We conducted a retrospective, observational, multicenter cohort study of 374 adult HD incident patients from our outpatient hemodialysis units (Las Palmas, Spain) between January 1, 2016, and January 1, 2021. The inclusion criteria were as follows: patients over 18 years of age from the Advanced Chronic Kidney Disease consultations, renal transplantation with graft dysfunction or patients with acute renal failure who required chronic hemodialysis. Patients belonging to the peritoneal dialysis or home hemodialysis programs were not included.

Electronic medical records (EMR) (Nefrosoft® and Selene-Drago®) of incident patients on chronic ambulatory hemodialysis were reviewed. Demographic data obtained were age, sex, type of vascular access, diuretics (use and dose), presence of comorbidities (arterial hypertension, diabetes mellitus), modified Charlson index (MCI) (calculated excluding subject's age and presence or absence of kidney disease),10 etiology of CKD and cause and date of death.

Mean dry weight (kg), interdialytic weight gain (kg), height (meter), body mass index (BMI) (kg/m2) and Kt (l) were obtained during one month after starting hemodialysis. Kt was determined by ionic dialysance, total body water (V) (l) by Watson's formula, body surface adjusted Kt (Ktbsa) by Lowrie's formula, and ΔKt was calculated by the difference between Ktbsa and Kt.11,12 The dose of loop diuretics was converted to its torasemide equivalent (8mg torasemide=40mg furosemide).1

The analytical data were obtained before the start of the hemodialysis session (the first value obtained within the first 3 months of starting the technique was taken into consideration) and were: Cl− (mEq/L), Na+ (mEq/L), K+ (mEq/L), C-reactive protein (CRP) (mg/L), bicarbonate (HCO3−) (mEq/L), anion gap (mEq/L) [AG=Na+-(Cl−+HCO3−)], serum iron (μg/dL), ferritin (ng/mL), albumin (g/dL) transferrin saturation index (TSI) (%), transferrin (mg/dl), absolute value of neutrophils (ANC) (103/μL), lymphocytes (ALC) (103/μL) and platelets (PLT) (103/μL), neutrophil–lymphocyte ratio (NLR=ANC/ALC), platelet–lymphocyte ratio (PLR=PLT/ALC) and systemic immunity-inflammation index (SII=ANC×PLT/ALC). Hypochloremia was considered when Cl− <98mEq/L.13

The analyses were performed at the hospital reference laboratory using standardized methods.

Flexia® or Formula® (Bellco®/Medtronic®, Mirandola, Italy) monitors were used in the outpatient hemodialysis units. Monitors were synchronized with Nefrosoft® EMR and session data was instantly downloaded. In all cases, an ultrapure dialysis fluid was used, defined according to the guidelines of the Spanish Society of Nephrology.14 The most frequent prescription was conventional high-flow hemodialysis, 4h, 3 times per week. The membranes used were made of high permeability and biocompatibility polyethersulfone with a surface area ≥1.7m2 (Elisium® 17H, 19H or 21H, Nipro Medical Corporation®, Osaka, Japan). Dialysate concentration of chloride was 108.3mEq/L. Prescribed blood flow (Qb) was the maximum allowed by the vascular access (300–450mL/min).

Mortality was the only outcome of interest. Cardiovascular mortality was defined as death due to heart failure, cardiac arrest, myocardial infarction, cerebrovascular accident, or peripheral vascular disease. Noncardiovascular mortality was defined as all other causes of death. Survival time was calculated from the date of entry into the chronic hemodialysis program until death, kidney transplantation, loss to follow-up or end of the study: January 1, 2021.

The baseline characteristics of the sample were described: frequency and percentage were used for categorical variables and mean and standard deviation (SD) for normal continuous variables and median and quartiles for non-normal continuous variables. Normality was tested using the Shapiro–Wilk test.

