Causes of death and associated risk factors among climacteric women from Southern Brazil: a population based-study

Participants were selected from the population-based cohort of the longitudinal menopause study that has been underway in the city of Passo Fundo, South Brazil, since 1995 [20, 21]. As part of this project, an initial cross-sectional study was performed between 1995 and 1997 to investigate ovarian volume in pre- and perimenopausal women. Sampling was carried out in two stages. First, 154 census sections (geographical subdivisions of the city defined by the Brazilian Institute of Geography and Statistics) were randomly selected. One block in each census section was picked by lot; two women were interviewed in each block after the randomization method described previously [20, 21]. A representative sample of 298 women aged 35 to 55 years who had menstruated at least once in the past 12 months was randomly identified.

In the second field visit, conducted between 2001 and 2002, 239 women from the baseline cohort were located and interviewed. In view of potential losses to follow-up and of the increasing city population, 119 additional women aged 35 to 62 years were sampled to guarantee enough statistical power for the analysis. The final sample was 358 women. The samples were selected at random based on the first census sections [22] (Figure 1).

In 2010, a third follow-up was initiated in order to assess cardiovascular risk and mortality rate. All 358 participants or their relatives were reached, and information regarding participant deaths was obtained for the period ending in November 2011. In addition to the interviews, the medical records of city hospitals and the Center for Health Information (NIS/RS-SES) were reviewed. All deaths between 1995 and 2011 were included in this analysis.

Medical records were reviewed to collect information on age at death, date, and cause of death. The causes of death were coded using the International Classification of Diseases, 10th revision [23]. Deaths were analyzed for all-cause and cardiovascular causes (ICD-10: I00-I99), neoplasms (ICD-10: C00-C97), and respiratory causes (ICD-10: J00-J99).

Ethics approval for the study was granted by the Research Ethics Committees at the University of Passo Fundo and the São Vicente de Paulo Hospital. All study participants signed an informed consent.

The participants were interviewed using a pretested standardized questionnaire covering demographic characteristics (age and self-reported race) and education (years of successful formal education, described as years at school). Skin color was classified according to self-report [21, 22]. Women were classified in terms of alcohol consumption as nondrinkers, social drinkers (1 to 15 g alcohol/day), or abusers (at least 15 g alcohol/day) [24]. Smoking status was categorized as current, ex-smoker, or nonsmoker [22]. Physical activity was investigated through a previously tested standardized questionnaire [25]; for each type of physical activity, the metabolic equivalent [26] and overall calorie expenditure were calculated. Women who expended at least 1000 kcal/week (approximately 3.5 hours per week walking, climbing stairs, swimming, playing sports, doing yard work, and so forth) were considered to be physically active, whereas the others were classified as sedentary [22].

Women were classified according to their baseline menopausal status: premenopause was defined as no change in menstrual frequency or flow; perimenopause was defined as changes in menstrual frequency or flow in the 12 months before the study; and postmenopause was defined as 12 months or more of amenorrhea, including as a result of medical interventions such as bilateral oophorectomy [21, 22]. A "hysterectomy" category was created for women who had undergone hysterectomy and whose menopausal status could not be classified [27].

The use of hormone therapy (HT), estrogen, estrogen plus progestogen or tibolone was verified by asking the women to show the medication box or the physician’s prescription [21, 22].

Body weight and height was assessed at the beginning of the study in 1995. Weight (kg) was measured to the nearest 100 g using a Filizola® scale, Model 31 (Ind Filizola-SA, São Paulo, Brazil), and height (cm) was measured to the nearest 0.1 cm with a wall-mounted fixed stadiometer. Special attention was taken to ensure that the participants were positioned with the Frankfort plane [28] horizontal and that they were barefoot. These were used to calculate BMI, dividing weight in kilograms by height squared (m²), and categorized as <25.0, 25.0–29.9, and ≥30.0 kg/m² [29].

Other anthropometric measurements were made in duplicate between 2001 and 2003, including waist circumference (WC) (measured at the midpoint between the lower rib margin and the iliac crest, perpendicularly to the long axis of the body, with the participant standing balanced on both feet, spread approximately 20 cm apart, with arms hanging freely), hip circumference (widest circumference over the buttocks), and waist to hip ratio (WHR) (waist circumference divided by hip circumference) [21, 22, 28]. All procedures followed standardized recommendations [30] and the equipment was periodically calibrated.

Previous diagnosis of DM was verified based on physician report or current use of anti-diabetic medication. Self-report of hypercholesterolemia or use of anti-cholesterol medication was used to define dyslipidemia.

Blood pressure was measured after a 10-minute rest. The same calibrated mercury manometer attached to a 12.5 × 23 cm inflatable cuff was used in all participants, and the fifth Korotkoff sound was adopted to determine diastolic pressure. Hypertension was defined as systolic blood pressure ≥140 and/or diastolic blood pressure ≥90 mmHg or current use of antihypertensive medication [31, 32].

