The Prognostic Implications of Night Sweats in Two Cohorts of Older Patients

Geriatric Continuity Clinic Database

The Geriatric Continuity Clinic (GCC) was a primary care continuity clinic serving patients older than 65, operated by 2 family physicians with certificates of added qualifications in geriatrics at the University of Oklahoma Health Sciences Center between 1989 and 2001. All patients seen in the GCC were required to complete a 34-page health history questionnaire and undergo a complete history and physical examination as a requirement for enrollment in the practice. The waiting list for an initial appointment was generally 6 to 12 weeks. As a result, patients were not usually acutely ill when first seen.

In 1999, the data from the initial evaluation of all patients enrolled during the prior 10 years (n = 740) were transferred to a research database. Deaths have subsequently been tracked clinically using the University's billing system and the Social Security Death Index. Permission was granted by the University of Oklahoma Health Sciences Center's Institutional Review Board to conduct the following analyses. This dataset has not been used in previous studies of night sweats.

Included in the initial questionnaire was a comprehensive review of systems consisting of 87 items of the form: “In the past 12 months have you noticed….” One of the items in the review of systems was “night sweats,” answered “yes” or “no.”

Age, sex, race, body mass index (BMI), number of grades completed in school, current cigarette use (yes/no), activities of daily living (ADLs) and instrumental activities of daily living (IADLs; 14-point scale for each), and several selected medical problems were also available for analysis. Frequencies and measures of central tendency were calculated for all variables. Fifty-eight questionnaires that were missing information about night sweats were dropped from the study.

Bivariate associations between reported night sweats (yes/no) and death were tested for statistical significance using the χ² test. Bivariate analyses and logistic regression were used to identify other potential predictors of mortality. The Cox proportional hazards method was used to identify associations between night sweats and earlier mortality, controlling for variables associated with death in the bivariate and/or logistic analyses. We were concerned that controlling for measures of health and functional status might obscure actual associations between night sweats and mortality through a causal chain. We therefore initially examined models that included only age, sex, income, education, BMI, and diagnosis of dementia (plus night sweats), and then examined more complete models that also included measures of health status and disability. We repeated the analyses for patients without dementia because their ability to report symptoms could be questioned.

Oklahoma Longitudinal Assessment of the Health Outcomes of Mature Adults Studies Dataset

The Oklahoma Longitudinal Assessment of the Health Outcomes of Mature Adults (OKLAHOMA) Studies dataset has been described previously.^2–4 Briefly, 834 individuals 65 years of age and older were recruited from the billing record lists of 23 family physicians in central Oklahoma in 1999 and 2000. Patients with known dementia and those living in nursing homes were specifically excluded. Those who chose to participate in the OKLAHOMA Studies were more likely than nonparticipants to be younger (P < .001); men (43% vs 37%; P = .002); better educated (P < .0001); and have better (self-reported) health (P < .0001).

Each participant completed a lengthy questionnaire that included sociodemographic information, health habits, past medical problems, current symptoms, self-rated health (100-point rating scale), and functional status instruments (ADLs and IADLs). A trained research nurse conducted several tests, including a 50-foot timed walk (at usual pace). Participants were re-enrolled annually for 4 years, at which times the questionnaires and examinations were repeated.

The following question about night sweats was included in the questionnaires completed in years 1 and 2: “During the last month, how much trouble have you had with night sweats?” Response options included, “no trouble,” “a little trouble,” “some trouble,” “a fair amount of trouble,” and “a great deal of trouble.” We converted these responses to 2 dichotomous variables called “night sweats” (including those with some, a fair amount, or a great deal of trouble with night sweats) and “severe night sweats” (including only those with a fair amount or a great deal of trouble with night sweats). An insufficient number of participants (n = 12) reported having “a great deal of trouble” with night sweats, so we were unable to accurately analyze this group separately. Deaths have been tracked by the study coordinator from participant contacts, primary care clinician reports, and the social security death index. Permission was granted by the University of Oklahoma Health Sciences Center's Institutional Review Board to conduct the analyses.

Descriptive statistics were calculated for all variables. Bivariate analyses were used to identify potential associations between “night sweats” and “severe night sweats” and mortality. We then conducted Cox proportional hazards analyses to determine whether either of the night sweats variables were associated with earlier mortality after controlling for other predictors. Again, we examined models that included only age, sex, income, education, and BMI (plus night sweats), as well as more complete models including measures of health status and disability.

GCC Data Analyses

Descriptive statistics for relevant variables can be found in Table 1. This was, on average, a very old population of patients. Their mean age was 84.1 years (SD, 7.2 years) and their mean BMI was 25.6 (SD, 5.5). The average duration of participation in the cohort from enrollment in the GCC through November 1, 2008 or until death was 7.3 years.

