Clinical features, outcomes, and HLA risk factors associated with nitrofurantoin-induced liver injury

Abstract

Background and Aim

Nitrofurantoin (NTF) is a widely used short-term for treating and long-term for preventing urinary tract infections. We aimed to describe the clinical characteristics, outcomes, and HLA risk factors for NTF-liver injury (NTF-DILI) among individuals enrolled by the Drug Induced Liver Injury Network (DILIN).

Methods

Seventy-eight patients with definite, highly likely, or probable NTF-DILI were enrolled into DILIN studies between 2004-2020. HLA alleles were compared between NTF-DILI and 3 control groups: population (n=14,001), idiopathic autoimmune hepatitis (AIH: n=231), and non-NTF DILI (n=661).

Results

Liver injury was hepatocellular in 69% and icteric in 55%. Compared to 18 short (≤7 days) and 16 intermediate (> 7 to <365 days) exposure cases, 44 long-exposure (≥ 1 year) NTF-DILI cases had greater frequency of AST > ALT (73% vs 33% vs 50%, P=0.018), ANA or SMA positivity (91% vs 44% vs 50%, P<0.001), and corticosteroid use (61% vs 27% vs 44%, P=0.06). In long-term NTF-DILI, bridging fibrosis, nodularity or cirrhosis, or clinical and imaging evidence for cirrhosis were present in 38%, and massive or sub-massive necrosis in 20%. No one in the short-exposure group died or underwent transplantation, whereas 7 (12%) patients from intermediate and long-term exposure groups combined died or had transplantation. After covariate adjustments, HLA-DRB1*11:04 was significantly more frequent in NTF-DILI compared to population controls (OR=4.29, P=1.15x 10⁻⁴), idiopathic AIH (OR=11.77, P=7.76 x 10⁻⁵), and non-NTF DILI (OR=3.34, P=0.003).

Conclusion

NTF-DILI can result in parenchymal necrosis, bridging fibrosis, cirrhosis, and death or liver transplantation, especially with long-term exposure, and is associated with HLA-DRB1*11:04. For mitigating against serious liver injury associated with NTF, the regulators should revise the prescribing information and consider other mitigation strategies.

Lay Summary

Liver injury due to nitrofurantoin (NTF) exhibits two patterns: one with short-term exposure and another with long-term exposure. Liver injury due to short-term NTF exposure resembles acute hepatitis, with a low frequency of autoantibodies and benign outcomes. Liver injury associated with long-term exposure resembles chronic hepatitis with moderate to severe inflammation, necrosis, and fibrosis. Most long-term cases have autoantibodies and other features of autoimmune liver disease and are treated with a limited course of corticosteroids, although a percentage remain on treatment for longer. Morbidity and mortality are high for liver injury associated with long-term NTF use. Certain genetic variants appear to increase the risk of developing NTF hepatotoxicity.

Introduction

Nitrofurantoin (NTF) is an oral antibiotic that is commonly used either short-term for acute urinary tract infections (UTI) or long term as suppressive therapy to prevent recurrent UTIs. NTF is particularly attractive because of its rapid oral absorption, extensive urinary excretion, and rarity of antibacterial resistance (1). In 2019, an estimated 1.6 million individuals received a total of 2.95 million prescriptions for NTF in the United States (2). NTF is also a well-known cause of drug induced liver injury (DILI) which can arise with either short-term or long-term use (2). NTF is also frequently listed as a cause of drug-induced acute liver failure both in the United States and abroad (3–5). The liver injury due to nitrofurantoin can also present with chronic liver injury that may result in cirrhosis and end stage liver disease (1).

First approved in the 1950’s, NTF was not linked to instances of liver injury until ten years later. Since, then there have been many case reports and several case series describing aspects of NTF liver injury, but mostly from the 70s and 80s from central registries of Denmark, Sweden, and New Zealand (1, 6–10). These older reports lacked in-depth description of latency, clinical characteristics, outcomes, and histological features.

HLA and non-HLA risk factors have recently emerged as important risk factors for liver injury caused by agents such as flucloxacillin (11), amoxicillin-clavulanate (12), terbinafine (13), and trimethoprim-sulfamethoxazole (14). NTF has been reported to be associated with carriage of HLA-B8 in a case report of a woman with chronic active hepatitis due to repeat courses of NTF (15). In a case series of 38 patients with likely NTF liver injury, HLA-DR2 and HLA-DRw6 were present in 56% vs 29% in a control group (6). In a more recent larger study, NTF liver injury was reported to not be associated with the HLA alleles typical of spontaneous idiopathic autoimmune hepatitis, namely, HLA-DRB1*03:01 and HLA-DRB1*04:01 (16). Systematic investigation of HLA alleles associated with NTF liver injury and with its characteristics and outcomes heretofore are lacking.

