Sauna bathing is associated with reduced cardiovascular mortality and improves risk prediction in men and women: a prospective cohort study

This is a self-archived version of an original article. This version may differ from the original in pagination and typographic details. Author(s): Title: Year: Version: Copyright: Rights: Rights url: Please cite the original version: CC BY 4.0 https://creativecommons.org/licenses/by/4.0/ Sauna bathing is associated with reduced cardiovascular mortality and improves risk prediction in men and women: a prospective cohort study © The Author(s), 2018. Published version Laukkanen, Tanjaniina; Kunutsor, Setor K.; Khan, Hassan; Willeit, Peter; Zaccardi, Francesco; Laukkanen, Jari Laukkanen, T., Kunutsor, S. K., Khan, H., Willeit, P., Zaccardi, F., & Laukkanen, J. (2018). Sauna bathing is associated with reduced cardiovascular mortality and improves risk prediction in men and women: a prospective cohort study. BMC Medicine, 16, Article 219. https://doi.org/10.1186/s12916-018-1198-0 2018 RESEARCH ARTICLE Open Access Sauna bathing is associated with reduced cardiovascular mortality and improves risk prediction in men and women: a prospective cohort study Tanjaniina Laukkanen 1,2 , Setor K. Kunutsor 3,4 , Hassan Khan 5 , Peter Willeit 6,7 , Francesco Zaccardi 8 and Jari A. Laukkanen 1,2,9* Abstract Background: Previous evidence indicates that sauna bathing is related to a reduced risk of fatal cardiovascular disease (CVD) events in men. The aim of this study was to investigate the relationship between sauna habits and CVD mortality in men and women, and whether adding information on sauna habits to conventional cardiovascular risk factors is associated with improvement in prediction of CVD mortality risk. Methods: Sauna bathing habits were assessed at baseline in a sample of 1688 participants (mean age 63; range 53–74 years), of whom 51.4% were women. Multivariable-adjusted hazard ratios (HRs) were calculated to investigate the relationships of frequency and duration of sauna use with CVD mortality. Results: A total of 181 fatal CVD events occurred during a median follow-up of 15.0 years (interquartile range, 14. 1–15.9). The risk of CVD mortality decreased linearly with increasing sauna sessions per week with no threshold effect. In ageand sex-adjusted analysis, compared with participants who had one sauna bathing session per week, HRs (95% CIs) for CVD mortality were 0.71 (0.52 to 0.98) and 0.30 (0.14 to 0.64) for participants with two to three and four to seven sauna sessions per week, respectively. After adjustment for established CVD risk factors, potential confounders including physical activity, socioeconomic status, and incident coronary heart disease, the corresponding HRs (95% CIs) were 0.75 (0.52 to 1.08) and 0.23 (0.08 to 0.65), respectively. The duration of sauna use (minutes per week) was inversely associated with CVD mortality in a continuous manner. Addition of information on sauna bathing frequency to a CVD mortality risk prediction model containing established risk factors was associated with a C-index change (0.0091; P= 0.010), difference in −2 log likelihood (P= 0.019), and categorical net reclassification improvement (4.14%; P=0.004). Conclusions: Higher frequency and duration of sauna bathing are each strongly, inversely, and independently associated with fatal CVD events in middle-aged to elderly males and females. The frequency of sauna bathing improves the prediction of the long-term risk for CVD mortality. Keywords: Sauna bathing, Prevention, Cardiovascular disease, Gender, Risk prediction * Correspondence: [email protected] 1 Institute of Public Health and Clinical Nutrition, University of Eastern Finland, P.O. Box 1627, FIN-70211 Kuopio, Finland 2 Central Finland Health Care District, Jyväskylä, Finland Full list of author information is available at the end of the article © The Author(s). 