Time-Dependent Associations between Iron and Mortality in Hemodialysis Patients

Materials and Methods

Results

The original 2-yr national database of all MHD patients included 69,819 cumulative subjects. After deleting MHD patients who did not maintain beyond 3 mo of hemodialysis treatment (5600 patients from the first seven quarters and 5870 patients from the last quarter), 58,349 MHD patients remained for analysis, 291 of whom had missing data. Hence, the resulting cohort included 58,058 MHD patients, 37,049 (64%) of whom originated from the first-quarter data set (q1) and the rest from the subsequent quarters (q2 through q8). Table 1 shows baseline demographic, clinical, and laboratory characteristics of the cohort according to the administration of intravenous iron. Diabetes was more prevalent among patients who received intravenous iron. Almost half of patients who received intravenous iron were among new patients (i.e., dialysis vintage of <6 mo), whereas only one quarter of those who did not receive intravenous iron were new. The BMI was 0.5 kg/m² higher whereas serum albumin and creatinine levels were slightly lower in the intravenous iron group. Among indices of iron and anemia, hemoglobin was 0.3 g/dl higher whereas serum ferritin was 162 ng/ml lower among those who received iron. Patients who received intravenous iron also received approximately 5000 units/wk higher dose of EPO.

Table 2 shows correlation coefficients between the three serum markers of iron stores (continuous values) and some clinically relevant variables at the baseline of the cohort. Serum ferritin had only weak to moderate correlations with serum iron and ISAT. Serum iron was positively associated with serum albumin level and TIBC. Blood hemoglobin had negative (inverse) association with serum ferritin but positive (direct) association with serum iron. Table 3 shows the 12 ferritin and 11 ISAT categories. Both all-cause and CV mortality exhibited increasing rates across increasing ferritin categories, whereas the opposite (inverse) association was observed for ISAT increments.

Table 4 and Figure 1 show the hazard ratios of death for time-varying ferritin categories. In unadjusted model, a serum ferritin >800 ng/ml during each quarter was associated with increased death rate, whereas in case-mix adjusted model, a tendency toward increased death rate first was observed when serum ferritin was >1000 ng/ml. However, after additional multivariate adjustment for the confounding effect of surrogates of inflammation and malnutrition, there was no increased death rate for ferritin levels as high as 1200 ng/ml. Figure 2 shows similar death risk analyses for time-varying ISAT categories. In all three models, a serum ISAT between 30 and 50% was associated with the lowest risk for all-cause mortality. Similarly, the risk for CV death was the lowest in the ISAT range of 35 to 50%. Figure 3 shows hazard ratios of death for time-varying serum iron categories. A moderately high serum iron between 60 and 100 μg/ml was associated with significantly better survival, whereas a serum iron <50 μg/ml displayed increased all-cause and CV death risk. Higher serum iron levels in the range of 100 to 120 μg/ml also tended to be associated with better survival, and levels even higher than 120 μg/ml still did not show any meaningful increase in risk for death when compared with the low serum iron levels <60 μg/ml. Figure 4 shows the associations between administered doses of intravenous iron and estimated relative risk for death. Compared with patients who did not receive any intravenous iron, administered intravenous iron up to 400 mg/mo was associated with lower relative risk for death, whereas doses >400 mg/mo correlated with increased death risk. To explore further the association between administering any intravenous iron (yes/no) and 2-yr survival and to explore possible interactions, we performed subgroup analysis using time-varying Cox model (Figure 5). The survival advantages of administering any intravenous iron seemed relatively consistent in four subgroups of serum ferritin and ISAT on the basis of cutoff values of 500 ng/ml and 50%, respectively,

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Association between serum ferritin and all-cause (top) and cardiovascular (CV; bottom) mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Association between serum iron saturation ratio (ISAT) and all-cause (top) and CV (bottom) mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Association between serum iron and all-cause (top) and CV (bottom) mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Association between administered intravenous iron and all-cause (top) and CV (bottom) mortality.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Association between administered intravenous iron and survival in different subgroups of patients on maintenance hemodialysis. Filled circles are based on unadjusted Cox models, whereas empty circles result from multivariate adjusted models. F, ferritin.