In this paper, we describe the clinical characteristics, phenotypes of liver injury, histology, and outcomes in a cohort of patients with well characterized NTF-associated liver injury who were enrolled in observational studies of the U.S. Drug Induced Liver Injury Network. Analyses also included characterization of HLA alleles associated with NTF-liver injury compared to population controls, patients with DILI due to other agents, and patients with idiopathic, non-drug related AIH.

Methods

The Drug Induced Liver Injury Network (DILIN) was established in 2003 to investigate the implicated agents, characteristics, risk factors, natural history, and outcomes of idiosyncratic DILI in the US. DILIN supports two ongoing studies: a Prospective Study (ClinicalTrials.gov Identifier: NCT00345930) and a Retrospective Study (ClinicalTrials.gov Identifier: NCT00360646). Their design and inclusion and exclusion criteria of the Prospective Study have been described previously in detail (17,18). The details of the DILIN Retrospective Study are described at: https://clinicaltrials.gov/ct2/show/NCT00360646. In brief, adults and children older than 2 years of age with suspected DILI are enrolled and undergo structured evaluation and assessment for competing etiology, disease severity, and causality. The process for adjudicating the causality and the causality and severity scores have been previously published (17, 18). A total of 78 patients included in these two studies were adjudicated to have definite, highly likely, or probable DILI due to nitrofurantoin (NTF-DILI). This study was approved by the institutional review boards of the participating institutions, and all participants have provided an informed consent prior to enrollment. Some of the patients described in this paper have been included in other DILIN publications (3, 13, 16, 18,19).

Genome wide single nucleotide polymorphisms (SNPs) data were generated for samples in DILIN as described previously (20). HLA sequencing was performed on DNA extracted from whole blood of DILI patients in DILIN using Illumina MiSeq platform at the Vanderbilt University Medical Center Immunogenomics, Microbial Genetics, and Single Cell Technologies core as previously described in Li et al. (14). Of 78 nitrofurantoin DILI cases, 73 have both HLA and genome-wide SNP data. To investigate HLA allelic association with DILI induced by nitrofurantoin, we compared their HLA profiles to three comparison groups: (1) population controls: assembled from five dbGaP GWAS datasets, (2) idiopathic AIH controls: patients with well characterized autoimmune hepatitis (AIH) enrolled at Indiana University, and (3) Non-NTF DILI controls: patients caused definite, highly likely or probably DILI due to drugs other than NTF from the DILIN studies. To assemble the population control dataset, we obtained the approval from NCBI dbGaP to access the following datasets: 1) eMERGE-I: Genome Wide Association Studies of Network Phenotypes (phs000360.v3.p1); 2) KORA: Genetic Epidemiology of Refractive Error in the KORA study (phs000303.v1.pl); 3) KIDRISK: Pooled Genome-Wide Analysis of Kidney Cancer Risk (KIDRISK, phs001271.v1.p1); 4) Ashkenazi: Genetics of Schizophrenia in an Ashkenazi Jewish Case-Control Cohort (phs000448.v1.p1); and 5) BioMe: The Charles Bronfman Institute for Personalized Medicine BioMe BioBank (phs000925.v1.p1). The population control dataset includes all subjects in eMERGE-1 and BioMe, controls from KIDRISK and Ashkenazi, and subjects without high myopia (defined by refractive error < −5 diopter of at least one eye) from KORA. Because all participants with HLA sequencing were females, we restricted all comparison groups to females only. Race stratified subsets of European American, African American, and Hispanics, were created. The race designation was based on the genetic ancestry inferred by genome-wide SNP data on the study subjects and the 1000 genomes project as previously described (14). Since HLA data were not available in population control and de novo AIH datasets, we imputed four-digit HLA alleles based on SNPs in the major histocompatibility region of chromosome 6 using the HLA Genotype Imputation with Attribute Bagging (HIBAG) program (20). Finally, principal components for each subject were computed by Eigensoft program (21) using genome-wide SNPs in low linkage disequilibrium (r² < 0.3) for the merged case-control dataset for European Americans.