2018 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Laukkanen et al. BMC Medicine (2018) 16:219 https://doi.org/10.1186/s12916-018-1198-0 Background Sauna bathing, a form of passive heat therapy, is a traditional activity in Finland and widely used for relaxation purposes and is becoming increasingly common in many other countries [1–4]. Emerging evidence suggests that sauna bathing is linked with several health benefits, including a reduction in the risk of high blood pressure or hypertension [5,6], stroke [7], neurocognitive diseases [8], and pulmonary diseases [9–11]. Sauna bathing has also been used in treating musculoskeletal pain [12,13] as well as chronic headache [14]. The beneficial effects of sauna bathing on these adverse events have been linked to its positive impact on circulatory and cardiovascular function. It has been suggested that regular heat therapy may improve cardiovascular function via improved endothelium-dependent dilatation, reduced arterial stiffness, modulation of the autonomic nervous system, and lowering of blood pressure [6,15–18]. We have shown that having frequent sauna baths is strongly associated with a reduced risk of fatal cardiovascular outcomes and all-cause mortality in a general population sample of middle-aged men [19]. To our knowledge, this is the only available study [19] on the prospective association between sauna habits and the risk of mortality outcomes. It is therefore unknown whether the additional cardiovascular benefits of frequent sauna bathing are also applicable to women and older individuals. In addition, there is no data on the associations of both weekly frequency and duration of sauna bathing with a risk of cardiovascular disease (CVD) in populations including men and women. Furthermore, given the strong independent association between sauna bathing and the risk of CVD, there is a possibility that adding information on sauna bathing habits to current CVD risk prediction algorithms might be associated with improvements in the ability to predict CVD risk. The potential utility of sauna bathing for CVD risk assessment has not yet been evaluated, and therefore, this warrants investigation. In this context, we aimed to evaluate the relationship between sauna bathing habits (both frequency and duration) and the risk of CVD mortality in a large population-based cohort of middle-aged to elderly men and women. We also investigated the extent to which information on sauna habits could improve the prediction of CVD mortality in our study population using measures of risk discrimination and reclassification. Methods Study design We employed the Kuopio Ischaemic Heart Disease (KIHD) Study, which is a population-based prospective cohort study designed to investigate sauna bathing habits and other risk factors for CVD [19,20]. The KIHD Study was initially based on a cohort of men aged 42–61 years who were living in Kuopio and the surrounding rural communities in the east of Finland. In the 11-year follow-up visit of the first cohort, women were invited to join this study. In this cohort which is being utilized for this analysis, participants (n= 2358) comprised a randomly selected sample of 1351 women and 1007 men aged 53.4 to 73.8 years. Of 2072 eligible participants, 1774 participated in the current prospective sauna study. We excluded 31 participants without information on assessment of sauna bathing habits at baseline examination. Of the remaining participants, complete data on sauna bathing, clinical characteristics, biomarkers, and fatal CVD outcomes were available for 1688 participants (867 women and 821 men) (Fig. 1). All baseline examinations were carried out between March 1998 and December 2001. This study was performed following the STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines for reporting observational studies in epidemiology (Additional file 1: Appendix) [21]. Assessment of sauna bathing In a traditional Finnish sauna, there is dry air with a relative humidity of about 10–20%. It is possible to increase the humidity temporarily by throwing water on the hot rocks of the sauna heater, although it usually remains below 20%. The recommended temperature for sauna is from 80 to 100 °C at the level of the bather’shead,butthe temperature is much lower at the floor level (about 30 °C) which keeps the ventilation of the sauna room efficient