Patients were divided into quartiles based on Cl− values [quartile 1 (<100mEq/L), quartile 2 (100–101mEq/L), quartile 3 (102–103mEq/L) and quartile 4 (>103mEq/L)], and a trend test was carried out (using Cuzick test for continuous variables and Cochran–Armitage test for binary variables) to evaluate whether there was an association between Cl− levels and the rest of the variables in a way that was less sensitive to the presence of outliers or possible non-linearity. In addition, the data were represented with Spearman correlations as they were less sensitive to the presence of extreme data.

Also, the sample were classified into two groups based on the median CRP as cut-off (high or low inflammation). The values of the continuous variables were compared using Student's T-test (for normal variables) or Mann–Whitney–Wilcoxon test (for non-normal variables). In the case of categorical variables, the comparison was made using the chi-square test.

Kaplan–Meier survival curves were used for the entire cohort and for groups defined by Cl− quartiles; and median Cl− or CRP, both used as cut-off point. Survival curves were compared between groups using the log-rank test. As a description of survival time, the probability of survival after 1, 2, 3, 4 and 5 years, respectively, was estimated.

Univariate and multivariate Cox proportional hazard regression models were constructed to examine the association between serum Cl− levels at the start of chronic hemodialysis and mortality (cardiovascular and all-cause).

To select variables in the multivariate model, initially, univariate Cox models were adjusted to all study variables. Those that were significant at the level of 0.25 were then used in an initial multivariate model. A stepwise procedure was used to minimize the Akaike Information Criterion (AIC), selecting the best model. The proportional hazards assumption was tested to validate the model.15

In all cases, differences with p-values of less than 0.05 were considered significant. Statistical analysis was performed using R program version 4.1.1 (2021) (R Foundation for Statistical Computing, Vienna, Austria. URL http://www.R-project.org/).

The study protocol (2020-428-1) was approved by the Ethics and Clinical Research Committee of the Dr. Negrín University Hospital of Gran Canaria (Las Palmas de Gran Canaria, Spain). The study was conducted following the guidelines of the Declaration of Helsinki. Data were collected in an anonymized database and given its retrospective study design and not affecting the clinical management of patients, informed consent exemption was granted.

The median age of patients was 68 years. Males accounted for 64.4%. Diabetes as a comorbidity was present in 52.9%. Patients were also grouped according to quartiles of Cl−. Those belonging to the first quartile of Cl− presented higher CRP levels, as well as other parameters associated with inflammation. There were no significant differences in MCI as a function of chloride quartiles (p=0.77) or cause of death (p=0.16) (Table 1) (Table S1).

The median Cl− level was 101mEq/L (q25–q75: 99–103) while the median Na+ level was 139mEq/L (q25–q75: 137–141). Both natremia and chloremia had no significant differences when compared by sex, vascular access, or etiology. There was a strong, positive, and significant correlation between Cl− and Na+ (r=0.64) (p≤0.0001) (Fig. S1).

The correlation between Cl− and diuretic dose was low (r=−0.05) and not significant (p=0.49) but there was an inverse correlation between Cl− and interdialytic weight gain (r=−0.15) (p=0.004) (Fig. S2).

The median CRP of the sample was 5.11mg/L (q25–q75: 2.04–16.05). This was the cut-off point chosen to divide the patients into two groups (“high” inflammation=CRP >5.11mg/L and “low” inflammation=CR≤5.11mg/L). Between the two inflammatory subgroup populations (high vs low), there were no significant differences in age, gender, etiology, presence of diabetes or hypertension as comorbidities, previous kidney transplant, interdialytic weight gain, type of vascular access, MCI, or use of diuretics.

The median value of both Cl− and Na+ were significantly lower in the group with “high” inflammation and there was a significant inverse correlation between Cl− or Na+ levels and Ln(CRP). This association was not affected by interdialytic weight gain, diuretic dose, or their previous transplant condition (Fig. S3) (Table 2).

There were significant differences in the median dose of diuretics, being higher in the most inflamed group. Significant differences were observed in many of the variables classically related to inflammation between subgroups (Table 2).

The median Kt was 48L (q25–q75: 43–54). The target Kt adjusted to body surface (Ktbs) was 50±3.80L and the Kt/V was 1.30±0.29. The mean of the differences (ΔKt) between the Ktbs and the Kt was −2.18±8.81L.