Statistical Analysis:

Summary statistics for demographic and clinical characteristics were computed for all nitrofurantoin DILI patients, those with exposure to nitrofurantoin for ≤ 7 days (short exposure), intermediate duration exposure (7-364 days), and those with latency ≥ 365 days (long exposure), respectively. That is, mean (standard deviation, SD) or median (interquartile range, IQR) were computed for continuous variables and frequency (percentage) were computed for categorical data as appropriate. Genetic association analysis was performed for all HLA alleles, the missense variant rs2476601 in PTPN22 (190), and rs1363907 in ERAP2 (22) for the full dataset. We used multivariable logistic regression models with covariate adjustment to test the allelic association with NTF-DILI. More specifically, for the comparison to population controls, logistic regression models with Firth correction were used to accommodate the small proportion of cases. For the comparisons to idiopathic AIH and non-NTF DILI groups, conventional logistic regression models were used. Covariates adjusted included age, PC1, and PC2 except for the comparison to idiopathic AIH group for which only age was adjusted. We excluded HLA alleles with population AF < 0.01 due to their rarity that may lead to unstable statistical results and not be practical in clinical usage such as screening high risk patients. False discovery rate based on Qvalue was computed to correct for multiple testing (23). We consider those HLA alleles or variants meeting qvalue < 0.05 as significant associations and the rest meeting qvalue < 0.20 as potential candidates for the susceptibility of developing DILI due to the use of nitrofurantoin.

Results

Overall Cohort

Between September 2004 and January 2021, a total of 2272 patients with suspected DILI were enrolled in the DILIN Prospective Study of which 2230 underwent formal adjudication by the DLIN Causality Committee. In these patients, NTF was listed as an implicated agent in 91, 68 of whom were considered probable (n=17), highly likely (n=37), or definitely (n=14) due to NTF. In the concurrently enrolled DILIN Retrospective Study, another 10 patients were identified with DILI which were probably (n=3), highly likely (n=6), or definitely (n=1) ascribed to NTF, which yielded a total analytic cohort of 78 patients.

Selected salient features of the 78 cases of NTF-DILI are shown in Table 1 and Supplemental Table 1. All but one were women, primarily Caucasian, with a mean age of 63 years (range 39 to 88 years). The duration of exposure prior to onset of documented liver injury (latency) varied greatly, ranging from 1 day to greater than 10 years (median = 463 days). Eighteen patients had short exposures (≤ 7 days), 16 had an intermediate duration (8 to 364 days), and 44 patients had longer exposures (≥ 1 year) to nitrofurantoin. The initial mean serum ALT was 801 U/L and AST 741 U/L. Interestingly, the initial AST value was higher than the ALT in 43 (55%) patients. The initial mean serum alkaline phosphatase (Alk P) was 243 U/L. Using the R-value, the biochemical pattern of liver injury at initial recognition was considered hepatocellular (R≥5) in 69%, mixed (R>2 and <5) in 22%, and cholestatic (R ≤2) in 9%. The initial mean serum total bilirubin was 6.2 mg/dL, and 55% of patients were jaundiced at DILI recognition.

Table 1:

Selected characteristics of individuals who presented with liver injury after receiving nitrofurantoin for short (≤ 7 days), intermediate (8-364 days), or longer duration (≥ 365 days)^¶

