Study design and population
A retrospective cohort study was performed, including 3073 incident CAPD patients from five peritoneal dialysis centers of three provinces in China between January 1, 2005, and December 31, 2018. To increase the generalizability of the findings in the CAPD settings, we only excluded those with age < 18 years or less than 3 months of follow-up. The Human Ethics Committee approved each research facility’s study, consistent with the Declaration of Helsinki’s ethical principles. Written informed consent was obtained from all eligible patients.
Data collection and definitions
Two trained nurses in each dialysis center recorded demographic data, comorbidities, medications, and laboratory parameters at baseline. The baseline was defined as one month before the first CAPD. If the patient has more than one measurement in this one month, the measurement closest to the first CAPD will be included. If parameters were missed at baseline, parameters closest to before the first CAPD were included. Medications included calcium channel blockers, beta-blockers, angiotensin II receptor blockers/angiotensin-converting enzyme inhibitors (ACEI/ARBs), diuretics, statins, and aspirin. Baseline laboratory parameters included hemoglobin, serum albumin, serum uric acid, estimated glomerular filtration rate (eGFR), cholesterol, triglyceride, high-density lipoprotein, and low-density lipoprotein, and high-sensitivity C-reactive protein [hs-CRP]). All laboratory parameters from fasting blood samples were measured in the department of the laboratory of each tertiary hospital.
Hyperlipidemia was defined as serum cholesterol levels ≥ 240 mg/dL, triglycerides levels ≥ 200 mg/dL, low-density lipoprotein levels ≥ 160 mg/dL, or when the patients were receiving lipid-lowering drugs . Patients with a history of CVD receiving lipid-lowering medications to prevent recurrence of CVD episodes were not considered hyperlipidemia. Diabetes mellitus was defined as ⑴ HbA1c ≥ 6.5%, ⑵ fasting plasma glucose ≥ 126 mg/dL, ⑶ 2-hour plasma glucose ≥ 200 mg/dL during an OGTT, ⑷ in a patient with classic symptoms of hyperglycemia or hyperglycemic crisis, a random plasma glucose ≥ 200 mg/dL, ⑸ or when the patients were receiving glucose-lowering drugs. In the absence of unequivocal hyperglycemia, criteria ⑴ to ⑶ should be confirmed by repeat testing. Hypertension was defined as systolic blood pressure (BP) > 140 mmHg or diastolic BP > 90 mmHg or taking antihypertensive medications. CVD was defined as coronary heart disease, congestive heart failure, arrhythmias, cerebrovascular disease, or peripheral vascular disease. Current smoking was defined as at least one cigarette a day, and current alcohol consumption was defined as > 20 g of ethanol a day. The comorbidity scores were calculated according to the Charlson comorbidity index calculator, which categorizes patients’ comorbidities. The more points are given, the more likely the predicted adverse outcomes are. eGFR was calculated using the Chronic Kidney Disease Epidemiology Collaboration equation..
Outcomes and follow-Up
The primary outcome was all-cause mortality. Two experts identified the exact death cause based on death certificates and medical records. We determined death causes and dates based on medical files of admission. If patients died out of hospitals, we determined death causes according to interviewing family members by telephone to acknowledge death’s circumstances, combined with information from medical records of peritoneal dialysis centers.
According to the KDIGO guideline, hyperlipidemia patients in our study not receiving lipid-lowering therapy did not receive a new lipid-lowering treatment when starting CAPD. Patients needed to return quarterly to each facility for an overall medical assessment for clinical purposes. The trained nurses conducted monthly face-to-face interviews or telephone interviews to assess the patient’s general conditions related to medications. If patients had significant serum lipid disorders during the follow-up period, we conducted an integrated approach rather than statins to improve serum lipid disorders. All patients were followed up until CAPD cessation, death, the end of the 8-year duration, or June 30, 2019. Transferring to hemodialysis, receiving renal transplantation, moving to other centers, loss of follow-up, or still survival with a follow-up period of 8 years or as of June 30, 2019, were considered censored.
We summarized baseline characteristics using descriptive statistics and expressed continuous variables as the mean (standard deviations) or median (interquartile range, [IQR]) and categorical variables as frequency (percentage). Multivariable binary logistic regression was conducted to estimate the association between baseline variables and hyperlipidemia. Covariables in multivariable binary logistic regression included age, sex, body mass index, systolic BP, diastolic BP, 24-hour urine volume, current smoking, current alcohol consumption, Charlson comorbidity index, diabetes mellitus, hypertension, and a history of CVD, hemoglobin, serum albumin, serum uric acid, eGFR, and hs-CRP.
Kaplan-Meier curve was conducted to examine the survival probability over the overall observational period. We constructed four Cox proportional hazards regression models to analyze the association between hyperlipidemia and all-cause mortality. Model 1, unadjusted; model 2, model 1 plus age, sex, body mass index, systolic BP, diastolic BP, current smoking, current alcohol consumption, 24-hour urine volume, Charlson comorbidity index, diabetes mellitus, hypertension, and a history of CVD; model 3, model 2 plus calcium channel blockers, beta-blockers, ACEI/ARBs, diuretics, statins, and aspirin; model 4, model 3 plus hemoglobin, serum albumin, serum uric acid, eGFR, cholesterol, triglyceride, high-density lipoprotein, low-density lipoprotein, and hs-CRP. Subgroups were stratified by age (< 65 or ≥ 65 years old), sex (men or women), diabetes mellitus (yes or no), hypertension (yes or no), and a history of CVD (yes or no). The interactions between subgroups and hyperlipidemia were examined by conducting a formal interaction test.
Many participants were censored due to hemodialysis transfer, kidney transplantations, moving to other centers, or loss of follow-up. Therefore, multiple subdistribution hazards models considering censoring events as competing risks were conducted as a sensitivity analysis. Additionally, the cumulative incidence was depicted using Gray’s test. We conducted a sensitivity analysis by excluding the patients already on lipid-lowering therapies from the whole cohort. The Cox proportional and subdistribution hazardresults were presented as the hazard ratio (HR) and the 95% confidence interval (CI). A two-sided P value < 0.05 was considered statistically significant. Statistical analyses were performed using Stata 15.1. statistical software (StataCorp, College Station, TX).