Impact of Iron Supplementation on Hospital Length of Stay for Pneumonia or Skin and Skin Structure Infections: A Retrospective Cohort Study

Abstract

Objectives: Pathogenic organisms utilize iron to survive and replicate and have evolved many processes to extract iron from human hosts. The goal of this study was to elucidate the impact of iron supplementation given in the setting of acute infection. Methods: This was a retrospective cohort study of Veterans Affairs patients who received intravenous antibiotics for pneumonia or skin and skin structure infections. Five-thousand subjects were included in each of the 2 cohorts: iron-receiving and non-iron-receiving. Data was analyzed using Fischer’s Exact test if categorical and independent t-tests if continuous. Primary and secondary objectives analyzed with Cox proportional hazard regression and outcome rates estimated utilizing Kaplan-Meier method. Results: Five-thousand patients were included in each cohort. The iron cohort was significantly older (Mean-years: Iron = 71.6, No-iron = 68.9; mean-difference = 2.7, P < .0001) with reduced renal function (Mean-eGFR[mL/min/1.73 m²]: Iron = 67.2, No-iron = 77.4; mean-difference = 10.2, P < .0001). For the primary outcome, the iron cohort had a significantly longer mean length of hospital stay (10.4 days) compared to the no-iron cohort (8.7 days) (mean difference 1.7 days, P < .0001). Secondary outcome analysis showed the iron cohort received intravenous antibiotics for longer (Iron = 8.2 days, No-iron = 7.1 days; mean-difference = 1.1 days, P < .0001) with a higher proportion of 30-day readmissions (Iron = 15.6%, No-iron = 12.8%; proportion difference = 2.8%, P < .0001). No significant difference was found between cohort proportions for 30-day mortality (Iron = 12.7%, No-iron = 11.3%, proportion difference = 1.4%, P = .052). Conclusions: Baseline characteristic differences between cohorts is representative of patients who would be expected to require iron replacement therapy. Given the magnitude of primary and secondary-outcomes, further studies controlling for these factors would be warranted.

Summary of Main Point

Patients who received iron have a significant increase in hospital length of stay, duration of intravenous antibiotic use, and 30-day readmission when treated for pneumonia or skin and skin structure infection compared to patients who did not receive iron.

Introduction

Iron is an essential dietary mineral that is a critical component of hemoglobin and myoglobin, proteins utilized in oxygen provision for tissues within the human body. Additionally, iron is required for physical growth, neurological development, cellular functioning, and synthesis of some hormones. ¹ In addition to iron’s importance for humans, almost all human pathogens require iron for survival and replication. An iron deficiency can cause abnormal immune functioning; however, iron overload can lead to acute infectious exacerbations as free iron becomes more readily available for pathogenic use. ² One study found that the presence of iron in growth environments promotes Klebsiella pneumoniae causing liver abscess growth, biofilm formation, and enhanced virulence. ³ Because of this pathogen iron requirement, the human immune system responds to infectious diseases via a method known as nutritional immunity which is a process of sequestering iron into storage compartments, including macrophages, which limits iron availability for circulating pathogens.^4,5 Part of the nutritional immunity process leading to a hypoferremic state is releasing hepcidin which sequesters iron intracellularly and is released due to pro-inflammatory cytokines, such as IL-6 ⁶ , ⁷ as well as production from neutrophils and macrophages which produce hepcidin in response to infectious agents. ⁸

Despite the nutritional immunity process, pathogens have evolved methods of overcoming the body’s iron sequestering and are able to obtain iron via various acquisition systems. In gram-negative bacteria, there are outer membrane receptors which bind compounds containing iron or heme and transport them into the periplasmic space. For gram-positive bacteria, anchored proteins within the cell wall are able to transfer iron and heme containing compounds to an adenosine triphosphate binding cassette transporter which subsequently delivers iron and heme to the cytoplasm. In addition to these mechanisms, bacteria acquire iron via siderophores which bind free iron ions and extract iron from mammalian iron-binding proteins. Once acquired by a siderophore, iron is then delivered to outer membrane receptors where the iron-siderophore complex can be reduced to iron alone or can be actively transported across the cell membrane. Some outer membrane receptors are able to recognize mammalian iron-binding proteins and extract iron from them without the help of a siderophore.^9,10 Because of the importance of iron acquisition for the survival and proliferation of all noteworthy gram-positive pathogens, pharmaceutical companies have started developing vaccines and antibiotics which target the iron-acquisition capabilities of gram-positive pathogens.^10,11