	Short exposure (N=18)	Intermediate exposure (N=16)	Longer exposure (N=44)	P-value
Age (years, mean [SD])	56 [12]	65 [11]	65 [10]	0.016
Women, n (%)	17 (94%)	16 (100%)	44 (100%)	0.4
Self-reported race, n (%)
White	15 (83%)	14 (87.5%)	42 (95%)	0.15
African American	3 (17%)	2 (12.5%)	1 (2%)
Other/Multiracial	-	-	1 (2%)
BMI (kg/m2, mean [SD])	27.7 [5.1]	27.1 [5.9]	27.9 (6.1)	0.9
Prior drug allergies, n (%)	10 (56%)	8 (50%)	27 (61%)	0.7
Rash, n (%)	4 (22%)	3 (19%)	7 (16%)	0.85
Preexisting Liver Disease, n (%)	0 (0%)	3 (18%)	2 (5%)	0.23
Diabetes mellitus, n (%)	2 (11%)	5 (31%)	13 (29.5%)	0.275
Latency* (days, median, [IQR])	9 (5-19)	89 [39-214]	765 (564-1037)	<0.001
Jaundice, n (%)	7 (39%)	9 (56%)	27 (61%)	0.27
Pattern of liver injury at presentation, n/N (%)
Hepatocellular	12/16 (75%)	11/14 (78%)	27/42 (64%)	0.64
Mixed	2/16 (12.5%)	3/14 (21%)	11/42 (9.5%)
Cholestatic	2/16 (25%)	-	4/42 (9.5%)
Eosinophilia (>500/μL), n (%)	3/16 (7%)	0/15 (0%)	2/38 (5%)	0.14
Initial AST > ALT, n (%)	6 (33%)	8 (50%)	32 (73%)	0.018
ANA or SMA positivity, n (%)	8 (44%)	8 (50%)	40 (91%)	<0.001
Liver biochemistries – Initial
ALT (U/L, mean [SD])	796 [809]	843 [599]	788 (644)	0.8
Alk P (U/L, mean [SD])	229 [112]	241 [129]	249 (143)	0.9
Total bilirubin (mg/dL, mean [SD])	5 [5.3]	8.4 [8.7]	5.8 (6.0)	0.8
INR (mean [SD])	1.4 [0.2]	1.3 [0.4]	1.6 (0.9)	0.6
Liver biochemistries – Peak
ALT (U/L, mean [SD])	854 [945]	1058 [729]	854 (650)	0.4
Alk P (U/L, mean [SD])	245 [93]	268 [145]	305 (255)	0.9
Total bilirubin (mg/dL, mean [SD])	6.7 [8.7]	13 [17]	8.7 (8.9)	0.8
INR (mean [SD])	1.3 [0.45]	2.1 [3.0]	1.8 (1.7)	0.4
Improvement in liver tests – median days
Peak ALT to below ULN	54.5	73	75.5	0.9
Peak Alk P to below ULN	22	218	76.5	0.08
Peak bilirubin to ≤ 2.5 mg/dL	14.5	73	33	0.06
Severity of Liver Injury, n (%)
Mild	8 (44%)	6 (37.5%)	17 (39%)	0.2
Moderate	8 (44%)	5 (31%)	11 (25%)
Severe	2 (11%)	2 (12,5%)	13 (29%)
Fatal	0 (%)	3 (19%)	3 (7%)
Causality, n (%)
Definite	2 (11%)	3 (19%)	10 (23%)	0.048
Highly Likely	6 (33%)	9 (56%)	27 (61%)
Probable	10 (56%)	4 (25%)	7 (16%)
Corticosteroids, n (%)	5 (27%)	7 (44%)	27 (61%)	0.06
Hospitalization or non-DILI hospitalization prolonged, n (%)	12 (67%)	9 (56%)	21 (48%)	0.4
Death**, at any time n/N (%)	0/16 (0%)	2 (14.3%)	3/38 (7.9%)	0.3
- Liver–related (%)	-	2 of 2 (100%)	2 of 3 (67%)	1.0
Liver Transplantation, n (%)	0 (0%)	1 (6.3%)	2 (4.5%)	0.8
DILI persistent at 6 months, n/N (%)	3/16 (19%)	0/13 (0%)	8/38 (21%)	0.19

^*Latency is the duration between nitrofurantoin start to DILI recognition.

^**Data on death are limited to the 68 cases identified in the Prospective Study.

^¶Univariate comparisons were made using standard ANOVA analyses. P value <0.05 was considered statistically significant.

Abbreviations: SD: Standard deviation; IQR: Interquartile range; BMI: Body mass index; ANA: Anti-nuclear antibody; SMA: Smooth muscle antibody; AST: Aspartate aminotransferase; ALT: Alanine aminotransferase; Alk P: Alkaline phosphatase; INR: International Normalized Ratio; ULN: Upper limit of normal; DILI: Drug induced liver injury

Anti-nuclear antibody (ANA) was positive in 45 (58%) patients, smooth muscle antibody (SMA) in 35 (45%), and neither in 8 (20%). No one had hemolytic anemia or peripheral neuropathy, but one woman on NTF for more than a year developed drug-induced pneumonitis which improved upon stopping. Corticosteroids were administered to 50% of patients. The liver injury severity was scored as mild in 40%, moderate in 30%, severe in 22%, and fatal (death or liver transplant within 6 months of onset) in 8%. There was evidence for persistent liver injury at 6 months in 16%.

Characteristics of liver injury according to the duration of NTF exposure

The exposure to NTF was 7 days or fewer in 18 patients (23%), 8-364 days in 16 patients (21%), and greater than 1 year in 44 patients (56%). Liver injury due to short NTF exposure had somewhat distinct characteristics, compared to other groups especially those with liver injury due to longer term NTF exposure. The characteristics of three groups of patients are described in Table 1 and individual patient level data for 18 patients with liver injury due to short NTF exposure are described in Supplemental Table 2. Patients with a short-term exposure were younger than those with a longer exposure (mean age 56 vs 65 years) and were less likely to have jaundice (39% vs 61%), an AST higher than ALT (39% vs 73%), to undergo liver biopsy (33% vs 73%), to receive corticosteroids (27% vs 61%), to be scored as having severe disease (11% vs 29.5%), and to die or undergo liver transplantation (0% vs 9%). Those with a short NTF exposure included two patients who developed liver injury after the first day of treatment. Interestingly, both patients had received multiple courses of NTF in the past, but without previous evidence of liver injury. Four patients with advanced liver injury due to short NTF exposure had a history of intermittent use of NTF and had the reversal of typical ALT to AST ratio.

