Propensity-Matched Comparison of New Onset Diabetes Mellitus between Incident Peritoneal Dialysis and Hemodialysis Patients
Chiu-Ching Huang
1,2, Che-Yi Chou
1,2, Kuang-Fu Cheng
3, Chih-Chia Liang
1, Huey-Liang Kuo
1,2, Chiz-Tzung Chang
1,2, Jiung-Hsiun Liu
1,2, Hsin-Hung Lin
1,2, I.-Kuan Wang
1,2, Ya-Fei Yang
1,2, Yi-Tzone Shiao
31
Kidney Institute and Division of Nephrology, Department of Internal Medicine, China Medical University Hospital, Taichung 40447, Taiwan
2College of Medicine, China Medical University, Taichung 40402, Taiwan
3
Biostatistics Center, China Medical University, Taichung 40402, Taiwan
Background
New-onset diabetes mellitus (NODM) is associated with poor outcome in patients treated with hemodialysis (HD) or peritoneal dialysis (PD). PD patients were exposed to extra glucose load from daily dianeal dwelling. We ask the question "Are PD patients more prone to develop NODM after dialysis?". We used a propensity score(PS) matching approach to study this subject. We aim to compare the incidence and outcome of NODM by intention-to-treat analysis in a propensity score-matched HD and PD cohorts.
Methods
Figure 1. Scheme of study design
NODM(-) n=10187 NODM(+) n=2553 P Age (year) 48.4 ±14.2 56.6 ±13.7 <0.001 Follow-up (year) 6.1 ±2.9 4.9 ±2.8 <0.001
Table 1. Clinical characteristics of patients with and without new onset diabetes mellitus (NODM)
HR 95% C.I. for HR p HD (v.s. PD) 1.85 1.57 2.20 <0.001
Age 1.01 1.007 1.014 <0.001
Table 2. Hazard ratio of risk factor for new-onset diabetes in chronic kidney disease patients receiving peritoneal dialysis (PD) or hemodialysis (HD) in univariate competing-risks analysis
All HD and PD patients in Taiwan Renal Registry Database (include all patients who survived ≥ 1M after dialysis) were analyzed from 1997 to 2005. Patients were followed until death, renal transplantation or up to December 31, 2008. We attempted to correct for differences between patient characteristics using a PS-matching method. We matched each PD patient with 4 HD patients having a similar PS. PS accounted for factors recorded in the database, including age, sex, body weight, primary renal disease and co-morbidity(Fig.1) . The risk for NODM was determined using competing-risks analysis. We compared the incidence and survival outcomes of NODM in PD and PS-matched HD patients. Predictors of NODM and mortality were studied using Cox models. All statistical analysis was performed with Stata version 12 SE (Statacorp, Texas, USA). P < 0.05 was considered as significant.
Results
Male gender n(%) 3707 (36) 985 (39) 0.04 HD n(%) 7971 (78) 2221 (87) <0.001 Mortality n(%) 2822 (28) 1276 (50) <0.001 Weight (kg) 69.4 ±5.8 70.2 ±7.8 0.08 Underlying disease n(%) CGN 5898 (58) 1470 (58) 0.77 Hypertension 1013 (10) 298 (12) 0.01 Others 3276 (32) 785 (31) 0.17 Age 1.01 1.007 1.014 <0.001 Male gender 0.936 0.835 1.05 0.266 HTN 0.958 0.855 1.07 0.465 FBG 1.006 1.005 1.007 <0.001 Serum albumin 1.25 1.11 1.41 <0.001 Phosphorus 1.05 1.001 1.098 0.022 CPP 0.980 0.968 1.093 0.232 Hematocrit 1.02 1.001 1.03 0.031Results
We analyzed 2548 non-DM incident PD patients and 10192 PS-matched HD patients with 5.9±3 years follow up. Mean age of PD patients were 50.2± 14.7y vs. 50.2± 14.4y of PS-matched HD patients. During follow up , 13% of PD patients and 22% of PS-matched HD patients developed NODM (Table 1). The crude incidence of NODM was 2.4/100 patient-years for PD patients and 4/100 patient- years for HD patients (p<0.001). The risk of developing NODM after starting dialysis is approximately 50% lower (HR 0.48) for patients treated with PD. Male gender, a lower serum albumin and lower hematocrit were independently linked to increased NODM risks. After adjusting for covariates, regardless of dialysis modality, patients who developed NODM had increased overall mortality (HR 3.628, p<0.001) than those did not.