Despite the well documented use of iron for human pathogens, as well as the clinical benefit of targeting pathogen iron acquisition, a literature search failed to identify studies that directly assessed the impact of iron replacement therapy given during acute infections on patient outcomes. One identified study by Cross et al looked at in-vitro bacterial growth in human serum and found an acute increase in growth rates from oral iron. The authors measured transferrin saturation (TSAT) in 48 iron replete males before and after iron supplementation with 2 mg/kg of ferrous sulfate. The TSAT mean increased from 42% to 76% in participants which was significantly associated with a profound increase in the serum growth of Escherichia coli, Yersinia enterocolitica, Salmonella enterica, and Staphylococcus epidermidis; however, no significant impact on the growth of Staphylococcus aureus, which preferentially scavenges heme, was shown. Cross et al ¹² concluded that highly soluble iron supplements such as ferrous sulfate could promote bacteremia due to accelerating bacterial growth prior to the induction of human immune defenses.

Due to the lack of knowledge on the impact of iron replacement therapy given in the setting of acute infection, this study aimed to evaluate if patients receiving concurrent iron supplementation and intravenous (IV) antibiotics had worse clinical outcomes compared to patients who did not receive iron replacement therapy during their acute infection.

Methods

This was a retrospective cohort study that utilized Department of Veterans Affairs (VA) data from the VA Corporate Data Warehouse (CDW) which is a repository of information from the VA’s electronic health record. ¹³ Data was collected from the period of 1/1/2010 to 1/1/2020 with use of the VA Informatics and Computing Infrastructure (VINCI). VINCI hosts all data available through CDW, as well as some unique data, and is utilized to query data from all hospitals included within the national VA healthcare system. ¹⁴ Institutional Review Board approval was obtained prior to the start of the study.

Inclusion criteria for VINCI database queries were veterans 18 years and older who were discharged from a VA hospital during the specified timeframe after receiving IV antibiotic therapy for pneumonia (PNA) or skin and skin structure infection (SSSI) with or without concomitant IV or oral iron supplementation. Only hospital admissions lasting 3 to 30 days were queried. Subjects were selected who had one hospital admission for PNA or SSSI during the study timeframe. Iron supplementation was not included if the source was from multivitamins or non-VA sources.

Subjects were excluded from VINCI queries if they received IV antibiotics for an infection other than PNA or SSI. These infections were chosen due to their high prevalence, which would represent a large proportion of infections, and a desire to limit query results. Additionally, patients were excluded from VINCI queries if they had a discharge diagnosis of sepsis based on ICD-9 code 995 and ICD-10 codes A40 and A41. Additional exclusion criteria included presence of an allergy to iron products, subject received oral antibiotics, IV antibiotic use for less than 3 consecutive days, IV vasopressor use at any point during hospitalization, and ICU admission at any point during hospitalization.

Once inclusion and exclusion criteria were applied to VINCI queries, the following data was collected: age, sex, height, weight, body-mass index (BMI), temperature, blood pressure, active problem list, admission diagnosis, discharge diagnosis, antibiotic(s) used (ie, drug names), iron dosage information (ie, drug name, route, and duration of use prior to and during admission), duration of IV antibiotics, length of hospital stay, 30-day readmission status, and 30-day mortality. Furthermore, the following labs were collected at time of admission: total iron, iron binding capacity, ferritin, iron saturation, serum creatinine, estimated glomerular filtration rate (eGFR), blood urea nitrogen (BUN), white blood cells (WBC), C-reactive protein, culture site during hospital stay, and culture results during hospital stay. Values for WBCs were excluded if greater or less than 3 standard deviations from the mean. Data returned from VINCI query was further limited to maximum cohort sizes of 5000 subjects selected at random. Random selection was done by assigning a random number to subjects via SQL version 17 random number generator and selecting the subjects associated with numbers 1 through 5000 for inclusion in study.