and sauna condition comfortable for sauna bathers [5]. The duration of stay in the sauna room depends on the comfort and temperature of the sauna bather, but it usually ranges from 5 to 20 min, although the sessions could be longer depending on the individual [22]. In the current study, sauna bathing was assessed at baseline by a self-administrated questionnaire based on weekly sauna sessions, duration, and temperature in the sauna room [19]. The assessment represents a typical sauna use during the week, and the temperature in the sauna room was measured using a thermometer. The questionnaires were checked by an experienced nurse at the time of baseline examination. Assessment of risk factors and baseline characteristics Risk factors and all other characteristics were assessed during the same visit at study entry. Baseline demographics and socioeconomic and living condition characteristics were assessed among the study participants. A participant who had ever smoked on a regular basis was defined as a smoker. The use of medications, baseline diseases, the level of physical activity, and socioeconomic status (SES) were assessed by self-administered questionnaires [23]. The total and energy expenditure of physical Laukkanen et al. BMC Medicine (2018) 16:219 Page 2 of 14 activity was assessed from a validated 12-month leisure time physical activity questionnaire [24,25]. This detailed quantitative questionnaire deals with the most common leisure time physical activities of middle-aged Finnish men. For the type of physical activity performed, participants were asked to document the frequency (number of sessions per month), average duration (hours and minutes per session), and intensity [26]. Energy expenditure was measured for each physical activity by multiplying the metabolic index of activity (in metabolic equivalent × hour/week) by body weight in kilograms. The diagnosis of chronic diseases and medication was assessed during a medical examination by a doctor. Alcohol consumption was assessed using the Nordic Alcohol Consumption Inventory [27]. Dietary energy intake was assessed using 4-day food recording (Nutricia); instructions were given, and completed food records were checked by a nutritionist. Resting blood pressure was measured between 8 and 10 a.m. with a random-zero sphygmomanometer. Participants were instructed to fast overnight, abstain from alcohol consumption for at least 3 days, and to keep away from smoking for at least 12 h prior to blood specimen collection. The cholesterol contents of serum lipoprotein fractions and triglycerides were measured enzymatically (Boehringer Mannheim, Mannheim, Germany). Serum high-density lipoprotein and its subfractions were separated from fresh serum samples using ultracentrifugation and precipitation. Body mass index (BMI) was computed as the ratio of weight in kilograms to the square of height in meters. Ascertainment of outcomes All CVD deaths that occurred by the end of 2015 were checked against the hospital documents, health center wards and death certificates, and medico-legal reports [28]. There were no losses to follow-up. All participants (just like every individual in Finland) have personal identity codes which are annually matched through computerized linkage with registries for hospitalizations, discharges, and deaths. Annual follow-up for outcomes is also done automatically using the personal identifiers. Registries are also regularly linked with the Central Population Register to ensure that the personal identity codes are correct. Cardiovascular disease deaths were coded using the Tenth International Classification of Diseases codes. Data on incident coronary heart disease (CHD) events from the beginning of the study were based on the national discharge registers [23]. The documents related to the death were cross-checked in detail by two physicians. Statistical analysis Differences in baseline characteristics were examined using the analysis of variance, the independent samples t test, and the chi-squared test. Descriptive data are presented as means (standard deviation, SD) and percentages. Hazard ratios (HRs) with 