Mortality Risk Factors in the China Dialysis Outcomes and Practice Patterns Study (DOPPS)

Background Mortality risk for hemodialysis (HD) patients varies by country and ethnicity. Here, mortality rate and its related risk factors in Chinese HD patients from the Dialysis Outcomes and Practice Patterns Study (DOPPS) were investigated. Methods Data from China DOPPS phase 5 (2012–2015) were used. Patients’ demographics, assigned primary causes of end stage Kidney disease (ESKD), comorbidities, dialysis prescription, laboratory values, date and cause for death were analyzed. Cox proportional hazards models were used to assess the association of patient characteristics and treatments with mortality. Results 1427 HD patients were enrolled. The mean age was 59.4 ± 14.9 years. The median follow-up time was 1.9 (1.1–2.1) years. There was total 205 deaths with at least 103 from cardiovascular disease (50.2%). The overall mortality rate was 8.8 per 100 patient-years. In the multivariate COX model, older, serum albumin (Alb < 4g/dl, blood platelets < 100*10 9 /L, pulse pressure (PP) > 63mmHg, and congestive heart failure history were independent risk factors for all-cause mortality. Conclusions Attention should be paid to patients who were older, with lower Alb and blood platelets level, higher PP and congestive heart failure history. Our results highlighted that there might be some modiable risk factors for patients’ survival, such as Hgb, Alb, blood platelets, and blood pressure management.


Introduction
Hemodialysis (HD) therapy is a life-saving and life-sustaining procedure that improves the life expectancy of patients with end stage Kidney disease (ESKD). However, the adjusted all-cause mortality rate was 6.5-7.9 times greater for dialysis patients than for individuals in the general population [1,2]. Understanding the mortality risk and its in uencing factors in HD patients are of great interest and value, especially those modi able risk factors. We previously reported that the crude mortality rate for maintenance HD patients in Beijing was much lower than that in the United State [3]. However, our previous analysis only included patients from one major city in China, and lacked information about risk factors for mortality, and was limited in sample size. Page 3/16 was the only country from Asia. China is a large developing Asian country and has an increasing HD population. However, the related information we hold about Chinese HD population was lacking. It is necessary to investigate the mortality rate and risk factors for Chinese HD patients and to optimize the treatment practices for its ESKD patients. Therefore, we analyzed the data of China DOPPS phase 5 (2012)(2013)(2014)(2015) to describe the overall death rate and explore modi able risk factors for patient mortality.
Results 1427 HD patients were enrolled in this study. The mean age was 59.4 ± 14.9 years. The median follow-up time was 1.9 (interquartile range = 1.1, 2.1) years. According to the outcome status, we divided patients into survived and died groups, and characteristics between groups were compared ( Table 1). Patients in died group differed from survived patients in many ways: for instance, they were older, with a higher proportion of catheter use, had lower Hgb and Alb. Meanwhile, died patients had higher pre-dialysis and post-dialysis diastolic blood pressure (DBP) than survived patients. Compared with survived patients, patients who died had a higher proportion of diabetes, CHF and other cardiovascular disease, cerebrovascular disease, fracture, hypertension, osteoporosis, lung disease and peripheral vascular disease (Table 1).  We use univariate COX regression analysis to select variables related to all-cause mortality from proposed clinically relevant ones (Table 3). Variables with P < 0.15 were entered as possible factors in the multivariate COX regression model. We found that age, Alb, blood platelets, history of CHF and PP were related to all-cause mortality in the nal model (Table 4). For each 10-years increasing of age, the risk of all-cause mortality rose by 52% (HR = 1.52, 95%CI: 1.35-1.71). Compared with Alb ≥ 4.0 g/dl, patients with Alb in 3.5-3.9 and < 3.5 g/dl had an increased risk of death after adjustment (HR = 1.50, 95%CI: 1.10-2.11 and HR = 2.45, 95%CI: 1.65-3.63, respectively). Meanwhile, low blood platelets (< 100*10 9 /L) were related to a higher death rate (HR = 1.97, 95%CI: 1.42-2.72) than normal blood platelets (100-299*10 9 /L). Patients who had a history of CHF (HR = 1.57, 95%CI: 1.16-2.11) tend to have a higher mortality risk than patients without this comorbidity. Patients with higher PP (≥ 63mmHg) had a 1.41fold higher risk for all-cause mortality (95% CI, 1.01-1.96; P = 0.042).

