All-cause and cardiovascular mortality in relation to lung function in the full range of distribution across four Eastern European cohorts

It is unclear whether the dose–response relationship between lung function and all-cause and cardiovascular mortality in the Central and Eastern European populations differ from that reported in the Western European and American populations. We used the prospective population-based HAPIEE cohort that includes randomly selected people with a mean age of 59 ± 7.3 years from population registers in Czech, Polish, Russian and Lithuanian urban centres. The baseline survey in 2002–2005 included 36,106 persons of whom 24,944 met the inclusion criteria. Cox proportional hazards models were used to estimate the dose–response relationship between lung function defined as FEV1 divided by height cubed and all-cause and cardiovascular mortality over 11–16 years of follow-up. Mortality rate increased in a dose–response manner from highest to lower FEV1/height3 deciles. Adjusted hazard ratios (HR) of all-cause mortality for persons in the 8th best, the 5th and the worst deciles were 1.27 (95% CI 1.08‒1.49), 1.37 (1.18–1.60) and 2.15 (1.86‒2.48), respectively; for cardiovascular mortality, the respective HRs were 1.84 (1.29–2.63), 2.35 (1.67–3.28) and 3.46 (2.50‒4.78). Patterns were similar across countries, with some statistically insignificant variation. FEV1/height3 is a strong predictor of all-cause and cardiovascular mortality, across full distribution of values, including persons with preserved lung function.

The prognostic value of impaired lung function in association with mortality is well documented [1][2][3][4][5][6] . It has been shown that reduced levels of forced expiratory volume measured in one second (FEV1) are a better tool in predicting mortality than forced vital capacity (FVC) 7,8 . FEV1 was also an independent predictor of all-cause and cardiovascular mortality and morbidity in population-based studies 9,10 . Impaired FEV1 had a stronger relationship with mortality and cardiovascular events than other common risk factors even at modest range of decrease in FEV1 11 independent from smoking and chronic respiratory diseases (COPD) 12,13 . Evidence is limited on predictive properties of FEV1 among persons with preserved lung function, but earlier studies suggested possible associations between lung function and mortality among those without lung function impairment [14][15][16] .
The conventional approach of lung function evaluation as the percentage of predicted value compared with reference values from healthy population has been criticised [17][18][19] . It less accounts for age-specific height variability and more prone to false positive results 19 . Global Lung Function Initiative (GLI) equation 20 based on standardised Z-score by taking into accounts age, sex, height and ethnicity are a more valid approach but it has some limitations in elderly population 17,19 . Some other methods of standardising spirometry indices including FEV1 divided by height squared and/or cubed have been found a better alternative to predict survival function

Results
Baseline characteristics of the study population. Altogether 36,106 individuals were recruited at baseline, of whom 24,944 met the inclusion criteria ( Fig. 1). For logistic reasons, spirometry was conducted on 25,224 persons. For the same reasons, in the Polish cohort spirometry tests were performed only on random 50% of respondents in the 2nd year of baseline survey. Another reason for missing spirometry data was due to non-response in clinical examinations in Czech and Polish participants. Compared with the top tertile of FEV1/ height 3 , persons in the lowest tertile were those at older ages with highest proportion of women and people with smoking history. The number of underlying chronic cardiovascular and lung diseases and diabetes was also greater for this group (Table 1). In comparison between countries, the prevalence of people in the lowest FEV1/ height 3 tertile was highest in Czech Republic and Lithuania, while a third of the Russian population were in the highest tertile group (Table 1).
Association between lung function and all-cause and cardiovascular mortality. In total, 5069 persons died during a mean follow-up period of 13 years (Supplementary Table 1). More than a third of total deaths was due to cardiovascular causes. Mortality rate increased in a dose-response manner from the highest to lower FEV1/height 3 deciles (mortality rate, 95% confidence interval (CI); 0.86, 0.76-0.96 per 100 person-years among people in the highest and 3.10, 2.91-3.31 per 100 person-years among people in the lowest FEV1/height 3 decile (Table 2)). In the Cox proportional hazards regression model 1 that was adjusted for age, sex and country (Table 2, Supplementary Fig. 3), risk of death was already elevated in the group of people in the 8th decile (HR 1.39, 95% CI 1.19-1.62) and it had a more than threefold increase in the lowest FEV1/height 3 decile (HR 3.33, 95% CI 2.89-3.82). In the model 2 that was adjusted for other baseline covariates (Table 2), being within the 8th decile of FEV1/height 3 was still associated with increased mortality risk (HR 1.27, 95% CI 1.08-1.49), and the risk was doubled in the lowest FEV1/height 3 decile (HR 2.15, 95% CI 1.86-2.48).
Cardiovascular mortality pattern was similar but with stronger association (Table 3, Supplementary Fig. 4). The strong association was observed even in the 9th decile compared to the 10th FEV1/height 3 decile (Table 3,   Supplementary  Fig. 5). In terms of cardiovascular mortality, the highest increase in the risk of death was in Poland where people in the 9th decile of FEV1/height 3 range had more than a threefold increased risk of mortality (Table 5, Supplementary Fig. 6, HR 3.38, 95% CI 1.25-9.16) and it was more than 10-times higher in the lowest FEV1/ height 3 decile (HR 11.1, 95% CI 4.36-28.23) ( Table 5, Supplementary Fig. 6). The country-specific confidence   www.nature.com/scientificreports/ intervals were wide and the differences in HRs between countries were not statistically significant (p for country heterogeneity = 0.31).

