Seasonal variations in mortality and clinical indicators in international hemodialysis populations from the MONDO registry

This study assessed seasonal variations in mortality and important clinical and laboratory parameters in dialysis patients in a large cohort of dialysis patients. This study adds to the current body of knowledge by exploring for the first time seasonal trends in dialysis patients on a global basis, including countries of both Northern and Southern hemispheres and of temperate and tropical zones. The present study expands our previous observation of major seasonal differences in mortality in a large cohort of chronic HD patients from the US [2].

At a global level, mortality was highest in winter and lowest in spring and summer. In our previously studied US cohort, mortality differences between seasons lost statistical significance when adjusting for seasonally variable risk factors such as blood pressure, albumin and inter-dialytic weight gain. However, the present study had far greater statistical power and clearly indicates that at a global level seasonal differences remained significant after adjustment for seasonally variable risk factors. Most likely, other factors, such as underlying diseases, influence seasonal mortality but are not reflected in the limited palette of clinical and laboratory parameters included in the present study.

Following stratification and adjustment, seasonal variations in mortality appeared more pronounced in northern temperate zones, whereas the variations in mortality in tropical zones did not reach significance, although the tendency to a higher winter mortality was also observed in tropical regions. Admittedly, patient characteristics differed between zones, but also after adjustment for multiple confounders, seasonal variations in mortality were only significant in the temperate zones, although this.

Several general hypotheses can be put forward to explain the apparent differences between climate zones. Firstly, in tropical regions seasons are often more characterized by differences in precipitation and not by classical four-season’s patterns, with accompanying pronounced differences in seasonal temperature. Secondly, seasonal changes in direct sunlight hours which might affect vitamin D levels are less pronounced in tropical as compared to temperate regions. Thirdly, both low winter temperatures and high summer temperatures are also associated with mortality in the general population [13, 14, 11], and the latter effect may be more pronounced in tropical regions.

Our findings are in line with a recent study in general population which showed that seasonal differences in mortality in a US population were lowest in Miami, which has a tropical climate [15]. On the other hand despite much lower seasonal temperature differences, the variation in cardiac mortality was comparable between Honolulu and US cities located in temperate regions [16]. Lastly, although the number of patients in each different zones was substantial, the smaller number of patients in tropical as compared to temperate regions might affect the margin of error in the confidence intervals of mortality differences between seasons, although the pooled tropical zones still contained 9811 patients. Still, the potential differences in seasonal mortality between the various climate zones should be interpreted with caution.

Even in the general population comparative studies of seasonal mortality in different global geographical zones are scarce. In the ISOTHURM study mortality was higher in winter periods, with more pronounced differences for temperate as compared to tropical latitudes [8], showing a significant relation between ambient temperature and mortality. In contrast to the ISOTHURM study, ambient outdoor and indoor temperatures were not available in the present study. Therefore, we cannot make firm statements about the relation between ambient temperature and mortality [8].

The pathophysiological mechanisms underlying the higher winter mortality rates in dialysis patients remain hypothetical. In our previous US study, we found that cardiovascular events, the most important cause of mortality in dialysis patients, contributed predominantly to the higher mortality in winter [2], an observation in agreement with findings in the general population [8]. Both fluid overload, as indicated by higher IDWG, increased peripheral vasoconstriction and release of pro-thrombotic factors [17], as well as a higher incidence of infectious complications and subsequent cardiovascular complications [18], could be responsible for the observed seasonal mortality differences. A recent study in the general population showed that influenza epidemics accounted for 18 % of the seasonal variation in cardiac mortality [16]. Regrettably we were not able to determine the exact causes of death in the present cohort. Also no data on vaccinations are available in our cohort.

Next to the differences in mortality, pre-SBP and IDWG showed seasonal changes in all geographical zones. These were observed both in temperate and - less pronounced - tropical zones. Higher IDWG and pre-SBP have been observed in winter periods in dialysis patients dwelling in Northern [2, 3, 5] and Southern temperate zones [6]. To the best of our knowledge only a single study from Brazil addressed seasonal differences in blood pressure in a tropical climate [19]. This study reported higher diastolic and mean blood pressure in winter, but no difference in IDWG between seasons. The mechanisms behind seasonal differences in SBP have been discussed previously and possibly include volume-related factors, differences in vascular tone or differences in vitamin D₃ levels [2, 3, 5, 6, 20]. The lower IDWG in summer could be related to increased perspiration and/or differences in fluid intake.