The primary outcome was length of hospital stay. Secondary outcomes included duration of IV antibiotic use, 30-day readmission, and 30-day mortality. Baseline demographic characteristics were described using mean and standard deviation for continuous variables and numbers and proportions for categorical variables using independent t-tests for comparison of continuous data and Fischer’s exact test for categorical data. The Charlson Comorbidity Index (CCI) was assessed by use of ICD-9 and ICD-10 codes pulled from the subject’s lifetime problem list. Primary and secondary outcome analysis was done utilizing Cox Proportional Hazards Regression. The cohort incidence rates of primary and secondary outcomes were estimated utilizing the Kaplan-Meier Method and log-rank test for comparison. Statistical significance was defined as a two-tailed alpha less than .05. Data was analyzed using SPSS version 17.

Results

Baseline characteristics of randomized patients are shown in Table 1. The vast majority of participants in both the iron and no-iron cohorts were male (96.6% and 97.1% respectively) which was expected given the sample population of United States veterans being predominantly male. Participants in the iron cohort had a significantly higher mean age than participants in the no-iron cohort (mean = 71.6 years in iron cohort and 68.9 years in no-iron cohort; mean difference (MD) = 2.7 years, P < .0001). Renal function was significantly reduced in the iron cohort compared to the no-iron cohort based on measures of total serum creatinine (mean = 2.8 mg/dL in iron cohort and 2.4 mg/dL in no-iron cohort; MD = 0.4 mg/dL, P = .027), and eGFR (mean = 67.2 mL/min/1.73 m² in iron cohort and 77.4 mL/min/1.73 m² in no-iron cohort; MD = 10.2 mL/min/1.73 m², P < .0001). Additional significantly different baseline characteristics were observed for weight, BMI, diastolic blood pressure, total iron, and iron binding capacity (see Table 1).

Table 1.

Baseline Characteristics.

	Iron (n = 5000) a	No-iron (n = 5000) a	P-value b
Age at admit, mean years (SEM)	71.6 (0.2)	68.9 (0.2)	<.0001
Elderly patients >70 y.o., n (%)	2687 (53.7)	2233 (44.7)
Male sex, n (%)	4829 (96.6)	4855 (97.1)
Weight (lbs), mean (SEM)	193.9 (0.8)	200.1 (0.8)	<.0001
BMI (kg/m2), mean (SEM)	28.4 (0.1)	29.1 (0.1)	<.0001
Serum creatinine (mg/dL), mean (SEM)	2.8 (0.1)	2.4 (0.1)	.027
eGFR (mL/min/1.73 m²), mean (SEM)	67.2 (0.7)	77.4 (0.7)	<.0001
BUN (mg/dL), mean (SEM)	30.4 (0.4)	24.1 (0.4)	<.0001
Systolic blood pressure (mmHg), mean (SEM)	129.0 (0.2)	128.7 (0.2)	.31
Diastolic blood pressure (mmHg), mean (SEM)	69.6 (0.1)	71.8 (0.1)	<.0001
WBC c (K/cmm), mean (SEM)	13.3 (0.7)	13.9 (0.7)	.517
C-reactive protein (mg/L), mean (SEM)	58.7 (2.6)	52.5 (2.6)	.094
Temperature (°C), mean (SEM)	36.7 (0.005)	36.7 (0.005)	.022
Total iron (μg/mL), mean (SEM)	29.8 (0.8)	42.8 (1.8)	<.0001
Iron binding capacity (μg/mL), mean (SEM)	224.4 (1.9)	210.8 (3.0)	.0004
Ferritin (ng/mL), mean (SEM)	356.3 (43.4)	442.2 (76.9)	.321
Iron saturation (%), mean (SEM)	15.3 (2.3)	29.1 (10.2)	.056

Note. SEM = standard error of the mean.

