been confirmed in humans. Additional prospective studies observing not only sex but also hormonal status (including exogenous hormone use), associated with a better stratification according to age and disease state and severity of illness, are needed to further characterize these differences.
Corresponding author and reprints: Jean-Louis Vincent, MD, PhD, Department of Intensive Care, Erasme Hospital, Route de Lennik, 808, B-1070 Brussels, Belgium (e-mail: email@example.com).
Accepted for publication January 31, 2003.