95% confidence intervals (CIs) for CVD mortality were calculated using Cox proportional hazard models after confirming the assumptions of the proportionality of hazards using Schoenfeld residuals [29]. Subjects were classified into groups on the basis of frequency of sauna bathing (1, 2–3, and 4–7 Fig. 1 Flowchart of the prospective study setting included in the analyses on sauna bathing and fatal cardiovascular outcomes Laukkanen et al. BMC Medicine (2018) 16:219 Page 3 of 14 times per week) and the total weekly duration of a sauna bathing (≤15, 16–45, > 45 min/week) to maintain consistency with previous reports [6,8,19,30]. In a subsidiary analysis, we categorized frequency of sauna bathing into 0–1, 2–3, and 4–7 times per week, including participants who did not use sauna at all (n= 43). Hazard ratios of the associations of frequency and duration of sauna bathing with CVD mortality were progressively adjusted for age and gender (model 1); BMI, smoking, systolic blood pressure (SBP), serum low-density lipoprotein cholesterol (LDL-C), alcohol consumption, previous myocardial infarction, and type 2 diabetes (model 2); total duration of physical activity per week and SES (model 3); and incident CHD events as a time-varying covariate (model 4), as it is a known factor in the pathway for development of CVD mortality. Covariates were selected on the basis of their previously established roles as well-defined predictive or confounding factors, evidence from previous research, or their potential as confounders based on known associations with cardiovascular outcomes and observed associations withsaunaexposureusingtheavailabledata[31]. The cumulative survival from CVDs according to the frequencyanddurationofsaunabathingwascalculated using the Kaplan-Meier method. We explored the shape of the relationship between the frequency of sauna bathing and CVD mortality, using restricted cubicsplinewithknotsatthe5th,35th,65th,and 95th percentiles of the distribution of sauna frequency in a multivariate-adjusted model. We also characterized the shape of the association between duration of sauna bathing and CVD mortality risk by calculating HRs within the quartiles of the duration of sauna bathing and plotted them against mean sauna bathing duration within each quartile using floating absolute risks. We performed subgroup analyses using interaction tests to assess statistical evidence of any differences in HRs across levels/categories of pre-specified clinically relevant characteristics such as age at survey, gender, BMI, SBP, total cholesterol, LDL-C, high-density lipoprotein cholesterol (HDL-C), total duration of physical activity per week, energy expenditure of physical activity, history of diabetes mellitus, smoking status, history of hypertension, and prevalent CHD. To minimize biases due to reverse causation, sensitivity analysis involved excluding the first 5 years of follow-up. To assess whether adding information on the frequency of sauna bathing (main exposure) to conventional cardiovascular risk factors would result in an improvement in the prediction of CVD mortality risk, we calculated measures of discrimination for censored time-to-event data (Harrell’s C-index [32]) and reclassification [33,34]. To investigate the change in C-index on the addition of frequency of sauna bathing, two CVD mortality risk prediction models were fitted: one model based on traditional risk factors (i.e., age, sex, SBP, history of diabetes, total cholesterol, HDL-C, and smoking) and the second model with these risk factors plus frequency of sauna bathing. Reclassification analysis was restricted to the first 10 years of follow-up and was assessed using the net reclassification improvement (NRI) [33] and integrated discrimination improvement (IDI) [33]. Reclassification analysis was based on predicted 10-year CVD mortality risk categories of low (< 1%), intermediate (1 to < 5%), and high (≥5%) risk as previously reported [35]. Given that Harrell’sC-index can be very insensitive in detecting differences in risk prediction analyses [36,37], to avoid discarding potential biomarkers that can be used in risk