^aBaseline parameters with different n values due to a lack of available data for pulled subjects: weight (n = 4973 for iron and 4958 for no-iron), BMI (n = 4940 for iron and 4912 for no-iron), Creatinine (n = 3765 for no-iron and 3760 for iron), eGFR (n = 4132 for no-iron and 4234 for iron), BUN (n = 3829 for no-iron and 3882 for iron), SBP (n = 4917 for no iron and 4949 for iron), DBP (n = 4916 for no-iron and 4949 for iron), WBC (n = 4868 for no-iron and 4940 for iron), CRP (n = 787 for no iron and 839 for iron), Temp (n = 4913 for no-iron and 4940 for iron), Total iron(n = 270 for no-iron and 759 for iron), TIBC (n = 519 for no-iron and 1695 for iron), ferritin (n = 39 for no-iron and 111 for iron), TSAT (n = 4 for no-iron and 13 for iron).

^bP-values derived from two-tailed independent t-tests.

^cValues for WBCs were excluded if greater or less than 3 standard deviations from the mean.

Similar proportions of participants were treated for PNA and SSSI and had similar amounts of resistant bacteria within the iron and no-iron cohorts (Supplemental Table 1). Results of the CCI are available in Supplemental Table 2. The CCI was significantly different for the iron cohort compared to the no iron-cohort (mean iron-cohort index = 3.4, mean no-iron cohort index = 3.1; MD = 0.3, P < .0001). Supplemental Table 2 shows comparison of ICD-9 and ICD-10 codes for each indication measured as part of the CCI.

For the primary outcome analysis, participants in the iron cohort were found to be hospitalized significantly longer compared to patients in the no-iron cohort. Mean days of hospital stay was 10.4 (SEM ± 0.1) for the iron group and 8.7 (SEM ± 0.1) for the no-iron group giving a mean group difference of 1.7 day (P < .0001) (Table 2). The Kaplan-Meier curves of each cohort for probability of participants remaining in the hospital at 30-day are shown in Figure 1. The difference between the 2 curves was significantly different (P < .0001) with the probability of remaining in the hospital at 30-day being significantly higher for the iron cohort.

Table 2.

Primary and Secondary Outcomes: Length of Hospital Stay, Duration of IV Antibiotic Use, 30-Day Readmission, 30-Day Mortality.

	Iron (n = 5000)	No-iron (n = 5000)	P-value a
Days of hospital stay, mean (SEM)	10.4 (0.1)	8.7 (0.1)	<.0001
Days of antibiotic use, mean (SEM)	8.2 (0.1)	7.1 (0.1)	<.0001
30-d readmission, n (%)	781 (15.6)	639 (12.8)	<.0001
30-d mortality, n (%)	633 (12.7)	566 (11.3)	.052

Note. SEM = standard error of the mean.

^aP-values derived from two-tailed independent t-tests for days of hospital stay and days of antibiotic use, Fisher’s Exact test for 30-day readmission, and by log-rank test of Kaplan-Meier curves for 30-day mortality.

Figure 1.

Thirty-day Kaplan-Meier plot of hospital length of stay comparing iron cohort (blue) to no-iron cohort (red). N = 5000 for each cohort. All subjects experienced a hospital discharge by day 30 giving a probability of remaining in hospital beyond 30-day of zero. Between cohort difference was statistically significant with P < .0001 calculated using the log-rank test.

Secondary outcomes for days of antibiotic use, 30-day readmissions, and 30-day mortality are shown in Table 2. The mean number of days of antibiotic use was significantly higher in the iron cohort (mean = 8.2, SEM ± 0.1) compared to the no-iron cohort (mean = 7.1, SEM ± 0.1) with a mean difference greater of 1.1 day (P < 0001). The Kaplan-Meier curves of each cohort are compared in Figure 2 with a significant difference noted for probability of antibiotic use days (P < .0001). The number of patients with a 30-day readmission was 781 (15.6% of total population) for the iron cohort and 639 (12.8% of total population) for the no-iron cohort with a between cohort difference of 2.8% (P < .0001) (Supplemental Table 3). No significant difference was found in the 30-day mortality rates with 633 participants (12.7% of cohort) in the iron cohort and 566 participants in the no-iron cohort (11.3% of cohort) who were marked as deceased before or at 30 days from initial hospitalization (cohort difference = 1.4%, P = .052) (Figure 3).

Figure 2.

Thirty-day Kaplan-Meier plot of remaining on IV antibiotics comparing iron cohort (blue) to no-iron cohort (red). N = 5000 for each cohort. All subjects were discontinued from IV antibiotics by day 30 giving a probability of remaining on IV antibiotics beyond 30-day of zero. Between cohort difference was statistically significant with P < .0001 calculated using the log-rank test.

