Comparative short-term safety of bolus versus maintenance iron dosing in hemodialysis patients: a replication study

Intravenous iron is used in combination with erythropoiesis-stimulating agents (ESAs) to treat the anemia of hemodialysis patients. Recent studies have reported a beneficial effect of iron dosing on anemia parameters in hemodialysis patients [1‐4]. While the safety of ESAs in chronic kidney disease has been examined in clinical trials [5‐7], less is known about the safety of intravenous iron. Several biological mechanisms suggest that sub-optimal use of iron could lead to adverse clinical events [8]. Frequent administration of iron may lead to oversaturation of transferrin and the release of free, catalytically active iron into the plasma [9]. Because iron is essential for bacterial growth, free iron in circulation may increase the risk of infection [10, 11]. Indeed, frequency of iron administration has been found to be associated with increased risk of infection-related mortality in ESRD [12]. Free iron is also known to catalyze the formation of highly reactive oxygen species [13, 14]. These could give rise to lipid radicals, which may damage tissue [15] and lead to atherogenesis [16], possibly increasing the risk of cardiovascular events [8, 17].

A recently completed a large-scale observational study examined the short-term comparative safety of intravenous iron dosing strategies in hemodialysis patients and found that patients receiving bolus (frequent large doses to achieve iron repletion) versus maintenance dosing had higher risks of infection-related hospitalization and infection-related death, and that these risks were greatest among the subgroup of patients with a dialysis catheter [18]. No association was found between large-dose intravenous iron treatment strategies and cardiovascular morbidity and mortality [19]. The study was conducted using data from a large U.S. dialysis chain merged with data from the United States Renal Data System (USRDS). Because the previous study was the first large epidemiologic study to address risks of adverse events associated with intravenous iron use, we sought to determine whether these findings could be replicated using the same methods with data from another U.S. dialysis chain where the patient demographics, case mix, and other aspects of medical practice may differ. Specifically, we examined short-term infection and cardiovascular risk associated with bolus versus maintenance intravenous iron dosing in a cohort of patients undergoing chronic hemodialysis.

6,605 patients met study entry requirements and contributed data on 48,050 exposure/follow-up periods (Figure 1). Sample characteristics are presented in Table 1, stratified by iron exposure group. Relative to the earlier study [18], the sample for this study had a greater proportion of Blacks and residents of the Northeast. Patients in this study were also on dialysis longer, had less catheter use, a higher proportion of comorbidities, and higher rates of infection at baseline. Demographic and clinical characteristics were generally similar among the dosing groups. Catheter use was highest in the bolus dosing group and lowest in the no iron group. TSAT and ferritin values were lowest in the bolus group and highest in the no iron group, as might be expected. The prevalence of comorbidities was generally highest in the bolus dosing group and lowest in the no iron group.

Table 2 presents the adjusted hazard ratios and rate differences for the comparison of bolus versus maintenance dosing for the entire sample. These findings are graphically represented in Figure 2. All of the adjusted hazard ratios were >1.00 for the infection-related outcomes, though the 95% confidence intervals included 1.00 for three of the six hazard ratios. The effects were greatest for hospitalized for infection of any major organ system (HR: 1.13) and use of intravenous antibiotics (HR: 1.08) and indicate that bolus dosing increases the risk of these events. On the additive scale, individuals who received bolus dosing had 59 more hospitalizations for major organ system infection, per 1000 person-years, than individuals who received maintenance dosing. Use of intravenous antibiotics was also higher among individuals who received bolus versus maintenance dosing (72 more events per 1000 person-years).

The adjusted hazard ratios for all-cause death and cardiovascular outcomes were relatively imprecise, with confidence intervals including the null. Relative to the rate differences for the infection outcomes, the rate differences for all-cause death and the cardiovascular outcomes were much closer to the null, though there was a slight suggestion of a protective effect on cardiovascular death (Table 2 and Figure 2). Our results were robust to variations in the length of the exposure and follow-up periods (Additional file 1: Table S3) and the addition of covariates beyond our a priori specified multivariable models (Additional file 1: Table S4).

The adjusted hazard ratios and rate differences for the infection-related outcomes in the catheter subgroup are presented in Table 3 with graphic representation in Figure 3. The effect measures all indicated increased risk, but were imprecise, with wide 95 percent confidence intervals that included the null. Three of the hazard ratios approached statistical significance, with the lower limit of the confidence interval 0.98 or 0.99.