Patients with liver injury due to longer NTF exposure often had autoimmune hepatitis-like features and evidence of advanced fibrosis or cirrhosis. The AST/ALT ratio > 1, suggesting advanced fibrosis or cirrhosis, was frequent. Liver biopsies were performed in 34 out of 44 patients. Histological bridging fibrosis, nodularity or cirrhosis, or clinical and imaging evidence for cirrhosis was present in 17 (38%) patients, whereas massive or sub-massive necrosis was present histologically in another 9 (20%) patients.

There were 16 patients who had DILI following intermediate duration exposure (8 to 364 days) to NTF. They were all women with mean age 65 years, 14 (87.5%) White, AST to ALT ratio was greater than 1 in 8 (50%) and ANA and/or SMA positive in 8 (50%). The pattern of liver injury at presentation was hepatocellular in 79% and mixed in 21%. This group had a mixture of acute hepatitis and chronic hepatitis like picture. Patients with few weeks of NTF tended to have acute hepatitis like picture whereas those with 6 months or longer NTF exposure had chronic hepatitis. Six patients had liver biopsies, which showed panlobular necrosis in one patient (9 days of NTF exposure) and chronic hepatitis with varying degrees of fibrosis in 4 other patients who received NTF for 6-9 months. Forty-four percent received corticosteroids and 12.5% died or received liver transplantation.

Outcomes

Five patients died and two others successfully underwent liver transplantation (Supplemental Table 3). Of 5 patients who died, one with cholestatic pattern of liver injury, died from pulmonary fibrosis, whereas another one died following liver transplantation due to multiorgan failure. Their ages ranged between 46 and 81 years and the pattern of liver injury was hepatocellular in 4, mixed in 2, and cholestatic in 1. The duration of exposure to nitrofurantoin was greater than 1 year in 5 individuals, 155 days in one, and 34 days in one. The pattern of enzyme elevations was hepatocellular in 4, mixed in 2 and cholestatic in one. AST was greater than ALT in 6 patients, and either ANA or ASMA was positive in all 7. Four patients received corticosteroids. Liver histology showed acute hepatitis with massive necrosis and collapse in 2 and severe acute hepatitis in 1 patient. One patient developed acute decompensation of possible preexisting cirrhosis likely due to prior nitrofurantoin exposure.

Other important aspects of NTF liver injury

Autoantibodies were highly prevalent in patients with NTF liver injury, especially those with liver injury due to long term NTF exposure. Further description of details of autoantibody positivity and their significance is provided in the Supplemental Results and Supplemental Table 4. Short term corticosteroid use was frequent although 15% were placed on long term azathioprine due to chronic hepatitis and fibrosis. Further details of corticosteroid and other immunosuppressant use are shown in Supplemental Results. Twenty-five liver specimens were centrally reviewed in a blinded fashion by an expert hepatopathologist (D.E.K). The pattern of injury was mainly either acute (panacinar) hepatitis (10 biopsies) or chronic (portal-based) hepatitis (9 biopsies) with one biopsy showing multiple non-necrotizing granulomas along with acute hepatitis. Bridging or multiacinar necrosis was seen in 13 biopsies. Increased numbers of plasma cells, suggestive of autoimmune hepatitis, were seen in 11 biopsies. Fibrosis was noted in 15 cases, with 5 cases showing bridging fibrosis and 3 showing cirrhosis. Table 2 and Supplemental Results provide further details of liver histology in patients with NTF liver injury.

Table 2.