Discussion
Using data from the population-based cohort of people from four Central and Eastern European countries, we found a robust dose-response relationship between lung function (defined as FEV1/height 3 ) and all-cause mortality with even stronger effects on cardiovascular mortality. The association was seen throughout all levels of lung function even in people with preserved function. The observed differences between countries were nonsignificant with higher all-cause mortality noticed in Lithuania and strongest association with cardiovascular mortality in Poland. Many previous studies investigated the link between impaired lung function and mortality 1-6 . The predicting properties of lung function in terms of mortality have been evaluated applying different approaches, and the choice of the best one is not straightforward. Lung function prediction equations based on reference values from the general population (FEV1% percent predicted and z-score) have been largely used for categorisation of airway obstruction, however, both have some limitations for the use in the elderly population [17][18][19][20] . The major pitfalls of these approaches arise from the lack of sufficient data on this age group, possible survival bias of the selected people and impact of comorbidities on body size 19 . In our study, we explored the trend in mortality risk and lung function defined as FEV1 divided by height cubed. It has been shown that FEV1 standardised with height is a valid approach for predicting mortality risk in the elderly population 17,23,24,28 . Studies based on FEV1 standardised with height are scarce and the majority of them assessed the mortality risk through the commonly used comparison between healthy and people with the impaired lung function from the lowest tertile or quartile 17,21,22,29 .
We ranked FEV1/height 3 into deciles in order to explore the risk of death across all levels of lung function including individuals with preserved lung function. The association was continuous, graded and considerably stronger for deaths from cardiovascular causes than for all-cause mortality. There is a paucity of studies evaluated the lung function across the whole spectrum in association with mortality. Ashley et al., using data on 2869 subjects from the Framingham Study, found a continuous association between FVC and all-cause mortality 14 . www.nature.com/scientificreports/ Analyses of data on 1195 people of the Buffalo Health Study and 1541 participants of the Rancho Bernardo Study also reported associations with mortality across FEV1 quantiles and quartiles, respectively 15,16 . More recently, Gupta et al. in the study of lifelong non-smokers has shown a consistent association with the risk of death per unit decrement in FEV1 and FVC z-scores as a measure of lung function but in more detailed assessment, the association across lung function quartiles was observed only with deaths from cardiovascular causes 12 . Moreover, the risk increase was more pronounced in groups corresponding to mild and mild to moderate decline in lung function compared to more severe impairment. Similar trend was also found in another study on mortality and cardiovascular adverse events in association with reduced lung function; which suggested that existing thresholds in defining lung function impairment might be misleading in terms of real impact of lung function on health risks 11 .
In line with previous studies, we also found a stronger association with cardiovascular mortality compared to deaths from all-causes. The exact mechanism underlying these associations remains unclear. Some epigenetic aging biomarkers may reflect lung function in elderly people 30 . Previous studies demonstrated that impaired lung function was associated with the severity of coronary atherosclerosis and vascular stiffness 31,32 . It was also an independent risk factor for predicting future cardiovascular disease and diabetes 16,33,34 . Lung function might act via similar mechanisms as other common risk factors (i.e., hypertension, obesity, level of physical activity and smoking) by triggering inflammatory processes that play a causal role in the development of chronic diseases leading to increased mortality 22,35,36 . In our study, the graded association with the risk of death remained strong after adjusting for potential confounders across all deciles suggesting the direct causal effect of lung function with mortality.
Most of the previous studies were performed in western populations, while studies investigating other populations are limited. In the large prospective international multi-ethnic (PURE) cohort study, the risk of death and cardiovascular events was elevated in persons with impaired lung function across all populations and it was highest in people from low-and middle-income countries 11 .
In our study, the association was similar across four countries with different socioeconomic status and health behaviour patterns, with non-significant variation between the countries. Compared to other populations from the Czech Republic and Russia, participants from Poland and Lithuania had higher number of pre-existing conditions that might partly explain the variation in results (Supplementary Table 2). Interestingly, the number of heavy smokers and people with COPD was the highest in Russia, where the effect of lung function was not particularly strong. Previous analyses have shown that lower socioeconomic status throughout a lifespan was associated with poorer lung function 37 and that larger socioeconomic inequalities and higher mortality risk were in Russia 25 . Our study confirmed that lung function is a strong independent predictor of mortality risk, while its contribution to variation in mortality between countries remains unclear.