prediction, it has been recommended to also use sensitive risk discrimination methods such as the −2 log likelihood test [36,37]. Therefore, in addition to Harrel’s C-index, we tested for differences in the −2 log likelihood of prediction models with and without the inclusion of frequency of sauna bathing. A Pvalue < 0.05 was considered statistically significant. Statistical analyses were performed using Stata version 12 (Stata Corp, College Station, TX). Results Baseline characteristics A summary of the baseline characteristics of overall study participants and according to the group of weekly frequency of sauna bathing is shown in Table 1. There were 867 (51.4%) female and 821 (48.6%) male participants. The mean (SD) age, BMI, and waist-to-hip ratio were 63 years (7), 27.9 kg/m 2 (4.4), and 0.91 (0.09), respectively. The median (interquartile range, IQR) frequency and duration of sauna bathing were two (one to three) sessions and 30 min (15–45) per week, respectively. The mean (SD) temperature of the sauna bath was 75.9°C (9.9). The average temperature of sauna room was slightly lower (74.8 °C) among participants who had four to seven sauna bathing sessions per week compared to those with only 1 sauna bathing session per week (77.4 °C). Participants with a frequency of sauna bathing of four to seven sessions per week had higher BMI and alcohol and energy intake, compared to those with 1 sauna session per week. When comparing men to women in terms of median frequency and duration of sauna bathing, the median (IQR) values were two (two to three) vs. two (one to two) sessions per week and 30 (20–45) vs. 20 min (13–30) per week, respectively; the mean (SD) temperature of the sauna bath was 77.1 (9.0) vs. 74.7 °C (10.5) for men and women, respectively. Sauna bathing and fatal cardiovascular events During a median (interquartile range) follow-up of 15.0 years (14.1–15.9) (23,601 person-years at risk), a Laukkanen et al. BMC Medicine (2018) 16:219 Page 4 of 14 Table 1 Baseline characteristics of overall study participants and according to frequency of sauna bathing Characteristics Frequency of sauna bathing (times per week) Overall (N= 1688) 1 (n= 455) 2–3(n= 1028) 4–7(n= 205) Pvalue for heterogeneity Mean (SD) or n (%) or median (IQR) Mean (SD) or n (%) or median (IQR) Mean (SD) or n (%) or median (IQR) Mean (SD) or n (%) or median (IQR) Sauna use Temperature, °C 75.9 (9.9) 77.4 (9.4) 75.4 (9.7) 74.8 (10.9) < 0.001 Duration, minutes/sauna session, median (IQR) 13 (10–15) 10 (10–15) 15 (10–20) 13 (10–15) 0.073 Duration, minutes/week, median (IQR) 30 (15–40) 10 (10–15) 30 (20–40) 60 (40–90) < 0.001 Demographics Age, years 63 (7) 64 (7) 63 (6) 60 (6) < 0.001 Male, n(%) 821 (48.6) 177 (38.9) 512 (49.8) 132 (64.4) < 0.001 Body mass index, kg/m 2 27.9 (4.4) 27.4 (4.4) 28.1 (4.4) 28.2 (4.6) 0.013 Systolic blood pressure, mmHg 136 (17) 137 (18) 136 (17) 135 (17) 0.356 Diastolic blood pressure, mmHg 81 (9) 81 (9) 81 (9) 82 (10) 0.259 Alcohol consumption, g/week, median (IQR) 12.20 (1.00–53.68) 8.75 (0.32–45.11) 12.53 (1.29–53.34) 24.00 (3.20–76.60) 0.004 Smokers, n(%) 221 (13.1) 70 (15.4) 132 (12.8) 19 (9.3) 0.091 Smoking, pack years* 3.05 (10.3) 3.7 (11.4) 2.9 (10.1) 2.1 (8.3) 0.174 Total physical activity per week, h, median (IQR) † 7.94 (4.60–13.21) 7.31 (4.16–12.05) 8.21 (4.80–13.21) 8.48 (4.78–14.76) 0.029 Physical activity, MET h/year, median (IQR) † 1817 (1077–2992) 1625 (948–2703) 1874 (1109–3055) 2012 (1256–3296) < 0.001 Energy expenditure of physical activity, kcal/day, median (IQR) † 383 (224–598) 325 (186–512) 399 (237–610) 449 (293–723) < 0.001 Mean intensity of physical activity, METs † 4.58 (1.02) 4.43 (1.02) 4.61 (1.02) 4.76 (1.03) < 0.001 Energy intake, kJ/day 7612 (2397) 7146 (2263) 7688 (2393) 8250 (2524) < 0.001 Previous myocardial infarction, n(%) 118 (7.0) 34 (7.5) 70 (6.8) 14 (6.8) 0.895 History of coronary heart disease, n(%) 474 (28.1) 128 (28.1) 289 (28.1) 57 (27.8) 0.996 Type 2 diabetes, n(%) 138 (8.2) 45 (9.9) 80 (7.8) 13 (6.3) 0.233 Hypertension, n(%) 702 (41.6) 196 (43.1) 421 (41.0) 85 (41.5) 0.746 Serum LDL cholesterol, mmol/L 3.59 (0.93) 3.56 (0.95) 3.61 (0.93) 