Seasonal differences in serum albumin and CRP are somewhat more difficult to interpret. At a global level, CRP levels were lower in the summer period in all geographical zones, which may point to a lower risk of infectious episodes in summer periods [8]. The lower CRP levels are in agreement with the lower neutrophil-to-lymphocyte ratio (NLR) in summer which we observed in our previous study [2]. A recent study from Dopico et al. [21] showed a seasonal expression profile for more than 4,000 protein-coding mRNAs in white blood cells and adipose tissue with inverted patterns observed between Europe and Oceania. They also found increased levels of soluble IL-6 receptor and CRP during European winter, risk biomarkers for cardiovascular and autoimmune diseases with peak in winter. With the exception of southern temperate regions, serum albumin levels were also lower in the summer period. While these results contrast with our previous study in US HD patients [2], in which no seasonal differences in serum albumin levels were observed, they are in agreement with findings from a Japanese dialysis population [7]. The fact that serum albumin levels, a strong marker of survival, were lower in the season when the relative risk of death, as well as CRP levels are also at their nadir appears contradictory. The interpretation of serum albumin levels is complicated since albumin concentrations are the compound resultant of inflammation, nutrition, and fluid status. It is possible that a higher dietary intake during winter [7] explains why serum albumin levels are higher in this season. Data from other studies are similarly inconclusive. In the study by Cheung in North American patients [3], protein intake appeared to be highest in early spring whereas serum albumin levels were highest in October. From the present results, we are not able to explain the discrepant results in Southern temperate zones as data on dietary intake are not available in our cohort. It is important to realize that these observations only imply that in the population studied albumin levels are higher during winter, and not that the relation between mortality and lower serum albumin levels would be different in the MONDO population. In a previous study from the MONDO cohort, we showed significant declining temporal trends in serum albumin levels before death [22].

The findings of the present study may have clinical implications. Both pre-SBP and IDWG are related to outcome in dialysis patients [23, 24]. Higher IDWG can result in hypertension, left ventricular hypertrophy, cardiac failure, and pulmonary congestion [25]. Although data on this aspect are scarce, a recent study in peritoneal dialysis patients suggests that intensive nursing care (including education on fluid management, home blood pressure measurements and intensified antihypertensive treatment) could reduce the seasonal differences in blood pressure [26]. Given the relatively large effect of influenza epidemics on cardiac mortality in the general population, special attention to vaccination for influenza and possibly pneumococcus in this vulnerable population appears warranted [27]. Also cholecalciferol supplementation particularly during sun-deprived seasons in temperate zones was shown to have a beneficial effect on inflammatory and cardiac parameters [28]. However, future studies should address whether the increased mortality during winter in temperate zones is amendable to intervention. Finally, seasonal differences in mortality may be of importance in the design of clinical trials and the interpretation of their results.

Our study has various limitations. Aside from to the absence of data on ambient temperatures, the division in geographical zones and regions is only a rough proxy of climatological and geographical differences. Therefore differences between smaller geographical regions are not captured in this study. In addition, despite the fact that the MONDO consortium includes countries within different geographical zones and the results presented here draw on the currently most diverse international patient-level hemodialysis data, they certainly do not reflect the global hemodialysis population and many important countries with large dialysis populations are currently missing or underrepresented. For example, neither dialysis clinics from Japan nor the large corporate US hemodialysis providers are members of the MONDO initiative. Moreover, because of data logistics and availability some MONDO data, in particular data from a large German provider and some smaller academic providers, were not included in the current analysis. For this reason, the data came primarily from three large and one medium-sized corporate dialysis provider. Despite that limitation we feel that geographical coverage and patient diversity are sufficient to support our conclusions. Nevertheless, it would be interesting to conduct a comparable study including public and academic dialysis centers. Another limitation is that we could not report the mortality rates by causes of death. Also, within each geographical area there is likely to be variation in mortality rates between providers. However, strict provider compliance regulations prevent the publication of their mortality rates. The cosinor method estimates parameters representing cyclic events by computing a least squares fit of a cosine function. While widely and successfully used in the analysis of seasonal data, the cosine method has certain shortcomings; for example, the parameter estimates may be biased if the raw data substantially depart from a cosine function. However, the fact that visual inspection of the data indicated seasonal patterns motivated and reassured us to use the cosinor method.