All participants were submitted, in 2003, to blood sampling between 8 and 10 a.m. after an overnight fast of 10 to 12 hours. Total cholesterol, high-density cholesterol (HDL-c), triglycerides, and glucose levels were determined by a colorimetric-enzymatic method (Architect C800, ABBOTT Systems). Low density lipoprotein cholesterol (LDL-c) was determined indirectly using the following formula: LDL-c = total cholesterol - (HDL-c + triglycerides/5) [33, 34].

Continuous variables are reported as means ± SDs. Categorical variables are reported as frequencies (%). Differences in baseline clinical characteristics between groups were analyzed by the Student t test (for continuous variables with normal distribution), Mann-Whitney’s U test (for continuous variables with skewed distribution), or χ² test (for categorical variables). Survival time for each participant was defined as the time between the date of study entry and the occurrence of death. Univariate predictors of mortality during follow-up were analyzed by Cox regression models and calculation of hazard rate ratios with 95% CI. For the estimation of hazard ratios, a Cox regression model was fit including relevant independent variables to estimate the associations between baseline characteristics and mortality. Multivariate analysis included only variables with a p-value of 0.05 or lower on univariate analysis. Adjusted hazard ratios (HR) and 95% confidence intervals (CI95%) for mortality were estimated. We calculated adjusted hazard ratios after multivariate correction for age and smoking. The final model included all variables with P < 0.05 or according to clinical plausibility. Collinearity for the model variables was evaluated using variance inflation factors and tolerances. No collinearity was found. Survival curves were estimated using the Kaplan-Meyer method. Log-rank p-values were calculated to test for significant differences between mortality and DM. All statistics analyses were performed using SPSS 20.0 software. P < 0.05 was considered statistically significant for all analyses.

A total of 358 women were studied during a mean follow-up of 13.4 ± 3.3 years. Table 1 shows the distribution of baseline characteristics in survivors and non-survivors. The mean age of the overall sample was 44.29 ± 6.0 years. Most participants were white (86.3%), with low levels of schooling (8.4 ± 4.7 years). Considering menopause status, 162 (47.1%) were premenopausal, 134 (39%) were perimenopausal, 37 (10.3%) were postmenopausal, and 11 (3.2%) had undergone a hysterectomy. Of the 358 individuals included in the study, 177 (49.4%) reported having hypertension; 56 (15.6%) used HT, and 14 (3.9%) were diabetic. Smoking, a well-known risk factor for CVD, was found in 96 (26.8%) of the overall sample, and in 8 (47.1%) deceased patients. Survivors were significantly younger and almost half were premenopausal. They also had higher education level and fewer cases of previous DM.

Table 2 describes the metabolic profile of participants. In general, they had low level of physical activity during their leisure time and showed baseline overweight and central adiposity. A relatively healthy lipid profile was observed, with normal or borderline values. WHR and HDL-c were significantly different in survivors and non-survivors.

The mean age at death was 57.8 ± 5.5 years. In the 13 years analyzed, 17 (4.7%), deaths from all causes were recorded. Among these, seven (41.2%) deaths were caused by CVD, including four cases of stroke and three myocardial infarctions. There were six (35.3%) deaths due to kidney (n = 1), breast (n = 1), lung (n = 2) and uterine (n = 2) cancer, and four (23.5%) deaths due to other reasons, such as polytrauma, asthma, and diabetes.

The survival for the entire cohort in 6, 12 and 15 years was 98.6%, 96.8% and 94.3%, respectively (Figure 2).

In a univariate analysis (Table 3), older age (HR = 1.141, 95% IC: 1.051-1.239, p = 0.002), menopausal status (postmenopause HR = 16.738, 95% CI: 3.644-16.895, p < 0.001), hysterectomy HR = 11.788, 95% CI: 3.644-121.382, p = 0.038), DM (HR = 10.439, 95% CI: 3.367– 32.367, p < 0.001), and WHR ≥ 0.85 (HR = 3.556, 95% CI: 1.253-10.094, p = 0.017) were associated with a major probability of death. Years at school was statistically meaningful in the univariate model (HR = 0.888, 95% CI: 0,792-0.997, p = 0.044), suggesting a protective association of higher level of education.

Table 4 presents the results of multivariate analysis resulting from the Cox regression model, taking into account the confounding factors (age and smoking). Previous diagnosis of DM, WHR ≥ 0.85, and postmenopause status remained the leading risk factors for all cause mortality, similar to those found in the crude analysis and increasing the risk of death. Diabetic women had higher HR for mortality (HR = 6.645, 95% CI: 1.938-22.79, p = 0.003). Postmenopausal status tended to be associated with, and entailed a six-fold increase in, the risk of death (HR = 6.216, IC: 0.963–40.143, p = 0.055). Alcohol intake was an independent risk factor for mortality (HR = 1.228, 95% CI: 1.014-1.487, p = 0.035). The magnitude of the association between abdominal obesity (WHR ≥ 0.85) and mortality became even greater after the adjustment for BMI (HR = 2.974, 95% CI: 1.039 – 8.510, p = 0.042).Figure 3 shows Kaplan-Meier curves for DM diagnosis. The results support the findings of the adjusted model, according to which mortality is increased in the presence of diabetes.