Discussion
In this large prospective study of 1,427 Chinese HD patients, we reported the all-cause mortality rate and explored related risk factors. The overall mortality rate was 8.8 per 100 patient-years in our patients. Cardiovascular death was the leading cause of death, which account for more than half of all deaths.
Patients that were older, with lower Alb, blood platelets, higher PP and had a history of CHF had a higher risk of all-cause mortality.
We reported a lower overall mortality rate compared with other western DOPPS countries, such as North America (14.3 per 100 patient-years) and Europe countries (13.1 per 100 patient-years) [9]. But it was a little higher than that in Japan (7.87 per 100 patient-years) [10]. Prior DOPPS analyses noted that many bene cial dialysis practices (such as high use of arteriovenous stula, longer or more frequent dialysis sessions, more effective volume management, good patient compliance, mineral and bone disorder control) were more common in Japan than in other countries [11]. However, in European and North American countries, kidney transplantation was more common among ESKD patients than Asian countries, such as China and Japan. In other words, many healthier and younger patients have opted for kidney transplantation instead of maintenance dialysis treatment in those countries. Therefore, transplantation may deplete the dialysis pool of healthier patients in North America and Europe, but not in China and Japan [11]. Although this may have impact on mortality rate among countries, mortality rate differences among countries may still exist after this factor adjusted.
Some widely recognized risk factors found in our results were consistent with previous researches, such as aging, low Alb, comorbidities. Not surprisingly, aging was known to be a signi cant risk factor for death. And the guideline of Kidney Disease Outcomes Quality Initiative (K/DOQI) suggested that the optimal serum Alb level of chronic kidney disease (CKD) patients was equal to or greater than the lower limit of the normal range (approximately 4.0 g/dL) [12]. We had 42.0% of our patients with serum Alb ≥ 4g/l, and 83.6% patients ≥ 3.5g/l. Patients with Alb ≥ 4g/l also have survival bene ts, compared with patients with Alb in 3.5-3.9 g/l. Serum Alb was regarded as the re ection of visceral protein stores, so it can act as a marker of protein malnutrition. Malnutrition was a strong contributor for increasing mortality in HD patients [13,14]. Deaths due to cardiovascular and infectious causes were increased with malnutrition and hypoalbuminemia [15,16]. Therefore, for most patients, they should try to make serum Alb meet the target of 4 g/l, and for patients with Alb < 3.5g/l, special attention should be paid to improving their nutritional status. Comparing the clinical data of the 2 groups of patients, we found that patients who died combined with more comorbidities than those who survived. In the COX model, patients with a history of CHF had higher mortality risk. As more than half of our patients died from cardiovascular causes, patients with a history of cardiovascular disease have a higher risk of death than other patients.
In our results, patients with low platelet counts (< 100*109/L) had a higher risk of all-cause mortality. Thrombocytopenia may have serious consequences, such as increasing the risk of internal and external bleeding, delaying in wound healing and coagulation defects. Previous studies have reported that thrombocytopenia was associated with an increased risk of all-cause death and cardiovascular death in general populations and patients with lung disease [17][18][19]. In one study about the association about mean platelet volume (MPV) and mortality in incident HD patients, the author also showed that lower baseline platelet counts were associated with higher mortality risk across all multivariable models, which was in line with our ndings [20]. Furthermore, there was a prothrombotic adverse drug reaction called Heparin-induced thrombocytopenia (HIT) [21]. Heparin is a common anticoagulant used during HD period to prevent clotting. Some patients may produce platelet-activating antibodies when they received heparin [22], and show unusual clinical features, such as thrombocytopenia, sometimes accompanied by disseminated intravascular coagulation and microvascular thrombosis. Several studies showed that HD patients with HIT were at a higher risk of cardiovascular mortality and arteriovenous stula thrombosis than patients without HIT [23,24].
And high PP was an independent predictor of all-cause mortality. Compared with PP < 63mmHg, PP ≥ 63mmHg was associated with a 64.8% increase in hazard ratio for death. Regarding blood pressure control, PP was not widely considered as a risk factor in ESKD patients. But several studies reported that PP was associated with all-cause mortality and/or cardiovascular mortality in HD patients [25,26].
Meanwhile, cardiac dysfunction was a widespread problem in ESKD patients, and cardiovascular death was the major cause in those patients. Elevated PP was usually due to increased central aortic stiffness and accompanied by increased pulse-wave velocity. Said et al found that PP was associated with an increased risk of developing cardiovascular disease and cardiovascular mortality in a large communitybased population [27]. Elevated pressure during systole induce left ventricular hypertrophy, then lead to left ventricular failure, whereas lower pressure throughout diastole had the possibility to limit coronary perfusion and lead to ischemia [25]. This may contribute to the high prevalence of cardiovascular disease in ESKD patients [28].
Overall, the mortality rate of our HD patients was comparatively low compared with other countries. Among China and other countries, especially developed countries, there were obvious differences in race, lifestyle, economic status, medical insurance policies and so on. Whether the risk of mortality is affected by these factors and how they affect need further investigation. Therefore, it was important to explore the risk factors related to death in Chinese HD patients and the modi able dialysis patterns.
However, our study had several limitations. Firstly, this is an observational study which might has inherent shortcomings such as selection bias and confounding factors. Secondly, the limited number of deaths in our study makes it impossible for us to perform some more detailed subgroup analysis. Thirdly, the China DOPPS study only included patients from three major cities. We didn't include patients receiving HD treatment in smaller cities or rural areas, and their conditions may be worse than what we reported. Therefore, the results can not represent for the whole country. Even though, the results of this study have irreplaceable signi cance for us to understand the current situation of Chinese HD patients and to explore modi able practice patterns to improve patients' survival.
In conclusion, results in this study were important and complement to previous ndings and added new knowledge to the understanding of mortality risks for prevalent HD patients globally. Our analysis highlighted that there were some modi able risk factors for Chinese patients' survival. Attention should also be paid to patients who were older, with lower Alb, low platelet counts, higher PP, and had a history of CHF.

