Article Mean Corpuscular Volume and Mortality in Patients with CKD Yao-Peng Hsieh,*†‡§ Chia-Chu Chang,*§ Chew-Teng Kor,* Yu Yang,*§ Yao-Ko Wen,* and Ping-Fang Chiu*§ Abstract Background and objectives Mean corpuscular volume is the measure of the average size of the circulatory erythrocyte, and it is principally used as an index for the differential diagnosis of anemia. Recently, mean *Division of corpuscular volume has been associated with mortality in many clinical settings. However, the association of Nephrology, mean corpuscular volume with mortality in patients with CKD has not been fully addressed. Department of Internal Medicine, Design, setting, participants, & measurements We conducted a retrospective observational cohort study of 1439 Changhua Christian Hospital, Changhua, patients with stages 3–5 CKD and baseline mean corpuscular volume values from 2004 to 2012 in a medical † – – Taiwan; PhD Program center. The study cohort was divided into the high mean corpuscular volume group and the low mean cor- in Translational puscular volume group by the median value (90.8 fl) of mean corpuscular volume. The baseline patient infor- Medicine, College of mation included demographic data, laboratory parameters, medications, and comorbid conditions. The Life Science, National independent association of mean corpuscular volume with mortality was examined using multivariate Cox Chung Hsing University, Taichung, regression analysis. Taiwan; ‡School of Medicine, Kaohsiung Results Of the 1439 participants, 234 patients (16.2%) died during a median follow-up of 1.9 years (interquartile Medical University, – fi – Kaohsiung, Taiwan; range, 1.1 3.8 years). The crude overall mortality rate was signi cantly higher in the high mean corpuscular § volume group (high–mean corpuscular volume group, 22.7%; low–mean corpuscular volume group, 9.7%; and School of P, – Medicine, Chung 0.001). In the fully adjusted models, the high mean corpuscular volume group was associated with higher Shan Medical risks of all-cause mortality (hazard ratio, 2.19; 95% confidence interval, 1.62 to 2.96; P,0.001), cardiovascular University, Taichung, mortality (hazard ratio, 3.57; 95% confidence interval, 1.80 to 7.06; P,0.001), and infection-related mortality Taiwan (hazard ratio, 2.22; 95% confidence interval, 1.41 to 3.49; P=0.001) compared with the low–mean corpuscular volume group. Correspondence: Dr. Yao-Peng Hsieh, – Division of Conclusions In patients with stages 3 5 CKD, mean corpuscular volume was associated with all-cause mortality, Nephrology, Internal cardiovascular disease mortality, and infection-associated mortality, independent of other factors. The under- Medicine, Changhua lying pathophysiologic mechanisms warrant additional investigation. Christian Hospital, 135 Nanxiao Street, Clin J Am Soc Nephrol 12: 237–244, 2017. doi: 10.2215/CJN.00970116 Changhua City, 500 Taiwan, Republic of China. Email: [email protected] Introduction erythrocytes (5,10,11). In addition, because reticulo- Anemia is a common comorbid condition in patients cytosis has been correlated with high mean corpuscular with CKD, especially for those with a moderate to volume (MCV), the use of erythropoiesis-stimulating severe stage (1). The underlying etiologies of anemia agents (ESAs) may be associated with macrocytosis, are multifactorial, and impaired endogenous erythro- whichissupportedbythestudybyTennankoreet al. poietin production by the dysfunctional kidneys is (12) that concluded that a higher ESA dosage was the predominant cause (2). Other reasons that have linked with macrocytosis in patients on chronic been identified include shortened erythrocyte lifespan, hemodialysis. iron deficiency, folate and vitamin B12 deficiency, and MCV is a measure of the mean size of erythrocytes bone marrow suppression by uremic toxins (3–6). Re- and has long been a useful index for approaching the cently, an increasing body of evidence has unveiled the differential diagnosis of anemia, and it is also a possible effect of inflammation on renal anemia (7). biomarker of bone marrow dysfunction. Recently, Although the typical renal–related anemia is nor- however, MCV has emerged as an independent risk mocytic and normochromic (8), a fair proportion of factor for mortality in several diseases. For example, patients with ESRD have macrocytosis (9). Several Zheng et al. (13) reported that preoperative MCV was pathophysiologic mechanisms have been suggested associated with mortality in patients with resectable to explain the observed phenomenon in patients on esophageal cell carcinoma. Unfortunately, little infor- dialysis, including intravenous iron supplementa- mation on the associations of MCV with clinical out- tion, megaloblastic anemia due to folate or vitamin comes among patients with CKD is currently available B12 deficiency, and dialysis-related changes in in the literature. For this reason, we conducted this www.cjasn.org Vol 12 February, 2017 Copyright © 2017 by the American Society of Nephrology 237 238 Clinical Journal of the American Society of Nephrology study to examine the