Introduction

During the past decade, several clinical and epidemiological studies have investigatedthe impact of gender on outcome in various clinical settings, yielding conflictingresults [1–10]. Sexual dimorphism in the immune response to noxious agents has beencorrelated to differences in sex steroid hormone concentrations that ultimatelydetermine the effect of gender on outcome [11–13]. Females have been observed to have more prominent hormonal and cell-mediatedimmune responses compared with males. Schröder and colleagues demonstrated thatmale patients with sepsis had testosterone levels that were consistently lower than thenormal range and that postmenopausal female patients had higher estradiol levels thanexpected [14]. These differences in hormonal secretion may play a key role in the improvedsurvival of critically ill women. Moreover, dysregulated proinflammatory andanti-inflammatory responses related to sexual immunomodulation of the cytokine networkare thought to be responsible for differences in susceptibility to sepsis and subsequentmultiorgan failure, which correlate with sex-based mortality rates [12, 15]. A recent French study, however, found that mortality was higher among femaleICU patients developing nosocomial infections than among male patients [10]. A higher risk of in-hospital death was also found for younger womenundergoing coronary artery bypass surgery [4] and for female trauma patients who acquired pneumonia during the ICU stay [3].

We conducted this post-hoc analysis to investigate the influence of gender onthe epidemiology of severe sepsis in a large cohort of ICU patients in the region ofPiedmont in Italy and its possible impact on morbidity and mortality in thesepatients.

Materials and methods

All adult patients (> 18 years old) admitted to the 24 Italian ICUs participating in thePiedmont Intensive Care Unit Network were included in this prospective multicenterobservational study conducted between 3 April 2006 and 29 September 2006 [16, 17]. These ICUs represent 75% of the ICUs in the region of Piedmont; inparticular, peripheral and central hospitals of the provinces of Torino, Cuneo, Asti andAlessandria. Recruitment for participation was by open invitation and was voluntary,with no financial incentive. The study was approved by the institutional review board ofthe coordinating center (San Giovanni Battista-Molinette Hospital, University of Turin,Italy) and adopted by the participating centers (Additional file 1). Informed consent was not required because of the observational natureof the study.

Data collection was performed using database-oriented software. For all variablescollected, precise definitions were provided in the relevant part of the software. Ineach ICU, a trained physician was responsible for data collection and entry. Centralsupport was provided by the department of anesthesiology and intensive care at theUniversity of Turin (coordinating center). Validity checks were made concurrent withdata entry in the electronic case record form, including plausibility checks for eachvariable and between variables. Data were further reviewed by the coordinating center,and any doubts clarified with the corresponding ICU.

For all patients, the following data were recorded on admission to the ICU: demographics(age, sex), admission diagnoses, admission category (medical, scheduled surgery,emergency surgery, or trauma) and origin (emergency, surgical or medical ward or anotherICU from the same hospital, or transfer from another hospital), comorbidities, surgicalstatus, reason for admission, and the components of the Simplified Acute PhysiologyScore (SAPS) II [18]. Daily data collection included the presence of systemic inflammatoryresponse syndrome. Patients with > 2 systemic inflammatory response syndrome criteriawere additionally screened for the presence of infection, and the parameters of organdysfunction/failure as assessed by the Sequential Organ Failure Assessment (SOFA) score [19] were recorded daily thereafter by the attending physician. Patients werefollowed up from the first day of admission until death or ICU discharge. Only the firstadmission to the ICU was considered.

Definitions

Sepsis syndromes were diagnosed according to the criteria proposed by the AmericanCollege of Chest Physicians/Society of Critical Care Medicine Consensus Conference [20]. ICU-acquired sepsis was defined as sepsis identified > 48 hours after ICUadmission and non-ICU-acquired sepsis as sepsis occurring within 48 hours of ICUadmission. Surgical admissions were defined as patients who had undergone surgerywithin 2 weeks preceding admission. Emergency surgery was defined as a nonscheduledoperation within 24 hours of the onset of symptoms or injury. Comorbidities includedthe presence of insulin-dependent diabetes mellitus (need for daily injection ofinsulin prior to ICU admission), chronic obstructive pulmonary disease, heart failureclass III or IV according to the New York Heart Association definitions, and chronicrenal failure (need for chronic renal support or history of chronic renalinsufficiency). Patients were also classified by the admitting physician according towhether they were admitted to the ICU for only short-term monitoring, had an expectedICU length of stay < 24 hours, or were admitted for intensive care treatment withan expected ICU length of stay > 24 hours.