Liver histology of liver injury due to nitrofurantoin

	Duration of Exposure	ANA or SMA	Corticosteroids	Onset to Biopsy (days)	Inflammation	Plasma Cells/Eosinophils	Fibrosis stage	Pattern*
1	≥ 365 days	Pos	No	40	Severe	Pos/Pos	3	Acute and granulomatous hepatitis
2	≥ 365 days	Pos	No	55	Moderate	Neg/Pos	2	Chronic hepatitis
3	≥ 365 days	Pos	No	66	Mild	Pos/Neg	5	Chronic hepatitis with chronic cholestatic changes
4	≥ 365 days	Pos	Yes	17	Severe	Pos/Pos	0	Acute hepatitis
5	≥ 365 days	Pos	Yes	11	Moderate	Neg/Neg	1	Acute hepatitis
6	10 months	Neg	Yes	85	Moderate	Neg/Pos	2	Chronic hepatitis with chronic cholestatic changes
7	≥ 365 days	Pos	Yes	9	Severe	Pos/Neg	3	Acute hepatitis with mild cholestasis
8	≥ 365 days	Pos	Yes	298	Severe	Pos/Neg	6	Chronic hepatitis
9	≥ 365 days	Pos	Yes	50	Severe	Neg/Pos	4	Acute hepatitis
10	6 months	Pos	No	15	Severe	Pos/Neg	3	Chronic hepatitis
11	≥ 365 days	Pos	No	34	Moderate	Pos/Neg	2	Chronic hepatitis
12	≥ 365 days	Pos	No	12	Mild	Neg/Neg	0	Acute hepatitis
13	≥ 365 days	Pos	Yes	41	Severe	Pos/Neg	2	Acute hepatitis
14	≥ 365 days	Pos	N/A	13	Severe	Neg/Neg	0	Acute hepatitis with massive necrosis
15	≥ 365 days	Pos	No	64	Moderate	Neg/Neg	3	Chronic hepatitis
16	7 days	Pos	Yes	4	Moderate	Pos/Neg	0	Acute hepatitis
17	≥ 365 days	Pos	No	23	Severe	Pos/Neg	1	Chronic hepatitis
18	≥ 365 days	Pos	Yes	32	Mild	Neg/Neg	6	Chronic hepatitis
19	7 days	Pos	Yes	8	Severe	Pos/Pos	1	Acute hepatitis
20	1 day	Neg	Yes	239	Mild	Neg/Pos	1	Chronic hepatitis
21	9 days	Pos	Yes	4	Severe	Neg/Neg	0	Acute hepatitis with mild cholestasis

Abbreviations: ANA, antinuclear antibody; SMA, smooth muscle antibody

HLA and genetic associations

Among 78 high confidence DILI cases caused by nitrofurantoin, 73 individuals (all females) had HLA sequencing and genome-wide genotype data. The comparison groups consist of 14,001 population controls, 661 non-NTF DILI, and 231 idiopathic AIH controls, all exclusively female. From the multivariable logistic (or Firth) regression analysis, eight HLA alleles met qvalue ≤ 0.20 criteria in at least one of comparisons, including two HLA alleles, HLA-DRB1*11:04 and DPB1*01:01 that showed significant association with nitrofurantoin liver injury in at least one comparison, i.e. meeting qvalue ≤ 0.05 after correcting for multiple testing (Table 3). The HLA-DRB1*11:04 was the most significant allele increasing the risk of developing nitrofurantoin liver injury (AF=0.08) with consistent results across all three comparison groups, including population controls (AF=0.02: OR (95% CI)= 4.29 (2.18,7.66), p=1.15x10⁻⁴); non-NTF DILI controls (AF=0.02: OR(95% CI)=2.88 (1.05, 3.39), p=0.013); and AIH controls (AF=0.009: OR(95% CI)=11.77(3.28, 58.79), p=7.76 × 10⁻⁵). There were 11 cases with HLA-DRB1*11:04 and all were European Americans, 9 had long-exposure, 2 had intermediate exposure, and none with short exposure. The clinical characteristics except for alkaline phosphatase levels were not different between those with and without HLA-DRB1*11:04 carriage (data not shown). Finally, the rs2476601 (A allele) in PTPN22, and rs1363907 (GG homozygous) in ERAP2 were not associated with nitrofurantoin liver injury in the comparison to population controls (P=0.56 for PTPN22, 0.56 for ERAP2) and non-NTF DILI controls (P=0.68 and 0.85). These two variants were not genotyped in the AIH controls. HLA analysis based on race-stratification and other secondary analyses are shown in Supplementary Materials. HLA-DRB1*11:04 remained significant (qvalue < 0.05) in the European Americans and patients with long-exposure of nitrofurantoin, respectively (Supplementary Table 5).

Table 3:

HLA alleles showing significant association with NTF-DILI in at least one comparison of all NTF-DILI cases to three comparison groups in females^¶