3.59 (0.87) 0.598 Serum HDL cholesterol, mmol/L 1.25 (0.31) 1.26 (0.33) 1.24 (0.30) 1.26 (0.33) 0.356 Fasting blood glucose, mmol/L 5.1 (1.2) 5.1 (1.3) 5.1 (1.2) 5.1 (1.2) 0.985 Socioeconomic and living condition characteristics Socioeconomic status, unit ‡ 10.9 (4.7) 10.3 (4.7) 11.3 (4.7) 10.3 (4.5) < 0.001 Annual income (1998–2001), €16,144 (11,715) 16,970 (10,474) 15,337 (10,703) 18,377 (17,403) 0.001 Academic degree (college or university), n(%) 93 (5.5) 47 (10.3) 40 (3.9) 6 (2.9) < 0.001 Daily working time (duration), h 8.1 (1.8) 7.9 (1.7) 8.1 (1.7) 8.7 (2.4) < 0.001 Physical strain of work, unit 2.40 (0.88) 2.35 (0.89) 2.42 (0.87) 2.43 (0.87) 0.402 Mental strain at work, unit 2.47 (0.69) 2.47 (0.72) 2.46 (0.68) 2.50 (0.68) 0.748 Type of residence, n(%) < 0.001 Family house 825 (48.9) 106 (23.4) 574 (55.8) 145 (70.7) Attached house 219 (13.0) 75 (16.5) 123 (12.0) 21 (10.2) Apartment house 643 (38.1) 273 (60.1) 331 (32.2) 39 (19.0) Summer cottage (own available), n(%) 804 (47.8) 190 (41.9) 512 (50.1) 102 (49.8) 0.013 Complete baseline information was available on 1688 individuals IQR interquartile range, SD standard deviation, LDL low-density lipoprotein, HDL high-density lipoprotein *Pack-years denotes the lifelong exposure to smoking which was estimated as the product of years smoked and the number of tobacco products smoked dailyat the time of examination † Physical activity was computed by multiplying the duration and intensity of each physical activity by body weight. Physical activity was assessed using the 12month physical activity questionnaire ‡ Socio-economic status is a summary index that combines measures of income, education, occupation, occupational prestige, material standard of living, and housing conditions, all of which were assessed with self-reported questionnaires Laukkanen et al. BMC Medicine (2018) 16:219 Page 5 of 14 total of 181 CVD deaths occurred. Cardiovascular mortality rates per 1000 person-years across the three frequency groups of sauna bathing (one, two to three, and four to seven times per week) were 10.1 (95% CI 7.9 to 12.9), 7.6 (6.3 to 9.2), and 2.7 (1.3 to 5.4), respectively. According to the frequency of sauna bathing, cumulative hazard curves demonstrated the lowest risk of CVD mortality among participants who had four to seven sauna sessions per week compared to other groups (P< 0.001 for the log-rank test; Fig. 2). A restricted cubic spline curve shows the risk of CVD mortality decreased linearly with increasing sauna sessions from one to seven (Pvalue for non-linearity = 0.932) (Fig. 3). In the analyses adjusted only for age and sex, compared to participants who had one sauna session per week, the HRs of CVD mortality were 0.71 (95% CI 0.52 to 0.98) and 0.30 (0.14 to 0.64) for participants with two to three and four to seven sauna sessions per week, respectively (Table 2). Additional adjustment for several established risk factors and potential confounders minimally attenuated the HRs: 0.77 (95% CI 0.56 to 1.07) for two to three sauna bathing sessions per week and 0.36 (0.17 to 0.77) for four to seven sauna bathing sessions per week. The corresponding HRs (for two to three and four to seven sauna bathing sessions per week) remained consistent after adjustment for incident CHD as a time-varying covariate: 0.75 (95% CI 0.52 to 1.08) and 0.23 (0.08 to 0.65), respectively (Table 2). The results remained similar to additional adjustment for the temperature of sauna bathing. In the analyses by gender, there was no statistically significant evidence of associations in women, which could be attributed largely to the low event rates in the sauna exposure categories (Table 2). A test of interaction showed that the association between sauna bathing frequency and CVD mortality was not significantly modified by gender (Pfor interaction = 0.524). Cardiovascular mortality rates per 1000 person-years of follow-up across the three groups of sauna bathing duration (≤15, 16–45, > 45 min/week) were 9.6 (95% CI 7.5 to 12.3), 7.6 (6.2 to 9.3), and 5.1 (3.4 to 7.7), respectively. Cumulative hazard curves demonstrated a greater risk of CVD mortality