Fourteen percent of patients in the cohort reported having experienced night sweats. Patients with dementia were less likely to report night sweats, presumably because of a lowered ability to report symptoms. Having reported night sweats was not associated with age, sex, race, BMI, education, or functional status (ADLs, IADLs). Eighty-one percent died during the study period. Bivariate associations between categorical variables and death are shown in Table 1. Both younger age (P = .02) and lower BMI (P < .001) were also associated with death in these analyses.

Based on the logistic models, predictors of death included younger age, male sex, a lower BMI, a lower ADL score, a lower IADL score, and a diagnosis of dementia. Night sweats was not a predictor (P = .61). The results were the same for patients without dementia except that age dropped out of the model.

The most parsimonious proportional hazards models included younger age, male sex, lower BMI, lower ADL and IADL scores, and dementia but not night sweats. Once again, removal of patients with dementia did not substantially affect the models. The models are shown in greater detail in Table 2.

OKLAHOMA Studies Data Analyses

Data from 843 patients were analyzed. The average duration of participation in the cohort from enrollment through November 1, 2008, or death was 7.5 years. At baseline, the mean age of this cohort was 80.9 years (SD, 6.2 years). Their average BMI was 28.7 (SD, 5.6) and their average self-rated health status, on a 0 to 100 scale, was 76.14 (SD, 17.02). It took them an average of 17.1 seconds (6.8 seconds) to walk 50 feet at their usual pace.

Ninety-one participants (9.7%) reported night sweats and 31 (3.3%) reported severe night sweats at baseline. Of the 59 patients reporting night sweats at baseline who re-enrolled in year 2, 28 (47%) still reported night sweats. Of the 22 reporting “severe night sweats” at baseline and who were re-enrolled, 13 (59%) still reported night sweats in year 2, and 8 (36%) still reported severe night sweats. Of the 566 re-enrolled participants who did not report night sweats at baseline, 27 (5%) reported night sweats in year 2.

During the period of time included in this study, 204 participants (24%) died. Descriptive statistics and bivariate associations between relevant categorical variables and mortality are shown in Table 3. Both younger (65–74 years old) and older (85+) participants were more likely to die. Lower BMI (P = .05), longer time to walk 50 feet (P < .0001), and lower health rating (P < .0001) were also associated with death. Neither night sweats nor severe night sweats were associated with death.

Variables associated with night sweats included younger age (P = .04), less education (P = .02), lower ADL score (P < .0001), lower IADL score (P = .002), and lower self-rated health (P < .0001). Having severe night sweats was associated with nonwhite race (P = .04), lower income (P = .04), lower ADL (P = 0005) and IADL (P = .003) scores, and lower self-rated health (P = .03).

After controlling for age, sex, BMI, education, and income, individuals who reported night sweats or severe night sweats were no more likely to die (Table 4). The most parsimonious proportional hazards models are shown in Table 4. Adding measures of functional status (ADL and IADL scores and gait time) to the logistic and proportional hazards models further reduced the strength of association between the night sweats variables and mortality.

Limitations

Our study has obvious limitations. Neither cohort was representative of the general population of the United States. The GCC cohort was essentially a referral population, even although most patients were self-referred. The OKLAHOMA Studies cohort, which was drawn from 9 different primary care practices, was biased toward healthier women with more education. The fact that night sweats were not associated with increased mortality in either cohort suggests that the same might be true of other populations of older individuals, but it certainly does not prove it.

Both cohorts were quite old. Other causes of death may have diluted the impact of less lethal conditions causing night sweats. Older people are less likely to sweat excessively for a variety of reasons, which could have skewed our results in either direction. Our ability to identify patients with severe night sweats was limited by the data and by the small number of people in the category of patients most bothered by night sweats within the OKLAHOMA Studies cohort.

The questions asked about “night sweats” were nonspecific. In other studies we have used the following definition: “sweating at night even when it isn’t excessively hot in your bedroom.” Others have used more narrow and precise definition, “episodes of significant nighttime sweating that soaks your bed clothes or bedding”⁸ or “drenching sweats that require the patient to change bedclothes.”⁹ Because the data used in this study had already been collected and was not specifically obtained for the purpose of studying night sweats, we had no way to influence the wording of the questions.

Regarding unexpected or confusing findings, we assume that the association between younger age and mortality in the GCC cohort reflected the fact that people chose to come to this special geriatric clinic for different reasons. Those who came at a younger age were evidently sicker. In the OKLAHOMA Studies sample female sex was associated with increased mortality, after controlling for other predictors. That was unexpected and must have something to do with selection bias. Women were evidently more willing to participate in the study despite their illnesses than were men, who may have used illness as a reason not to enroll. The association between more education and mortality in the OKLAHOMA Studies cohort is harder to explain, but suggests some more complex form of selection bias or an interaction not obvious to us.