Figure 3.

Thirty-day Kaplan-Meier survival plot of mortality comparing iron cohort (blue) to no-iron cohort (red). N = 5000 for each cohort. Results did not reach statistical significance for proportion of iron cohort subjects (12.7% of subjects) compared to no-iron cohort subjects (11.3%), P = .052 calculated using the log-rank test.

Discussion

Our study has shown that patients who receive iron therapy while on IV antibiotics for treatment of either PNA or SSSI have a significantly longer length of hospital stay compared to patients who do not receive iron while on IV antibiotics for these diseases. Furthermore, there is a significantly greater number of days that antibiotics are given and a higher likelihood of readmission after 30-days when patients are receiving iron therapy.

In order to avoid selection bias and the potential skewing of results, after meeting inclusion criteria for their respective cohorts, we randomly selected participants for each group. This method of selection led to some significant differences in the baseline characteristics with a significantly older iron cohort with an overall worse renal function compared to the no-iron cohort. It has been well documented that renal function declines with age,^15-17 and the prevalence of anemia increases with declining renal function and worsening stages of CKD.^18-20 Thus there is an expected higher incidence of reduced renal function and anemia prevalent in elderly patients.^21,22 Therefore, the differences observed in the baseline characteristics between the iron and no-iron cohorts are likely reflective of which patients are expected to be anemic and consequently be initiated on iron therapy.

Similarly, the expected difference in age and reduced renal function of the iron cohort is reflected in the CCI which was significantly higher compared to the no-iron group. As shown in Supplemental Table 2 the CCI was assessed utilizing 19 unique factors pulled from ICD-9 and ICD-10 codes. Of those 19 factors, 9 were significantly different with 8 of the 9 favoring the no iron-cohort (iron cohort was favored in the factor of mild liver disease with significantly fewer patients represented). Of the factors which favored the no-iron cohort, 2 of them were for renal disease with more patients having mild and severe renal disease represented in the iron cohort, which was an expected outcome. Five of the remaining 6 factors were related to cardiovascular outcomes (MI, CHF, PCD, CVD, Plegia) which are thought to have increased incidence rates when renal function decreases.^23-25

Despite the differences in baseline characteristics, the magnitude of differences for the primary outcomes is likely greater than what would be seen based on age and renal function alone. A previous study ²⁶ showed that mortality from community acquired pneumonia (CAP) increased with age (65-74 years, 6.9%; 75-84 years, 8.9%; >85 years, 17.1%; P < .001); however, this was primarily due to comorbidities and multidrug resistant pathogens. One study that looked at risk factors associated with 30-day readmission from CAP showed that variables associated with pneumonia-unrelated hospital readmission included age ≥65 years-old and a CCI ≥2. ²⁷ Similarly, for SSSI, studies have shown that higher age and chronic renal disease is associated with increased mortality.^28-30 We believe that while age and renal function likely played a role in the between cohort difference, the magnitude of difference for the primary outcomes is unlikely accounted for by these factors alone. This is in-part due to the aforementioned studies finding an impact of age at ≥60 years-old which was met for both groups and the actual magnitude difference in age being relatively minimal with a mean difference of 2.7 years which, while statistically significant, may ultimately not be clinically significant. Likewise, while the difference in renal function was significant between the 2 cohorts, the magnitude of difference is likely not sufficient enough to account for the observed differences in the primary outcome.

Conclusion

Our study showed that in patients for whom iron replacement therapy is being given, there is a statistically significant increase in hospital length of stay, duration of IV antibiotic therapy, and 30-day readmission rates for PNA and SSSI compared to patients for whom iron replacement therapy is not given. Additionally, while not statistically significant, there was an overall higher rate of 30-day mortality in patients receiving iron replacement therapy. Our results were unclear as to whether or not the observed effects were directly related to increased free serum iron levels from iron replacement therapy or due to the presence of expected comorbidities of patients who would require iron replacement therapy. Our results would warrant a well-designed, randomized controlled trial to better elucidate the direct impact of iron replacement therapy on patient outcomes from acute infections.