Conclusions
Figure 2. Risk of developing new onset diabetes mellitus in all hemodialysis (HD) patients (A) and( ) ( ) Co-morbidity n(%) Hypertension 3878 (38) 870 (34) <0.001 CHF 455 (4) 134 (5) 0.09 Ischemic heart 410 (4) 121 (5) 0.11 CVA 193 (2) 63 (2) 0.07 Liver disease 297 (3) 85 (3) 0.27 Cancer 158 (2) 42 (2) 0.73 i-PTH 1.000 0.999 1.000 0.112
FBG: fasting blood glucose, CPP: calcium-phosphate product, i-PTH: intact parathyroid hormone
Table 3. Hazard ratio of risk factor for new-onset diabetes in chronic kidney disease patients receiving peritoneal dialysis (PD) or hemodialysis (HD) in multivariate competing-risks analysis
In a PS-matched Chinese cohort, incident patients receiving HD carry a higher risk of developing NODM than those receiving PD. The patients with NODM are associated with increased overall mortality reardless of dialysis modality chosen.
y ( ) p ( )
matched HD patients (B) compared to PD patients Tuberculosis 84 (1) 22 (1) 0.85
Others 658 (6) 146 (6) 0.17 Hematocrit (%) 29.4 ±3.6 28.9 ±3.8 <0.001 Albumin (g/dl) 3.9 ±0.4 3.8 ±0.5 <0.001 Phosphate (mg/dl) 5.1 ±1.3 4.9 ±1.5 <0.001 Calcium (mg/dl) 9.6 ±0.8 9.6 ±0.8 0.14 CPP (mg/dl)2 48.9 ±13.1 46.7 ±14.7 <0.001 FBG (mg/dl) 96 ±14 168 ±69 <0 001 HR 95% C.I. for HR p HD (v.s. PD) 1.93 1.46 2.56 <0.001 Age 1.01 1.004 1.01 <0.001 FBG 1.01 1.005 1.01 <0.001 Serum albumin 1.36 1.14 1.62 0.001 Phosphorus 1.07 1.02 1.13 0.007 Hematocrit 1.02 0.995 1.05 0.108 analysis
Discussion
Why more NODM in HD patients?
We speculate that PD patients may have better physical activities in their daily life than HD patients. Increased physical activities may prevent the development of diabetes. In addition, the blood-membrane interaction during HD may induce cytokine production, such as C-reactive protein and interleukins-6. Chronic inflammation induced by blood-membrane interaction may play a critical role in the development of diabetes in HD patients.
FBG (mg/dl) 96 ±14 168 ±69 <0.001
i-PTH (ng/dl) 262.6 ±265.6 223.7 ±262.3 <0.001*
HD: hemodialysis, CGN: chronic glomerulonephritis, HTN: hypertension, CHF: congestive heart failure, CVA: cerebral vascular accident, FBG: fasting blood glucose, CPP: calcium-phosphate product, i-PTH: intact parathyroid hormone
* Mann-Whitney U test
FBG: fasting blood glucose
References
1. Tien KJ, Lin ZZ, Chio CC et al. Epidemiology and Mortality of New-Onset Diabetes Mellitus After Dialysis: Taiwan national cohort study. Diabetes
Care2013.
2. Verduijn M, Grootendorst DC, Dekker FW et al. The analysis of competing events like cause-specific mortality--beware of the Kaplan-Meier method.
Nephrol Dial Transplant2011; 26: 56-61.
3. Haubitz M, Brunkhorst R, Wrenger E et al. Chronic induction of C-reactive protein by hemodialysis, but not by peritoneal dialysis therapy. Perit Dial
Int1996; 16: 158-62.