among participants having a sauna bath of ≤15 min/week compared with the other groups (P=0.028 for the log-rank test; Fig. 2). In the analysis adjusted for (i) age and sex and (ii) BMI, smoking, SBP, serum LDL-C, alcohol consumption, previous myocardial infarction, and type 2 diabetes, an inverse association was found between duration of sauna bathing and CVD mortality risk, which was potentially consistent with either a curvilinear or linear shape (Fig. 4). However, statistical tests suggested a fit with a non-linear shape (Pfor non-linearity = 0.005). After adjustment for age and gender, HR was 0.49 (0.30–0.80) for CVD mortality among participants in the highest weekly duration (> 45 min/week) compared with the lowest weekly duration (≤15 min/week) of sauna bathing (Table 3). The respective HR was 0.57 (0.35–0.94) after adjustment for several established CVD risk factors, and potential confounders. The respective HRs remained consistent on further adjustment for the temperature of sauna bathing. In gender-specific analyses, there was no statistically significant evidence of associations in both men and women, which could be attributed to the low event rates (Table 3). A test of interaction showed that the association between the duration of sauna bathing and CVD mortality was not significantly modified by gender (Pfor interaction = 0.314). The associations of both frequency and duration of sauna bathing with CVD mortality risk remained Fig. 2 Cumulative Kaplan-Meier curves for cardiovascular mortality according to the frequency and duration of sauna bathing per week Laukkanen et al. BMC Medicine (2018) 16:219 Page 6 of 14 consistent in the analyses that excluded the first 5 years of follow-up (Additional file 2: Tables S1–S2). In a subsidiary analysis which compared four to seven sauna sessions per week with zero to one sauna session per week, the associations were similar (Additional file 2: Table S3). Associations in subgroups Figures 5and 6show the associations of frequency and duration of sauna bathing with the risk of CVD death in clinically relevant subgroups. Except for the evidence of effect modification by diabetes status for the association between sauna frequency and CVD mortality (Pfor interaction = 0.021), the associations did not vary significantly by levels or categories of several clinically relevant characteristics. Frequency of sauna bathing and CVD mortality risk prediction A risk prediction model for CVD mortality containing conventional risk factors yielded a C-index of 0.7716 (95% CI 0.7382–0.8049; P< 0.001). After addition of information on the frequency of sauna bathing, the C-index was 0.7807 (0.7486–0.8128; P< 0.001), representing a significant increase of 0.0091 (0.0022–0.0160; P= 0.010). In addition, when investigating differences in the -2 log likelihood of the risk score with and without the inclusion of frequency of sauna bathing, the -2 log likelihood was significantly improved on the addition of information on the frequency of sauna bathing to the model (Pfor comparison = 0.019). There was a significant improvement in the classification of participants into predicted 10-year CVD mortality risk categories (NRI: 4.14%, 1.30–6.97%; P= 0.004). The IDI was 0.0037 (0.0002–0.0072; P= 0.041). Fig. 3 Restricted cubic spline model of the hazard ratios of cardiovascular mortality with the frequency of sauna bathing. Restricted cubic spline functions were analyzed with knots located at 5th, 35th, 65th, and 95th percentiles of sauna bathing frequency distribution, with the reference category set at one session/week; adjusted for age, gender, body mass index, smoking, systolic blood pressure, serum low-density lipoprotein cholesterol, alcohol consumption, previous myocardial infarction, and type 2 diabetes. The dashed lines represent the 95% confidence intervals Table 2 Hazard ratios of cardiovascular mortality according to the frequency of sauna bathing, overall and among men and women Frequency of sauna bathing (sessions/week) Events/ total Model 1 Model 2 Model 3 Model 4* HR (95% CI) Pvalue HR (95% CI) Pvalue HR (95% CI) Pvalue HR (95% CI) Pvalue Overall