Our present research is a prospective cohort study in consecutive patients with ESRD who were undergoing maintenance dialysis at Xinqiao Hospital, Army Medical University (Third Military Medical University) from March 1, 2016, to July 1, 2019. A total of 129 consecutive patients were included in our cohort according to the inclusion and exclusion criteria. Most of the subjects were followed up within 39 months (fewer of them were followed up within 2 months). Median and quartiles of the follow-up time were 39 (39–39) months. During a median of 39 months of follow-up, nine patients were excluded or lost to follow-up.

The inclusion criterion was ESRD patients who were undergoing maintenance dialysis. The exclusion criteria were as follows: severe hepatic dysfunction (n = 2), death during hospitalization (n = 2), advanced cancer (n = 1), or severe infection (1). Additionally, three patients were lost to follow-up in the cohort, as shown in Figure 1.

All patients provided written informed consent. Our present research complied with the Declaration of Helsinki with respect to human investigations and was approved by the ethics committee of Xinqiao Hospital, Army Medical University (Third Military Medical University).

Venous blood samples were obtained from the patients in the early morning after at least 12-h fasting at both cross sections (after entrance to the cohort and follow-up cross section) within a median of 39 months of follow-up (from 2 to 39 months).

First, we performed routine blood examination [tests of red blood cell count (RBC), mean corpuscular volume (MCV), concentration of hemoglobin (Hb), mean corpuscular hemoglobin (MCH), hematocrit (HCT), MCH concentration (MCHC), red blood cell distribution width (RDW), white blood cell count (WBC), platelet count (PLT), plateletcrit (PCT), and platelet distribution width (PDW)] by using an automated hematology corpuscle analyzer (AU400; Olympus Optical, Co., Tokyo, Japan).

Second, we measured plasma creatinine (Cr, enzyme method), UA (colorimetry), parathyroid hormone (PTH; chemiluminescent immunoassay), and ferritin (chemiluminescent immunoassay) concentrations by using Roche Diagnostics GmbH products (Abbott, i2000, USA).

Third, we also performed examinations of serum iron concentration (FERENE methods, Beckman AU5821), blood urea nitrogen concentration (BUN), potassium concentration (K⁺), and sodium concentration (Na⁺) by using indirect ion-selective electrode methods (EX-Z, JOKOH, Japan). Serum calcium concentration (Ca²⁺) was measured by using Tri-azo methods, and phosphate concentration (P) was measured with a phosphomolybdate ultraviolet method (Roche Diagnostics GmbH, USA).

Finally, EPO, anti-EPO antibody, and anti-EPOR antibody were measured using ELISA kits (Recombinant Human Erythropoietin, BioLegend and Recombinant Human R Erythropoietin, RD Systems, USA). Each sample was assayed in duplicate to measure the exact concentrations of endogenous EPO, anti-EPO antibody, and anti-EPOR antibody.

All of the biochemical variables were measured from blood specimens in the Clinical Laboratory Department, Xinqiao Hospital.

KT/V was calculated as ln(posturea/preurea) – 0.008 ^*ultrafiltration time + (4 – 3.5 ^*posturea/preurea) ^*(ultrafiltration volume/postweight). Follow-up EPO demand was assessed by using the endogenous EPO level divided by Hb (EDI). Anti-EPO antibody positive (EPOA+) was defined as an optical density (OD) >0.5526 [the mean ± 3 standard deviations (SDs) in 55 age-matched healthy volunteers]; otherwise, it was defined as anti-EPO antibody negative (EPOA–). Similarly, EPOR antibody positive (EPOA+) was defined as an OD >0.5356 (the mean ± 3 SDs in 55 age-matched healthy volunteers); otherwise, it was defined as anti-EPOR negative (EPORA–).

Continuous variables (such as age and BMI) that are normally distributed were expressed as the means ± SD. The differences between two groups or among four EDI intervals of these variables were compared using independent Student's t-test or one-way ANOVA [comparisons between two intervals were made using least significant difference (LSD) methods]. The Wilcoxon rank sum test was used to compare the non-normally distributed variables (comparisons among four intervals were performed by using one-way ANOVA after transformation to normality). Furthermore, the chi-square test or Fisher's exact test (if there were less than five cases) was employed to assess dichotomous variables. Univariate logistic regression analyses were performed to primarily screen the risk factors for EPO demand. Variables with p < 0.05 in comparisons among different EDI groups or in the univariate logistic regression were included in multivariate (adjusted) logistic regression analyses to identify independent associated factors of higher EPO demand. The regressions were performed using ordinal logistic regression analyses, and we have just listed the Q4 EDI vs. other intervals (Q1–Q3) owing to too many data. Furthermore, the generalized linear model's analyses have also been performed to determine the OR and its 95% CI. All of the statistical analyses were performed with SPSS 22.0 for MAC statistical software (California, USA).

The EDI was divided into four quartile intervals, and we compared the differences among various groups. The patients in the Q4 and Q3 groups were significantly older than those in the Q2 and Q1 groups (p-values were <0.05; Table 2).