We also examined the epidemiology of severe septic syndromes in two a priori defined age subgroups of patients (≤ 50 years and > 50 years), assumingthat a 50-year cutoff value represented a reasonable physiological limit betweenpremenopausal and postmenopausal periods for women.

Outcome parameters

The primary outcome parameter was death in the ICU. Secondary outcome parameters werethe development of sepsis syndromes in the ICU and the ICU length of stay.

Statistical analysis

Data were analyzed using SPSS 17.0 for Windows (SPSS Inc., Chicago, IL, USA).Discrete variables are expressed as counts (percentage) and continuous variables asmeans ± standard deviation or median and interquartile range unless statedotherwise. Categorical data were compared using the chi-square test with Yates'correction, by Fisher's exact test or by the Cochran-Armitage trend test, asappropriate. A Kolmogorov-Smirnov test was used to verify the normality ofdistribution of continuous variables. Continuous variables conforming to a normaldistribution were compared using analysis of variance and Student's t test;otherwise the Kruskal-Wallis and Mann-Whitney U tests were applied. A Bonferronicorrection was done for multiple comparisons. Kaplan-Meier survival curves stratifiedaccording to gender were plotted over the 28 days following admission to the ICU andwere compared using a log-rank test.

To investigate the impact of gender on ICU mortality adjusting for differences inbaseline characteristics and severity of illness, we performed a logistic regressionanalysis with ICU mortality as the dependent factor in the overall population.Variables included in this analysis were age, comorbid diseases, SAPS II onadmission, the referring facility, the type of admission to the ICU, the presence ofsepsis syndromes and the time of acquisition of sepsis. Collinearity between thevariables was excluded prior to modeling. A Hosmer and Lemeshow goodness-of-fit testwas performed, and odds ratios (ORs) with 95% confidence intervals (CIs) werecomputed. We adjusted for the center effect in the final model by introducing this asa covariate, with the center that included the largest number of patients as thereference category. A similar model was constructed in patients with severe sepsis.The source of infection, SOFA scores, and the time of acquisition of sepsis were alsoconsidered in this model. In the whole cohort, the multivariate analysis included allcovariates. In patients with severe sepsis, however, covariates were included ifP < 0.2 in a univariate logistic regression analysis in order toreduce the number of covariates in the model because of the relatively small numberof patients in this subgroup.

All statistics were two-tailed and P < 0.05 was considered statisticallysignificant.

Results

Characteristics of the study cohort

During the study period 3,902 patients were admitted to the participating centers, ofwhom 2,479 (63.5%) were male. The characteristics of the study group are shown inTable 1[16, 17]. Female patients were significantly older than male patients (66 ± 16years vs. 63.4 ± 15.6 years, P < 0.001). There were no significantdifferences between men and women in comorbidities, SAPS II, type and reason ofadmission, or referring facility. More of the patients admitted with trauma were male(15.8% vs. 8.6%, P < 0.001).

Table 1 Characteristics of the study cohort on admission to the ICU stratifiedaccording to gender
Full size table

Impact of gender on the epidemiology of severe sepsis

The frequency of severe sepsis, including septic shock, during the ICU stay was lowerin women than in men (6.0% vs. 8.9%, P = 0.001) irrespective of the agegroup, mainly because of the lower occurrence of these syndromes in women within 48hours of admission to the ICU (2.3% vs. 4%, P = 0.005) (Table 2). The prevalence of septic shock was extremely low in female patientsaged ≤ 50 years (0.9%). Among 305 patients who had severe sepsis or septicshock during the ICU stay, 220 (72.1%) were male and only 85 (27.9%) were female.Female patients with severe sepsis were older than male patients (67.7 ± 14.3years vs. 63.1 ± 15 years, P = 0.004), but other baselinecharacteristics, the source of infection, and SOFA scores were similar irrespectiveof gender (Tables 3 and 4).