Once 63/455 Ref Ref Ref Ref 2–3 110/1028 0.71 (0.52 to 0.98) 0.035 0.78 (0.57 to 1.08) 0.133 0.77 (0.56 to 1.07) 0.121 0.75 (0.52 to 1.08) 0.120 4–7 8/205 0.30 (0.14 to 0.64) 0.002 0.36 (0.17 to 0.76) 0.007 0.36 (0.17 to 0.77) 0.008 0.23 (0.08 to 0.65) 0.005 Men Once 39/177 Ref Ref Ref Ref 2–3 71/512 0.61 (0.41 to 0.90) 0.013 0.70 (0.47 to 1.03) 0.073 0.69 (0.46 to 1.03) 0.069 0.68 (0.43 to 1.09) 0.111 4–7 8/132 0.33 (0.15 to 0.71) 0.005 0.39 (0.18 to 0.84) 0.016 0.39 (0.18 to 0.84) 0.016 0.26 (0.09 to 0.75) 0.013 Women Once 24/278 Ref Ref Ref Ref 2–3 39/516 0.95 (0.57 to 1.57) 0.830 1.03 (0.59 to 1.77) 0.929 1.00 (0.57 to 1.74) 0.997 0.88 (0.48 to 1.60) 0.676 4–7 0/73 NE NE NE NE Model 1: adjusted for age and gender Model 2: model 1 plus body mass index, smoking, systolic blood pressure, serum low-density lipoprotein cholesterol, alcohol consumption, previous myocardial infarction, and type 2 diabetes Model 3: model 2 plus physical activity (duration per week) and socio-economic status Model 4: model 3 plus incident coronary heart disease as a time-dependent covariate CI confidence interval, HR hazard ratio, NE not estimated because of zero event rate; analysis is based on 1688 participants and 181 cardiovascular deaths *The model was limited to the population at risk and did not include those who already had coronary heart disease Laukkanen et al. BMC Medicine (2018) 16:219 Page 7 of 14 Discussion The findings of this long-term prospective study of over 14 years follow-up suggest that the cardiovascular benefits of sauna bathing may exist in both men and women. Our new results show that addition of information on the frequency of sauna bathing improved the prediction and reclassification of the long-term risk for CVD mortality. A higher frequency of sauna bathing sessions per week was related to a decreased risk of fatal CVD events independent of conventional cardiovascular risk factors as well as several other potential confounders. The risk of fatal CVD events decreased with increasing sauna (A) (B) Fig. 4 Hazard ratios for cardiovascular mortality by quartiles of the duration of sauna bathing. aAdjusted for age and gender. bAdjusted for age, gender, body mass index, smoking, systolic blood pressure, serum low-density lipoprotein cholesterol, alcohol consumption, previous myocardial infarction, and type 2 diabetes. CI, confidence interval Table 3 Hazard ratios of cardiovascular mortality according to the weekly duration of sauna bathing, overall and among men and women Duration of sauna bathing (minutes/week) Events/ total Model 1 Model 2 Model 3 Model 4* HR (95% CI) Pvalue HR (95% CI) Pvalue HR (95% CI) Pvalue HR (95% CI) Pvalue Overall ≤15 62/463 Ref Ref Ref Ref 16–45 96/906 0.69 (0.50 to 0.96) 0.027 0.77 (0.55 to 1.06) 0.112 0.77 (0.55 to 1.07) 0.123 0.74 (0.51 to 1.09) 0.132 > 45 23/319 0.49 (0.30 to 0.80) 0.004 0.57 (0.35 to 0.93) 0.025 0.57 (0.35 to 0.94) 0.028 0.60 (0.34 to 1.05) 0.074 Men ≤15 29/157 Ref Ref Ref Ref 16–45 72/461 0.83 (0.54 to 1.28) 0.406 0.92 (0.59 to 1.43) 0.711 0.92 (0.59 to 1.43) 0.705 1.05 (0.61 to 1.83) 0.846 > 45 17/203 0.50 (0.27 to 0.91) 0.023 0.57 (0.31 to 1.04) 0.068 0.57 (0.31 to 1.05) 0.073 0.68 (0.33 to 1.43) 0.311 Women ≤15 33/306 Ref Ref Ref Ref 16–45 24/445 0.53 (0.31 to 0.89) 0.017 0.62 (0.36 to 1.08) 0.093 0.61 (0.35 to 1.07) 0.084 0.51 (0.28 to 0.91) 0.024 > 45 6/116 0.58 (0.24 to 1.39) 0.222 0.75 (0.31 to 1.83) 0.528 0.75 (0.31 to 1.85) 0.532 0.66 (0.25 to 1.76) 0.408 Model 1: adjusted for age and gender Model 2: model 1 plus body mass index, smoking, systolic blood pressure, serum low-density lipoprotein cholesterol, alcohol consumption, previous myocardial infarction, and type 2 diabetes Model 3: model 2 plus physical activity (duration per week) and socio-economic status Model 4: model 3 plus incident coronary heart disease as a time-dependent covariate CI confidence interval, HR hazard ratio; analysis is based on 1688 participants and 181 cardiovascular deaths *The model was limited to the population at risk and did not include those who already had coronary heart disease Laukkanen et al. BMC Medicine (2018) 16:219 Page 8 of 14