During a median follow-up of 20 months (q25–q75: 10–32), 83 (22.2%) of the 375 patients died. The causes of death were: 47 (56.6%) cardiovascular, 22 (26.5%) infection and 14 (16.9%) malignancies. Although it appears that death due to infectious causes and malignancy occurs earlier (12 months), the log-rank test was not significant (p=0.87) (Fig. S4). There was no significant difference between Cl− and Na+ values according to cause of death (Table S1).

The Kaplan–Meir survival analysis stratified by baseline Cl− levels showed that patients in the 1st quartile had a higher both cardiovascular mortality (log-rang test, p=0.0085) and all-cause mortality (log-rank test, p=0.019). The separation of the 1st quartile curve from the rest is more pronounced from the second year onwards (Fig. 1).

Fig. 1.

(A) Kaplan–Meier estimates of cumulative all-cause survival rates. (B) Kaplan–Meier estimates of cardiovascular mortality. In both cases patients were stratified into four groups according to Cl− quartiles.

(0.33MB).

Patients in the group with high CRP and low Cl− also had higher all-cause mortality compared with other groups (log-rank test, p=0.00044). In addition, of the two curves classified as “low CRP”, the one belonging to the lower chloride separates from the other from month 25 onwards (Fig. 2). There were no significant differences between these groups in terms of MCI (p=0.35).

Fig. 2.

Kaplan–Meier estimates of cumulative all-cause survival rates. Patients were stratified into four groups according to median CRP and Cl−.

(0.34MB).

To identify which variables were significantly associated with survival, univariate Cox proportional hazard regression models were fitted to each one of the variables in the study. Only those variables for which the regression coefficient has p-value <0.25 are shown (Table 3).

All these variables were initially entered in a multivariate model. A stepwise procedure seeking to minimize the AIC criterion was used for selecting variables in the final model. Variables selected were age, LnCRP, Cl− (as continuous variable), Na+, albumin, sex, use of diuretics and MCI. The LnCRP variable did not meet the proportional hazards hypothesis in the model (p=0.002). For this reason, we stratified the Cox model according to this variable; to stratify we considered two groups, the one formed by subjects whose CRP exceeds the median CRP value (CRP High), and those whose CRP is less than or equal to the median value (CRP Low). Model was again adjusted by a stepwise procedure. In the stratified model it is assumed that the effect of each independent variable is the same in the two strata defined by the CRP value. It is possible that this value differs between strata, i.e. that there is an interaction between the level of PCR and any of the independent variables. To check this, we also test for the presence of interactions. As interaction was not significant (p=0.99), the final fitted model was the stratified one, which coefficients are shown in Table 3.

After adjusting for the rest of variables (age, Na+, albumin, being male, taking diuretics and MCI), the coefficient of Cl− was negative (−0.172, with se=0.045), which means that higher baseline Cl− was associated with an increase in the probability of survival (HR 0.83; 95% CI 0.77–0.92; p<0.0001).

Increasing age (HR 1.02; 95% CI, 1.01–1.04; p=0.023), lower albumin (HR 0.51; 95% CI, 0.31–0.84; p=0.0083), not taking diuretics (HR 0.58; 95% CI, 0.37–0.92; p=0.02) and higher MCI (HR 1.12; 95% CI, 1.01–1.23; p=0.033) were also significantly associated with all-cause mortality (Table 3).

Likewise, in a multivariate analysis of cardiovascular mortality, higher baseline Cl− (HR 0.9; 95% CI, 0.82–0.97; p=0.0057) and albumin levels (HR 0.41; 95% CI, 0.22–0.76; p=0.043) were significantly associated with better survival (Table 4).

A similar adjusted Cox model was designed using Cl− quartiles. Compared with the 4th quartile of Cl− (>103mEq/L), the 1st quartile (<100mEq/L) had 2.51-fold (95% CI, 1.25–5.07; p=0.01) and 2.94-fold (95% CI, 1.20–7.18; p=0.018) higher risk of all-cause and cardiovascular mortality, respectively (Tables S2 and S3).