association between MCV and mortal- median and interquartile range. Comparisons of distribu- ity among patients with stages 3–5 CKD. tion between the two groups were assessed using the paired t test for parametric data and the Mann–Whitney U test for nonparametric data as appropriate. Categorical Materials and Methods variables are shown as number (n) and percentage, and the Participants and Measurements difference of the two groups was compared using chi- – We conducted a single center retrospective study from the squared test or Fisher exact test as appropriate. The com- medical records and electronic data in Taiwan. Between Jan- parison of survival status between the two groups was uary 1, 2004 and December 31, 2011, patients who joined the done using the Kaplan–Meier curve with log rank test to integrated CKD care program at the outpatient clinic were determine significance levels. Cox proportional hazards fi screened for eligibility. The diagnosis of CKD was de ned model were used for the analyses of predictors for mortal- according to National Kidney Foundation Kidney Disease ity. We implemented five models for the adjustments of Outcomes Quality Initiative criteria. Renal function status the covariates: model 1, adjusted for sex, age categories, fi – was determined by eGFR using the simpli ed four variable educational level, marital status, and BMI; model 2, ad- fi Modi cation of Diet in Renal Disease (MDRD) Study equa- justed for all variables in model 1 plus alcohol and smok- tion as follows: eGFR in milliliters per minute per ing; model 3, adjusted for all variables in model 2 plus 2 3 21.1543 20.2033 1.73 m =186 serum creatinine age 0.742 (if laboratory parameters; model 4, adjusted for all variables 3 the patient is a woman) 1.212 (if black patient). After ex- in model 3 plus medications; and model 5, adjusted for all cluding those with stages 1 and 2 CKD, younger than 20 years variables in model 4 plus comorbid conditions. fi of age, or older than 80 years of age, a nalcohortof1439 One sensitivity analysis was done with the hazard ratio fi patients was left for nal analysis. All of the study partici- (HR) of MCV adjusted for quintiles of the propensity score pants were followed until death or the end of study on De- in addition to all of the covariates in model 5 (model S1). cember 31, 2012. The study was approved by the institutional The propensity scores were estimated by using the logistic review board of Changhua Christian Hospital, and all of the regression model to control for the differences between the clinical investigation was conducted according to the princi- high-MCV ($90.8 fl)andlow-MCV( ,90.8 fl) groups; ples of the Declaration of Helsinki. ,10% of our patients had serum levels of folate and vitamin For each patient, the collected baseline data at enrollment B12. We performed another sensitivity test with adjustment included demographic features (sex and age), smoking, for iron profile in 417 patients who had blood levels of iron alcohol status, comorbid conditions, the cause of CKD, profile (model S2). All statistics were two-sided tests, and educational status, body mass index (BMI), medications, the results were considered significant if the P value was and laboratory parameters. The comorbidities encompassed ,0.05. All statistical analyses were carried out using the diabetes mellitus (DM), hypertension, coronary artery disease, statistical package for Windows, SAS 9.2 (SAS Institute congestive heart failure, cerebrovascular disease and periph- Inc., Cary, NC), and SPSS 16.0 (SPSS, Chicago, IL). eral artery disease, cancer, dementia, autoimmune disease, chronic lung disease, and liver cirrhosis. The medication history included angiotensin–converting enzyme inhibitors, Results angiotensin II receptor blockers, iron, folic acid, vitamin Baseline Characteristics of the Study Cohort B12, lipid-lowering agents (statin and fibrate), and ESAs. The median MCV level among the 1439 participants was The laboratory measurements included blood levels of 90.8 fl, with a range of 58.4–114.5 fl. The mean age was BUN, creatinine, albumin, white blood cell counts, hemoglo- 64.1612.2 years old, 795 patients (55.3%) were men, and bin, red cell distribution width, MCV, platelets, cholesterol, the median follow-up was 1.9 years (interquartile range, triglycerides, glutamic-pyruvic transaminase, uric acid, cal- 1.1–3.8 years). The three most common causes of CKD in- cium, phosphate, and 24-hour proteinuria. The median value cluded DM, hypertension, and chronic GN. Primary school of MCV for our study population was 90.8 fl.Thestudyco- was the most common educational level, and 20.4% of the hort was stratified into two groups according to this median participants were current smokers. Additional information value: the high-MCV group ($90.8 fl) and the low-MCV was collected according to the median (90.8 fl) of the MCV group (,90.8 fl). Regarding the causes of death, the two values (dividing the cohort into low- and high-MCV groups) most common reasons were infection and cardiovascular dis- and is presented in Table 1.