Discussion

Examining prospectively collected historical data of a large national database of 58,058 MHD patients in the United States with comprehensive time-varying quarterly laboratory values and medication doses for up to 2 yr, we found that after extensive time-dependent and multivariate adjustment for case mix, administered intravenous iron and EPO doses, and surrogates of nutritional status and inflammation, serum ferritin levels between 200 and 1200 ng/ml, serum iron levels between 60 and 120 μg/ml, and ISAT between 30 and 50% were associated with the lowest all-cause and CV death risks. We also found that compared with those who did not receive intravenous iron, administered intravenous iron up to 400 mg/mo was associated with improved survival. Although patients who received intravenous iron had significantly different demographic, clinical, and laboratory features at baseline, the survival benefits of intravenous iron were relatively consistent in different subgroups of MHD patients, including those who had high ferritin but low ISAT values. Hence, the previously observed associations between moderately high levels of serum ferritin and mortality in unadjusted and case-mix adjusted models (36) seem to be mostly due to the confounding effects of malnutrition and inflammation, because additional adjustments for surrogates of MICS mitigates or even reverses some of these associations.

High serum ferritin levels have been found to be associated with increased hospitalization, and a recent rise in serum ferritin is reported to be an imminent death risk in MHD patients (36). An association between dialysis morbidity, including risk for infection, and iron overload reflected by a high serum ferritin have also been reported (37). However, the foregoing studies had small sample size and did not control extensively for confounders. Hyperferritinemia-associated morbidity may be due to non–iron-related factors. Because serum ferritin is a positive inflammatory marker (10), hyperferritinemia-associated increased risk for infection and death may be a mere epiphenomenon. Thus, considering high ferritin levels as the primary cause of increased mortality in the setting of inflammation or infection and preventing optimal anemia management with intravenous iron for serum ferritin levels >500 ng/ml (17) or 800 ng/ml (16) may be flawed. Indeed, the recommended 800-ng/ml threshold as the upper level for serum ferritin in K/DOQI guidelines was mostly opinion based without adequate evidence (16).

We found that ISAT in range of 30 to 50% was associated with the best survival. This range is essentially within the recommended guidelines (16,17). However, it is important to note that ISAT is not a measured entity but a mathematical derivation of two other measures (serum iron divided by TIBC). Serum TIBC is a negative acute-phase reactant and a marker of MICS and poor outcome in MHD patients (9,18,36). In iron deficiency, the TIBC level tends to be elevated in individuals without kidney disease, whereas serum iron and TIBC levels change parallel to each other in MHD patients as shown in our study (r = 0.30). In addition to uremia, ISAT can be significantly confounded by MICS, because the denominator of ISAT (TIBC) is a nutritional and/or inflammatory marker (38,39). This feature of TIBC as a component of ISAT may explain why in our study a more prominent J-shaped association was observed with ISAT than with serum iron (Figures 2 and 3). Indeed, our found association between ISAT levels >50% and increased death risk may be an artificial finding and a mere reflection of the known association between low TIBC levels (which erroneously increases the calculated ISAT value) and death risk in MHD patients.

Among the three indicators of iron stores, the serum iron concentration, which is usually measured monthly in US dialysis facilities, is used less frequently by clinicians. Similarly, K/DOQI guidelines include inadequate practical recommendations that pertain to the interpretation of serum iron for clinical application (16). Serum iron levels may be decreased in conditions besides iron deficiency; the most frequent is probably the anemia associated with chronic conditions such as uremia and inflammation (40–42). Hypoferremia (i.e., low serum iron; not to be confused with hypoferritinemia) is observed during inflammation, neoplasia, trauma, myocardial infarction, surgery, and stressful conditions (41). Hence, our found association between low serum iron (<50 μg/ml) and increased death risk may be due to inflammation. Although, we showed the same associations between low iron and poor outcome even after adjustment for available surrogates of MICS, such conventional inflammatory markers as CRP were not available in our study. Nevertheless, in a recent study of a prospective 12-mo cohort of 1283 MHD patients, low baseline values of serum iron and ISAT were also associated with higher rates of mortality and hospitalization, and in the subcohort of the same study, these associations were also independent of inflammation as measured by serum CRP (8).