Table 2 Frequency of severe sepsis according to gender in the whole population andstratified by age
Full size table
Table 3 Characteristics of patients with severe sepsis stratified according togender
Full size table
Table 4 Characteristics of infections in patients with severe sepsis stratifiedaccording to gender
Full size table

Impact of gender on outcome

The overall ICU mortality rate was 20% and did not differ significantly between maleand female patients (20.1% vs. 19.8%, P = 0.838). Twenty-eight-day survivalrates, censored at ICU discharge, were similar in male and female patients (log-rankP = 0.148, Figure 1). The median ICU length of stayin the whole cohort was 3 (1 to 9) days (survivors vs. nonsurvivors, 18 (9.5 to 30)vs. 9 (4 to 21), P < 0.001) and was similar in men and women (Table 1). In a multivariate logistic regression analysis with ICUmortality as the dependent variable, female gender was not independently associatedwith an increased risk of death in the ICU (OR = 1.07, 95% CI = 0.87 to 1.34) (fullresults of the regression analysis are shown in Additional file 2).

Figure 1
figure 1

Kaplan-Meier survival curves representing 28-day survival according togender in the whole cohort.

Full size image

In patients with severe sepsis, ICU mortality was 51.1% and was higher in women thanin men (63.5% vs. 46.4%, P = 0.007). A Kaplan-Meier analysis showed reduced28-day survival in female compared with male patients with severe sepsis (log-rankP = 0.004, Figure 2). However, the ICU length ofstay was similar in men and women (Table 3). In a multivariatelogistic regression analysis in patients with severe sepsis with ICU outcome as thedependent variable, age (OR = 1.03, 95% CI = 1.01 to 1.05, P = 0.02), femalegender (OR = 2.33, 95% CI = 1.23 to 4.39, P = 0.009) and SAPS II (OR = 1.03,95% CI = 1.01 to 1.05, P = 0.001) were independently associated with ahigher risk of ICU death. Other factors associated with a higher risk of death inthis population were SOFA respiratory and SOFA cardiovascular subscores, referralfrom a surgical ward, an emergency department or another ICU, and an abdominal siteof infection (Table 5).

Figure 2
figure 2

Kaplan-Meier survival curves representing 28-day survival according togender in patients with severe sepsis.

Full size image
Table 5 Logistic regression analysis with ICU mortality as the dependent variable inpatients with severe sepsis
Full size table

Discussion

The main finding of our study was that, although the overall prevalence of severe sepsiswas lower in female patients admitted to the ICU than in male patients, female genderwas independently associated with a higher risk of death in the ICU in patients withsevere sepsis.

In our study, there were more male than female ICU admissions. This finding has beenconsistently reported in all the large epidemiologic studies in ICU patients [21–23]. The reason for this finding is unclear. One proposal has been that there maybe a gender-related bias in the provision of care [5, 8, 24]. In a large multicenter Austrian cohort including 25,998 adult ICU patients,despite a higher severity of illness in women, men received an increased level of careand underwent more invasive procedures [5]. Another large single-center retrospective study, including 24,778 criticallyill patients, reported that among patients 50 years or older, women were less likelythan men to receive life-supporting treatments [8]. The differences in provision of care are probably not responsible alone forthe higher ICU admission rates in men than women. Gender-related differences in immuneresponse and in the presentation of critical illness cannot be excluded as anexplanation for this finding [14, 25].

The possible impact of gender on outcome from critical illness has been reportedpreviously but with conflicting results [6, 8, 10]. In our study, gender had no impact on the ICU mortality rate in a largecohort of critically ill patients admitted to the region of Piedmont in Italy. To thebest of our knowledge, our study is the first study to investigate the impact of genderin a representative sample of ICU patients admitted to a specific region. The absence ofgender-related differences in outcome in our study does not preclude possibledifferences in outcome in specific subgroups of ICU patients. In agreement with ourresults, Valentin and colleagues found that outcome was similar in critically ill menand women admitted to 31 Austrian ICUs despite differences in the therapeutic approachaccording to gender [5]. However, Fowler and colleagues reported that female patients were morelikely to die after critical illness [8] - although their study was limited by its single-center nature andretrospective design. Discrepancies between the results of these studies can also beexplained, at least in part, by the differences in case mix.