Results of our analyses using IPTWs are presented in Additional file 1: Table S5 and Additional file 1: Table S6. The IPTWs balanced the sample characteristics for the bolus and maintenance groups with absolute standardized differences ranging from 0 – 0.15 with all differences < =0.05 with the exception of baseline iron which was 0.15. (Additional file 1: Table S5). These findings indicate that the weighting was successful in balancing the bolus and maintenance groups on baseline characteristics. The mean of the stabilized weights was 1.00, as expected [25], and there were no extremely small or large weights (0.25 – 9.1). A comparison of the results of the multivariable and IPTW-adjusted analyses is presented in Additional file 1: Table S6. For the full sample, there was little difference in the multivariate adjusted and IPTW adjusted results, particularly in regard to the infection outcomes. For the catheter subgroup, the point estimates for the two methods differed slightly more, relative to the full sample, and were much less precise. In most cases the point estimates were in the same direction for both analyses.

Consistent with previous work [18, 19], we found evidence of an association between bolus dosing and infection outcomes and no association between bolus dosing and cardiovascular outcomes, in a sample of patients receiving center-based hemodialysis. Patients receiving bolus versus maintenance dosing were more likely to have infection-related events, with the most notable rate differences for use of intravenous antibiotics and hospitalization for any infection. While some of the hazard ratios for the infection outcomes had confidence intervals that included the null, all of these hazard ratios were >1.00 (HR’s ranging from 1.08-1.17) and similar in magnitude to the hazard ratios in the previous study examining infection risks [18] (HR’s ranging from 1.05-1.11). All of the hazard ratios for the cardiovascular outcomes included the null, with one greater than 1.00 and the others less than 1.00. These hazard ratios ranged from 0.82-1.07, similar to those in the previous study examining cardiovascular risks [19] (HR’s ranging from 0.98-1.05).

The effect measures for the catheter subgroup (Table 3) were imprecise due to small sample size, but the magnitude of the rate differences suggests that this group, in particular, may be at greater risk of infection-related outcomes with bolus versus maintenance dosing. Of particular note are the rate differences for hospitalization for pneumonia, sepsis, or vascular access infection and the combination of these outcomes with infection-related death. Patients with a catheter who received bolus dosing experienced approximately 100 more events per 1000 person-years for these outcomes relative to patients with a catheter who received maintenance dosing. These findings also agree with the earlier study, which reported rate differences of approximately 75 more infection-related events per 1000 person-years when comparing bolus versus maintenance dosing in this subgroup [18].

Using the identical epidemiologic design and statistical methods as in the previous research, we examined the comparative short-term safety of bolus versus maintenance dosing using data on a different sample of U.S. center-based hemodialysis patients. While the sample for this analysis was much smaller (N = 48,050) than in the previous study (N = 776,203), our findings were remarkably similar. The characteristics of the two cohorts were similar in some respects, but differed in regard to some demographic and clinical characteristics. Also, a majority of patients in the current study were from the Northeastern part of the U.S., while most patients in the previous study resided in the South. The proportion of individuals that received bolus versus maintenance dosing in the two studies were remarkably similar (13.9% received bolus dosing in this study vs 12.6% and 49.3% received maintenance dosing vs 49.2%). While not particularly innovative, replicating studies provide important scientific information that may add to a body of evidence or call into question previous findings. Our study has several limitations, including its non-experimental design, its focus on short-term events only, and the potential for unmeasured confounding. We did examine our results with the addition of more covariates and found little change in the effect measures (Additional file 1: Table S3), suggesting that we adequately controlled for confounding due to measured covariates. Another limitation is that our study design required survival for at least 9 months following the start of dialysis, which limits the generalizability of our findings to incident hemodialysis patients. We also did not validate the categorization of our exposure variable against a gold standard. Rather, we examined the patterns and amount of iron dosing and used clinical expertise to arrive at our method of categorization.

Strengths of our study include the sound design, in which baseline covariates were identified before the exposure period and outcomes were ascertained after the exposure period, and our rich data base, which contained clinical data merged with administrative health care claims data. Building on earlier analyses, the results of this study provide further evidence of the potential risks of bolus dosing of intravenous iron. These risks should be considered in light of the reported benefits of bolus dosing, which include diminished ESA requirements and improved anemia management [1‐4].

We examined the short-term comparative safety of bolus versus maintenance dosing in a sample of center-based hemodialysis patients and found that bolus dosing was associated with increased infection risk, particularly in the subgroup of patients with a catheter. We found no association between dosing practices and cardiovascular outcomes.