HLA Allele	All Cases (N=	vs. Population Controls (N=14,001)			vs. Non-NTF DILI (N=661)			vs. AIH (N=231)
	AF	AF	OR (95% CI)	P	AF	OR (95% CI)	P	AF	OR (95% CI)	P
All NTF-DILI cases (N=73)
DRB1*11:04	0.08	0.02	4.29 (2.18, 7.66)	1.15x10−4*	0.02	2.88 (1.05, 3.39)	0.013	0.01	11.77 (2.28, 58.8)	7.76x10−5*
DRB1*07:01	0.08	0.12	0.62 (0.33, 1.06)	0.083	0.13	0.57 (0.27, 0.98)	0.015	0.16	0.42 (0.20,0.80)	0.007**
DQA1*02:01	0.07	0.12	0.51 (0.26, 0.92)	0.022	0.13	0.48 (0.21, 0.84)	0.004**	0.16	0.34 (0.16, 0.67)	0.001**
DQA1*03:01	0.17	0.09	1.87 (1.185, 2.83)	0.008	0.17	1.06 (0.59, 1.87)	0.999	0.09	2.36 (1.31, 4.27)	0.004**
DQB1*03:03	0.01	0.03	0.39 (0.08, 1.12)	0.087	0.05	0.28 (0.03, 0.79)	0.005**	0.05	0.34 (0.06, 1.09)	0.072
DPB1*01:01	0.04	0.14	0.56 (0.22, 1.18)	0.134	0.10	0.63 (0.52, 1.42)	0.265	0.14	0.27 (0.10, 0.59)	3.82x10−4*
DPB1*03:01	0.03	0.08	0.37 (0.14, 0.80)	0.008	0.08	0.39 (0.12, 0.83)	0.005**	0.08	0.38 (0.12, 0.95)	0.039

^*Meeting qvalue ≤ 0.05;

^**qvalue < 0.20.

Abbreviations: NTF: nitrofurantoin; DILI: drug induced liver injury; AIH: autoimmune hepatitis; AF: allele frequency; OR, odds ratio; CI: confidence intervals

^¶For comparison vs population controls, logistic regression models with Firth correction were used to accommodate the small proportion of cases. For the comparisons to idiopathic AIH and non-NTF DILI groups, conventional logistic regression models were used. Covariates adjusted included age, PC1, and PC2 except for the comparison to idiopathic AIH group for which only age was adjusted.

Discussion

This paper, based on an ongoing prospective observational study, comprehensively characterizes the HLA risk factors, clinical features and outcomes in a large cohort of individuals with liver injury due to NTF. Key observations are (a) liver injury due to NTF is associated with significant morbidity and mortality; (b) the nature and outcomes of liver injury due to NTF are distinct, based on whether the exposure is short or long term; (c) autoantibodies are frequent, especially in those with liver injury due to long term exposure; but their presence was not associated with different liver injury characteristics, outcomes, or use of immunosuppression; (d) long term immunosuppression is needed in a small proportion of patients; and (e) there are HLA alleles associated with NTF liver injury.

The mortality associated with NTF liver injury was 7.4%, and 80% of such cases were adjudicated to be liver-related deaths. In addition, 4% of patients underwent liver transplantation. This frequency of death is slightly higher than what we reported for the entire DILIN cohort (6.2%) and, more importantly, the proportion of deaths that were considered liver-related was higher for NTF than other agents in the DILIN cohort (80% vs 64%) (3). In long-term NTF-liver injury, bridging fibrosis, nodularity or cirrhosis, or clinical and imaging evidence for cirrhosis was present in 38% and histologic evidence of massive or sub-massive necrosis in and additional 20%. For comparison, the liver injury played a primary role in leading to mortality in 64% of all patients who died within 2 years in the entire DILIN prospective cohort (24). Liver injury due to green tea extract, the most common herbal and dietary supplement to cause liver injury in the DILIN prospective study, was not associated with any deaths but 7.5% underwent transplantation (25). The package inserts for NTF (Macrodantin®, Furadantin®, Macrobid®) (26–28) recommend that patients should be monitored periodically for changes in biochemical tests that would indicate liver injury, but in our experience this recommendation was seldom followed. Among the 78 patients with NTF-DILI described in this report, none were detected by prospective monitoring. To mitigate the risk of worse outcomes and serious liver injury due to NTF, we believe a strong caution or a black-box warning in the label and a well-defined Risk Evaluation and Mitigation Strategy (REMS), consisting of educating prescribing clinicians and dispensing pharmacists should be considered by the regulators. This may be particularly important as NTF was suggested to be a better option than fluoroquinolones as first-line therapy for uncomplicated lower urinary tract infections to reduce the risk of bacterial resistance to fluoroquinolones (29).

This study further builds on previous descriptions that distinguish between acute and chronic forms of liver injury due to NTF (1). Liver injury associated with short exposure to NTF was abrupt in onset, largely hepatocellular, and arose in some cases after a single dose. Autoantibodies were infrequent, liver histology showed little or no fibrosis, and corticosteroids, while often used, could usually be withdrawn eventually without relapse. Importantly, liver injury was less severe in the short-exposure; none of the patients died or required liver transplantation. In contrast, the liver injury associated with long term exposure to NTF more commonly presented with jaundice (61%) and autoantibody positivity (91%). Nearly 70% of patients with liver injury after long term NTF use had an AST to ALT ratio >1, suggesting significant fibrosis. Evidence for bridging fibrosis or cirrhosis was present in 38%. Corticosteroids were used in a majority and azathioprine in a quarter of patients with long-exposure associated NTF liver injury. A limitation of our study is that we lacked precise dose and duration of corticosteroid use as well as specific indications for their use in this cohort.