Study design and subjects
The DOPPS is an international prospective cohort study of in-center HD patients ≥ Demographic and clinical characteristics were abstracted from medical records using a web-based data collection tool. Death events and reasons were recorded during follow-up period. The study was approved by the Ethics Committee of Peking University People's Hospital (ethical approval number: 2018PHB028-01). All patients signed the written informed consent. The authors con rm that all the methods used in this study comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration.

Statistical analysis
Baseline characteristics were reported as continuous or categorical variables. Normally distributed continuous variables were presented as mean and standard deviation (SD), otherwise, expressed as median (25th, 75th ). Crude mortality rates were estimated as the number of deaths per 100 patient-years.
Patients were divided into 2 groups, survived or died. Differences in baseline data between survived and died groups were evaluated by using independent sample t-test or Wilcoxon rank-sum test respectively.
Differences in categorical variables between 2 groups were estimated by using chi-square test.
We examined proposed clinically relevant variables by univariate Cox regression models, and relatively signi cant variables (P < 0.15) were added into multivariate Cox regression analysis. The variables included in the multivariate COX regression were age, vascular access, pulse pressure (PP), serum albumin (Alb), hemoglobin (Hgb), platelet counts, spKt/V, serum corrected calcium, phosphate, comorbidities ((congestive heart failure, CHF) cancer, other cardiovascular disease, cerebrovascular disease, fracture, history cirrhosis of the liver, diabetes, hypertension, hepatitis, lung disease, peripheral vascular disease). PP was de ned as the difference value of the mean pre-dialysis systolic blood pressure and the mean diastolic blood pressure. The PHREG procedure based on the Cox proportional hazards model were used to analysis association of patient characteristics with death. Patients from the same facility tend to have similar practice patterns, so we accounted for facility clustering using a robust sandwich estimator of the covariance. After adjustment, age, serum Alb level, platelet counts, comorbidities (CHF) and PP were signi cant and included in the nal COX model.
We converted continuous variables considered as risk factors of all-cause mortality in nal Cox model into categorical variables to investigate the differences in death risk among subgroups of each variable. Alb was divided into 3 groups by < 3.5g/l, 3.5-3.9g/l, and ≥ 4.0g/l; platelets counts were divided into 3 groups by < 100*10 9 /l, 101-299*10 9 /l, ≥ 300*10 9 /l; and PP were divided into 2 groups by < 63 mmHg and ≥ 63mmHg.
We performed MI procedure to impute the missing data, and continuous and categorical variables were imputed by fully conditional speci cation (FCS) regression and logistic regression, respectively. After 25 steps of imputation, 25 data sets were combined for the nal analysis of Cox regression model.
Percentages of missing for most variables were < 10%, except for Kt/V (37.8%) and transferrin saturation (40.6%). P value < 0.05 was seen as statistically signi cant. Statistical analysis was performed with SAS, version 9.4 (SAS institute, Cary, NC; USA), and the forest plot of subgroup analysis were performed by R software, version 3.2.3.