Beyond Hy's Law: Importance of Markedly Elevated Aminotransferases without Hyperbilirubinemia

Abstract

Elevated aminotransferases with concomitant jaundice in drug‐induced liver injury (DILI) are associated with adverse outcomes, but the prognostic significance of very high aminotransferases without jaundice is unclear. We examined selected outcomes of patients with DILI and markedly elevated aminotransferases but normal or near‐normal total bilirubin (TBL). We analyzed the data from the prospective Drug‐Induced Liver Injury Network (DILIN) and the Spanish DILI registries. Patients were categorized based on peak alanine or aspartate aminotransferase (ALT/AST) levels—very high (500–1,000 U/L) or towering (>1,000 U/L)—and concomitant normal (≤1.0 mg/dL) or near‐normal (1.1–2.5 mg/dL) TBL levels. Primary outcome was subsequent development of nR‐based Hy's Law (nR ≥5 with TBL >2.5 mg/dL). A total of 964 patients were analyzed (692 DILIN; 272 Spanish DILI). In both cohorts, very high aminotransferases with normal TBL were rarely followed by nR‐based Hy's Law. In DILIN, only 1.3% of patients with very high ALT and normal TBL developed nR‐based Hy's Law, whereas risk increased markedly with minimally elevated TBL (odds ratio (OR) 7.8). Towering aminotransferases conferred increased risk regardless of TBL, with the highest risk observed when accompanied by minimally elevated TBL (OR 23.6). Similar inflection points were observed in the Spanish DILI Registry, in patients with very high aminotransferases and minimally elevated TBL or towering aminotransferases with or without elevated TBL. In conclusion, patients with DILI and very high aminotransferases may tolerate liver injury well when bilirubin is normal, but even minimal bilirubin elevation or towering aminotransferases signal increased risk for poor outcomes.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Drug induced liver injury (DILI) is an important event that may derail the clinical development of a new therapeutic agent. The importance of elevated total bilirubin along with elevated aminotransferases (Hy’s law) is well understood as an important safety signal in drug development. However, the importance of elevated aminotransferases in conjunction with normal or near normal bilirubin is unclear.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This study examined selected outcomes of patients with DILI and markedly elevated aminotransferases but normal or near‐normal total bilirubin (TBL).

• WHAT THIS STUDY ADDS TO OUR KNOWLEDGE?

Patients with DILI and very high aminotransferases may tolerate liver injury well when bilirubin is normal, but even minimal bilirubin elevation or towering aminotransferases signal increased risk for poor outcomes.

• HOW THIS MIGHT CHANGE CLINICAL PHARMACOLOGY AND THERAPEUTICS?

This study findings improve our approaches to monitor liver safety during clinical drug development.

Drug‐induced liver injury (DILI) is the most common cause of acute liver failure in the United States and is a leading reason for drug development attrition and regulatory action. ¹ , ² Clinicians and regulators rely heavily on early laboratory features to assess the prognosis of a DILI event. ² , ³ Since the registration clinical trials are generally not statistically powered for detecting rare adverse events, the regulators depend on a series of signals and criteria based on liver biochemistries and other approaches to assess for DILI risk from a compound in clinical development. ² , ⁴ , ⁵ These include Temple Corollary (imbalance in alanine aminotransferase [ALT] or aspartate aminotransferase [AST] >3 times upper limit of normal [ULN] between active arm and the comparator), Hy's Law (hepatocellular injury with total bilirubin (>2 times ULN)), hyperbilirubinemia with or without other symptoms, and eDISH. ² , ⁵ While the ALT or AST >3 times ULN threshold is sensitive for identifying future risk for DILI, it is not specific and there are many examples of medications which can cause low‐grade elevations in liver enzymes but without significant risk of DILI, for example, statins.

Hy's Law is an important criterion used by the regulators and drug developers for early identification of DILI risk of a compound in development. ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ In principle, Hy's Law represents hepatocellular liver injury in conjunction with jaundice, and it has consistently been shown to predict subsequent risk of acute liver failure and liver‐related mortality. Since the initial description by the late Dr Hy Zimmerman, there have been many modifications, including the definition used by the FDA, R‐based Hy's Law, and nR‐based Hy's Law (nR‐Hy's Law). ⁴ , ⁸ , ⁹ Although not as widely used, nR‐Hy's Law offers better prediction for poor liver outcomes than other Hy's Law versions. ⁸ , ⁹ The nR‐Hy's Law is defined as nR value ≥5, T >2 ULN, and no alternative explanation for such elevated liver tests. ⁸ , ⁹

While Hy's Law captures cases with aminotransferase elevations ≥3 times ULN plus jaundice, many patients present with extremely high ALT values but normal or near‐normal total bilirubin (TBL). The prognostic significance of this presentation is uncertain. For example, in a clinical trial, if there are five cases of unexplained ALT elevations greater than 10 times ULN in the active arm with none in the placebo group but no cases of Hy's Law in either group, how should one assess the risk of serious DILI from this agent?