Strengths and limitations
Our study included people from urban communities of four Central and Eastern European countries. While the study cohort is in general representative for urban populations, it does not include rural areas, and the results are thus not generalizable to entire population of included countries. The response rate was comparable and follow-up time was balanced between countries. Although, some participants with less healthy status from Czech and Polish cohorts did not underwent the baseline clinical examination and that could lead to the underestimation of our results. Large sample size, standardised data collection across all populations, adjustment for a large number of investigating covariates in the association of lung function and mortality is the particular strength of this study. However, the self-report of information in the questionnaire may be a source of reporting bias.
Limited data is available on the use of FEV1/height 3 in defining lung function and further studies needed to validate this approach. However, such an approach does not require reference values and considered as a better tool for predicting mortality risk in the elderly population 17,23,24,28 .
Finally, the nature of our design cannot entirely exclude reverse causation, although the longitudinal design, extensive covariate adjustment of models and long follow-up likely to minimize these limitations.

Conclusion
This study explored the association of lung function with all-cause and cardiovascular mortality in the Central and Eastern European populations. We showed a significant dose-response relationship of lung function with risk of death across the whole spectrum of FEV1/height 3 . The results emphasize the important contribution of lung function to cardiovascular and total mortality risk. If further research confirms the use of FEV1/height 3 in assessing lung function in the elderly population, this measure may have a useful role in predicting future mortality risk in clinical practice.

Study design and participants. The prospective Health, Alcohol and Psychosocial factors in Eastern
Europe (HAPIEE) cohort study has been designed to investigate risk factors for high rates of mortality and cardiovascular diseases in four Central and Eastern European countries (e.g., Czech Republic, Poland, Russia and Lithuania) 38  Spirometry. Lung function test was performed using a Micro-Medical Microplus spirometer. Participants with acute pulmonary infections and illnesses (e.g., vomiting and nausea), recent surgical procedures and cardiovascular conditions (e.g., myocardial infarction and stroke) were excluded from testing 39 . Two or more measurements of FEV1 within 150 ml variation considered for the study 39 . For each participant the highest value of FEV1 was selected. FEV1 was standardised by height and defined as FEV1 divided by height cubed (FEV1/ height 3 ). It was expressed at levels of tertiles and deciles of their distribution for further analyses.
Outcome. The primary outcome was all-cause mortality. Dates of death were obtained from the national or regional (Novosibirsk) death registers in each country. All registers have been shown a complete coverage of deaths 38 . In this study cause-specific mortality was based on underlying causes of death which are determined according to the selection and application rules of ICD-10 maintained by the World Health Organization (WHO).
Covariates. Data on covariates was obtained from questionnaires and medical examination. The selection of variables was based on their known association with mortality 40,41 . For the adjustment we considered age, sex, country, education (primary, secondary education, college or university degree), occupation (employed as entrepreneur or freelancer, housewife, farmer, retired or unemployed), deprivation scale (graded from 1 as a least deprived up to 12 as a most deprived), smoking status (never, current or past heavy smoker (> 30 cigarettes per day), moderate smoker (11-29 cigarettes per day), or light smoker (< 10 cigarettes per day)) 42 , alcohol consumption (never, graduated frequency from 1 to 3 drinks monthly or 1-5 drinks weekly) and physical activity (as number of hours demanding physical activity per week). We also identified the following self-reported comorbidities: stroke, myocardial infarction, ischemic heart disease, hypertension (defined as measured blood pressure > 140/90 mm Hg and self-reported treated hypertension), diabetes (treated and/or untreated), asthma and COPD, any type of surgery (in last 3 months), cancer.
Information on pulmonary symptoms such as cough with or without phlegm (for 3 months) and chest pain were also included in the analyses as markers of respiratory diseases. Confounders obtained during medical examination were weight, height, body-mass index (BMI), blood pressure and cholesterol level.
Statistical analyses. All analyses were performed with Stata (Version 14; StataCorp). Descriptive statistics are presented as means with standard deviations (SD) or frequencies with proportions and were compared across FEV1/height 3 tertiles.
The risk of death in association with FEV1/height 3 was compared by Cox proportional hazards regression models. Deciles of FEV1/height 3 were entered into the model as categorical variable accounting for level of lung function with highest category set as a reference. We used robust variance estimator to account for possible interactions between groups and multiple comparison. Proportional hazards assumptions were confirmed by exploring parallelism of log negative and log estimated survival curves for each covariate (Supplementary Figs. 1,  2). Hazard ratios (HR) with their corresponding 95% confidence intervals (CI) were estimated by crude (included age, sex and country) and confounder-adjusted models.
Dose-response relationship between FEV1/height 3 and mortality was also assessed by entering FEV1/height 3 as a continuous variable into the model. Linear trend was explored by adding quadratic terms and splines.
In addition to all-cause mortality, we investigated cardiovascular mortality. These analyses were restricted to diseases of circulatory system (ICD10 codes I00-I99).
We also conducted separate stratified sensitivity analysis for each country.

Data availability
The data that support the findings of this study are available from the register maintainers but restrictions apply to the availability of these data, which were used under license for the current study, and so are not publicly available. Data are however available from the authors upon reasonable request and with permission of the register maintainers.