Thirty-nine (50%) patients received corticosteroids during the initial presentation. Most of these patients had autoantibodies and primarily hepatocellular pattern of liver injury and were deemed to exhibit AIH-like liver injury. Although corticosteroids were discontinued by 6 months in all but two patients, 12 others who initially received corticosteroids were maintained on azathioprine beyond 6-months. Our observations are somewhat at odds with the general understanding that AIH-like liver injury associated with NTF improves without long term immunosuppression upon stopping NTF (30).

Our HLA analyses identified a novel allele, HLA-DRB1*11:04 to be significantly associated with NTF liver injury, in comparisons to 3 control groups. The AF of HLA-DRB1*11:04 was 4-fold higher in NTF liver injury patients than in population controls in European Americans. Interestingly, this allele was observed only in patients with liver injury due to long or intermediate exposure to NTF in our dataset.On the other hand, among 12 patients with liver injury due to short NTF exposure, we found higher prevalence of the extended haplotype of HLA-A*01:01-B*08:01-C*07:01. If confirmed in future validation studies, these HLA allele and haplotype will facilitate the early detection of individuals at high risk for liver injury due to NTF. Further, combining HLA data with the duration of NTF exposure may improve the early and confident diagnosis of NTF associated liver injury, especially in cases where recovery is incomplete after NTF is stopped.

Several different HLA alleles have been implicated in NTF liver injury in the past, including HLA-DR2 [currently known as DRB1*15 or *16], HLA-DRw6 [currently known as DRB1*13 or *14], HLA-DRB1*03:01, and HLA-DRB1*04:01. The HLA-DRB1*15:01, -DRB1*03:01, and-DRB1*04:01 alleles were found in the current cohort, but none was more frequent in the NTF cases compared to controls (e.g., the AF of DRB1*03:01 was 0.14 in NTF-related cases and 0.12 in population controls). HLA-DRw6 supertype was not present in any of the 73 subjects with NTF DILI in our cohort. Previous analyses from an international group, including DILIN, have reported two non-HLA genetic variants, namely, rs2476601 in PTPN22 and rs1363907 in ERAP2, to be significantly associated with overall DILI and amoxicillin-clavulanate liver injury, respectively (19,22). However, these non-HLA genetic variants were not overrepresented in this NTF liver injury cohort compared to population controls.

In summary, liver injury due to NTF is predominantly hepatocellular in nature and can occur after only a few days of intake, but more commonly after at least a year of exposure. Liver injury injury after long exposure to NTF typically resembles chronic hepatitis with autoantibody production, AST/ALT > 1, histology showing chronic injury and fibrosis, and frequent need for steroids. Morbidity and mortality are significant for patients with liver injury due to long term NTF therapy. The class II HLA allele DRB1*11:04 is overrepresented in patients with NTF liver injury. Monitoring ALT levels at 3-to-6-month intervals is prudent in patients on long-term NTF.

Highlights.

• There were 78 patients with nitrofurantoin liver injury (NTF DILI) in the DILIN studies in the United States from 2004-2020. Of them, liver injury followed ≤ 7 days (short term) exposure in 18, 8-364 days (intermediate) in 16, and ≥ 365 days (long term) in 44 patients.

• Compared to short and intermediate exposure cases, long-exposure NTF-DILI cases had greater frequency of AST > ALT, ANA or SMA positivity, and corticosteroid use.

• In long-term NTF-DILI, bridging fibrosis, nodularity or cirrhosis, or clinical and imaging evidence for cirrhosis were present in 38%, and massive or sub-massive necrosis in 20%.

• No one in the short-exposure group died or underwent transplantation, whereas 12% patients from intermediate and long-term exposure groups combined died or had transplantation.

• There was significant association between NTF DILI and HLA-DRB1*11:04

Abbreviations

NTF

Nitrofurantoin

DILI

Drug induced liver injury

DILIN

Drug Induced Liver Injury Network

UTI

Urinary tract infection

ALT

Alanine aminotransferase

AST

Aspartate aminotransferase

Alk P

Alkaline Phosphatase

T Bili

Total bilirubin

ANA

anti-nuclear antibody

SMA

smooth muscle antibody

HL

Highly likely

HC

Hepatocellular

CS

Cholestatic

GWAS

genome wide association study

HLA

Human leukocyte antigen

SNP

Single Nucleotide Polymorphism

OR

Odds Ratio

CI

confidence interval

MHC

Major Histocompatibility Complex

AF

Allele frequency

CF

Carriage frequency