In our study, we found an unexpected association between moderate doses of intravenous iron (up to 400 mg/mo) and improved survival. This finding seems to be inconsistent with the notion that excessive intravenous iron is deleterious by leading to oxidative stress and predisposition to infection. Such concerns have led judiciously to relatively conservative policies including in K/DOQI guidelines for iron administration to dialysis patients (16). However, most reports concerning adverse effects of iron in dialysis patients are based on in vitro studies (43). Very few in vivo studies have shown an association between iron administration or higher iron indices and poor outcome in MHD patients, and most of these studies have a small sample sizes and high likelihood of selection bias (44). In another study, by Feldman et al. (45), although there was a tendency to increased mortality in MHD patients who had received higher doses of intravenous iron, additional analyses by the same investigators using time-varying marginal structural models did not confirm previous findings (46). A recent study showed that intravenous iron administration to a group of dialysis patients reduced levels of circulating TNF (47). Hence, it is possible that the observed survival advantage of intravenous iron is due to its mitigating effect on inflammation. It is of utmost importance to appreciate that examining associations between a prescribed medication and outcome is amenable to bias by indication (48); such treatments are usually the result of sophisticated processes of active decision making by well-trained and specialized physicians. Many nephrologists may practice according to the K/DOQI or other guidelines. Indeed, we found that MHD patients who did not receive intravenous iron had substantially higher serum ferritin levels, which may indicate that intravenous iron was not prescribed to such high-ferritin patients on the basis of the current guidelines. Although use of such novel epidemiologic tools as marginal structural modeling may mitigate the degree of bias by indication (49), such models were not examined in our study. Hence, the found associations should not be interpreted as causal relationships between the administered intravenous iron and improved survival.

Our study should be qualified because it is observational, rather than interventional, and because a mixed incident/prevalent MHD population was examined. Nevertheless, because essentially all MHD patients of the DaVita dialysis facilities were included in our analyses, the likelihood of selection bias is minimal. Moreover, all dialysis facilities were under uniform administrative care, and all laboratory tests were performed in one single laboratory with optimal quality assurance monitoring. Furthermore, we used 3-mo averaged measures rather than one single measure at baseline, and we adjusted for dialysis vintage in all multivariate models. However, a switch from iron gluconate to iron sucrose in the middle of the cohort might have confounded some associations. Moreover, the actually administered dose may be different from the billed dose, and the discrepancy may differ according to the type of intravenous iron medication used. Hence, caution should be practiced in interpreting our medication dose data. Another limitation of our study is the possible inclusion of cases with gastrointestinal bleeding, other sources of blood loss, or malignancies, which may lead to low serum iron and poor outcome. Moreover, MHD patients with intercurrent infection or systemic inflammatory diseases, in whom inflammation-induced hypoferremia can be observed, were not excluded. However, these cases are not too frequent to cause major confounding, especially because the entire national database was examined and because the laboratory values were 13-wk averaged values.

Although our study is based exclusively on mere observational data and, hence, no cause-effect association can be inferred with certainty, it is consistent with the thesis that moderately high levels of serum ferritin are not per se indicators of iron overload and should not be regarded as means to restrict medications for optimal anemia management. A low iron status may be as harmful as, if not more harmful than, the so-called iron overload. Indeed, in the previously reported cases of hemochromatosis and/or hemosiderosis among dialysis patients, the observed serum ferritin levels were well above the currently observed ranges in MHD patients, usually in 3000- to 10,000-ng/ml range (50). With widespread EPO administration to dialysis patients since the early 1990s, there has been much fewer reported, if any, cases of hemochromatosis or hemosiderosis in MHD patients despite rigorous use of intravenous iron (50). Although our results may have major clinical implications, the observational nature of our study prompts caution in interpreting and generalizing our findings. Interventional studies including randomized clinical trials are required to ascertain whether (1) in MHD patients with a low serum iron and/or moderately high serum ferritin, intravenous iron administration can effectively increase serum iron and (2) whether such an interventional increase in serum iron improves clinical outcome in these individuals.