In the routine blood tests, only Hb showed significant differences among different EDI intervals. The Hb concentration was significantly higher in the Q1 EDI group at 10.95 (9.82–12.02). Other routine blood test parameters, including HCT, MCV, MCH, MCHC, RDW-CV, WBC, PCT, PLT, mean platelet volume (MPV), and PDW, showed no differences among the various intervals of EDI.

The serum iron, ferritin, and PTH levels were similar in the groups (all p-values were >0.05).

In the dialysis parameters, the dialysis duration showed no differences among the four intervals (p = 0.179). Predialysis, the potassium ion concentration was significantly higher in the Q4 population than in the other three populations (all p-values were <0.05). However, other predialysis parameters, including the concentrations of urea, Cr, uric acid, calcium ion, serum phosphate, and serum sodium, as well as TCO₂, showed no differences among these populations (all of the p-values were >0.05). Furthermore, the postdialysis-related variables were all similar among various populations, including concentrations of urea, Cr, uric acid, calcium ion, serum phosphate, potassium ion, and serum sodium, as well as TCO₂.

Unfortunately, we found no difference in either ultrafiltration volume or KT/V value (p > 0.05). Focusing on the target level of anti-EPO and anti-EPOR antibodies, only anti-EPO antibody showed significant differences among the four intervals (p = 0.006); that is, the levels of anti-EPO antibody were significantly higher in the Q2, Q3, and Q4 populations than in the Q1 subgroup (all p-values were <0.05). However, the anti-EPO antibody levels were similar among the three subgroups (Q2, Q3, and Q4, Table 2).

We further analyzed the differences in baseline parameters between EPO/EPOR antibody-positive and EPO/EPOR antibody-negative groups. First, age and BMI showed no difference between the EPOA+ and EPOA– groups or between the EPORA+ and EPORA– groups.

In the routine blood examination parameters, we found that only RDW was significantly higher in the EPOA+ group than in the EPOA– group (14.87 ± 1.57% vs. 14.28 ± 1.31%, p = 0.046). Only one variable (MPV) was significantly higher in the EPORA+ group than in the EPORA– group (9.65 ± 1.70 vs. 8.94 ± 1.14 fl, p = 0.010). However, none of the other routine blood examination parameters showed any differences between the EPORA+ and EPORA– individuals (p > 0.05).

Moreover, neither clinical data (serum iron, ferritin, and PTH) nor dialysis-related parameters showed any differences between the EPOA+ and EPORA+ populations.

Regarding the EPORA groups, we found that ferritin and TCO₂ (both predialysis and postdialysis) showed differences between the EPORA+ and EPORA– populations. Other clinical data and dialysis-related parameters were similar in the two groups (Table 3).

To identify the associations between EDI and anti-EPO/anti-EPOR antibody levels, we further compared the EDI between the EPOA + and EPOA– groups (as well as between the EPORA+ and EPORA– groups). As shown in Figure 2, the baseline EPO antibody level is correlated with follow-up EDI (r² = 0.0377, p = 0.03), whereas the baseline anti-EPOR antibody level was not correlated with EDI (r² = 0.0009, p = 0.73). Furthermore, in Figure 3, we found that the follow-up EDI was significantly higher in the EPOA+ group than in the EPOA– group (p = 0.02, Figure 3A). However, there were no significant differences between the EPORA+ and EPORA– groups. In addition, follow-up EDI was similar in the EPOA+ combined with EPORA+ group and EPOA– combined EPORA– group (Figures 3B–D). The correlations between the follow-up EDI and other parameters are shown in Table S1.

To identify the predictive roles of EPO and other potential factors, we performed univariate logistic regression for each parameter between the higher and lower erythropoietin demand (Q4 of EDI vs. Q1 of EDI). Other analyses between Q3 EDI and Q1 EDI are shown in the Appendix.

In the univariate logistic regression analyses, we found that age (OR = 1.071, 95% CI 95% CI: 1.021–1.123, p = 0.005), ferritin (OR = 1.001, 95% CI: 1.000–1.003, p = 0.038), potassium ion concentration before dialysis (OR = 2.781, 95% CI: 1.255–6.162, p = 0.012), dialysis duration (OR = 1.025, 95% CI: 1.002–1.048, p = 0.030), and EPOA OD value (OR = 7.694, 95% CI: 2.109–67.277, p = 0.004) were potential associated factors for higher erythropoietin demand (Table 4).

To identify the independent associated factors of higher erythropoietin demand, variables that showed differences among various intervals of EDI or that had a p-value of <0.05 in univariate logistic regression analyses were included in the adjusted logistic regression. Ultimately, age (OR = 1.072, 95% CI: 1.008–1.140, p = 0.026), potassium ion concentration before dialysis (OR = 3.425, 95% CI: 1.297–9.040, p = 0.013), and EPOA OD value (OR = 5.27, 95% CI: 2.577–6.733, p = 0.007) were found to be independent associated factors of higher erythropoietin demand (Table 5).

The ordinal logistic regression analyses have found that the baseline EPO antibody level (OR = 5.328, 95% CI: 1.072–8.735, p = 0.045), ferritin (OR = 6.746, 95% CI: 1.028–9.976, p = 0.044), and PTH (OR= 0.064, 95% CI: 0.009–0.148, p = 0.035) were independently associated with higher EDI (Tables S4, S5).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.