In our study, the prevalence of severe sepsis was lower in women than in men. The designof our study does not enable us to elaborate on the possible pathophysiologic reasonsfor this finding. It has been reported that an increased estradiol level may enhanceimmune function in females [14, 26, 27]. Moreover, a predominance of anti-inflammatory mediators in women [14] may be responsible for a protective effect in female critically ill patientsin terms of development of severe sepsis. Our results confirm those reported by Adrieand colleagues, in which women had a lower prevalence of severe sepsis [28]. Other results from Wichmann and colleagues showed a significantly lowerincidence of severe sepsis/septic shock in female ICU patients between 60 and 79 yearsold compared with male patients [1]. In contrast, in a recent prospective multicenter cohort study of adulttrauma patients with hemorrhagic shock, Sperry and colleagues noted that female genderwas associated with a significant reduction in rates of multiple organ failure andnosocomial infection [12].

Differences in gender-related outcome may also exist in patients with severe sepsis.Schröder and colleagues reported higher survival rates in women with surgicalsepsis than in men [14, 25]. These authors also analyzed sex-related hormonal secretion and differentpatterns of proinflammatory and anti-inflammatory mediators in response to severe sepsisand found a more favorable hormonal and immunologic profile in women than in men. Thisstudy was limited, however, by the small number of patients and the inclusion of onlysurgical sepsis. In a large case-control study including 1,692 patients with severesepsis, Adrie and colleagues found that mortality was higher in men than in women,especially in the subgroup of patients > 50 years old [7]. Several experimental studies have also reported a survival benefit in femaleseptic animals compared with males [13, 29]. Sexually dimorphic cytokine profiles, such as increased levels ofproinflammatory cytokines, have been suggested to be responsible for this phenomenon [13, 29]. Sex steroids can also modulate the inflammatory response and maysubsequently influence outcome after septic challenge [30].

The results of our study are in contrast to some previous findings [7, 14, 25], with female gender being independently associated with a higher risk ofdeath in patients with severe sepsis. Indeed, our data suggest a protective effect offemale gender in terms of developing sepsis, with a lower prevalence of severe sepsisand septic shock. We suggest that females with an unfavorable immunologic profile arethose who are more liable to develop severe sepsis and subsequently have a worseprognosis, but this hypothesis needs to be confirmed in large, prospective studies. Ourresults agree with those of Eachempati and colleagues, who demonstrated that femalegender was an independent predictor of increased mortality in critically ill patientswith documented infection [31]. More recently, Combes and colleagues analyzed the gender-related outcome ofa mixed population of patients who developed nosocomial infections in the ICU and alsoreported that female gender was associated with an increased risk of ICU mortality [10].

Although this cohort of ICU patients is a representative sample from a specific region,our study has some limitations. First, the epidemiology of sepsis in this region may notbe extrapolated to all ICUs in Italy. These results can also not be extrapolated toother parts of the world because genetic polymorphisms that could have an impact onmortality are not taken into account. Second, we considered only short-term outcome interms of ICU mortality, and cannot therefore comment on the potential impact of genderon in-hospital mortality or longer term outcomes. Third, the relatively small samplesize in the subgroup of patients with severe sepsis may be another limitation of ourstudy. Finally, the multivariate analysis is limited to the variables considered forthis analysis; however, we included a large number of variables relevant to outcome inthis population, and adjusted for the center effect.

Conclusions

In this large, regional Italian cohort of ICU patients, there were more male than femaleICU admissions. The prevalence of severe sepsis was lower in women than in men, butfemale gender was independently associated with a higher risk of death in the ICU inpatients with severe sepsis.

Key messages

  • In this cohort, the overall prevalence of severe sepsis was lower infemale patients admitted to the ICU than in male patients.

  • Female gender was independently associated with a higher risk ofdeath in the ICU in patients with severe sepsis.