To address this knowledge gap, we systematically analyzed DILI cases from two large, prospective cohorts—the United States (U.S.) Drug‐Induced Liver Injury Network (DILIN) prospective study and the Spanish DILI Registry. Our objective was to examine the relationship between very high aminotransferases with normal or near TBL and subsequent development of nR‐Hy's Law and liver‐related poor outcomes.

METHODS

This study is based on post hoc analysis of data from two well‐known prospective DILI registries: the U.S. Drug Induced Liver Injury Network (DILIN) Prospective Study and the Spanish DILI Registry. The design of these two long‐standing studies and their primary results have been published previously. ¹⁰ , ¹¹ , ¹² , ¹³ , ¹⁴ The participants enrolled from September 2004 to August 2024 into the DILIN prospective study and 1994 to October 2024 into the Spanish DILI Registry were eligible for inclusion in this analysis. The study protocols of both registries were approved by the local ethics committee, and all subjects gave written informed consent.

We included only cases with a causality score of probable, highly likely, or definite, following a structured causality adjudication process as described previously. ¹¹ , ¹⁴ We defined ALT or AST 500 to ≤1,000 U/L as very high, whereas ALT or AST >1,000 U/L as towering, respectively. Our primary analysis focused on peak ALT or AST, and we then constructed five groups based on their peak ALT or AST levels in U/L and serum TBL in mg/dL: Category 1: peak ALT or AST <500 U/L with TBL ≤1 mg/dL up to peak ALT or AST values; Category 2: very high ALT or AST with TBL ≤1 mg/dL up to peak ALT or AST values; Category 3: very high ALT or AST with TBL 1.1 to ≤2.5 mg/dL up to peak ALT or AST values; Category 4: towering ALT or AST with TBL ≤1 mg/dL up to peak ALT or AST values; and Category 5: towering ALT or AST with TBL 1.1 to ≤2.5 mg/dL up to peak ALT or AST values. Participants with TBL >2.5 mg/dL prior to peak ALT or AST value were excluded.

Our primary outcome of interest was the subsequent meeting of nR‐Hy's Law criteria: nR value ≥5 along with total bilirubin >2.5 mg/dL. Other outcomes of interest included liver‐related mortality, liver transplantation, and chronic DILI. Chronic DILI was defined at 6 months after DILI onset in the DILIN Prospective Study, whereas at 12 months after DILI onset in the Spanish DILI Registry. As an ancillary analysis, we also examined the significance of very high and towering ALT or AST values with normal or near‐normal TBL at DILI onset.

Statistical analysis

The DILIN and Spanish DILI Registry cohorts were investigated separately as there were some differences in the causality adjudication process and differing enrollment periods. To characterize the study groups and outcomes, quantitative data were presented using mean and standard deviation (SD) or median, while categorical data were described using frequency distributions. Logistic regression analyses were conducted to determine the risk of developing nR‐Hy's Law in Categories 2–5 compared to Category 1. Odds ratios (OR) and 95% confidence intervals (CI) were reported. All statistical analyses were conducted using Stata version 18 (Stata Corporation). A two‐sided P‐value lower than 0.05 was deemed statistically significant.

RESULTS

During the study periods, there were 692 individuals from the DILIN Prospective Study and 272 individuals from the Spanish DILI Registry with well characterized DILI and meeting predefined eligibility criteria.

Very high and towering peak ALT with normal or near‐normal TBL

In the DILIN cohort, there were 210 participants in Category 1, 157 participants in Category 2, 121 in Category 3, 67 in Category 4, and 137 in Category 5. Their selected demographic and clinical characteristics are shown in Table 1 . Top 5 agent classes and top 10 single causative agents are described in Table S1 . There were no participants in Category 1 who subsequently met nR‐Hy's Law whereas only two out of 157 participants in Category 2 met nR‐Hy's Law. Category 3 appeared to be an inflection point in terms of risk for the subsequent development of nR‐Hy's Law. Since there were no instances of nR‐Hy's Law in Category 1, we used Category 2 as the reference, which produced OR of 7.8 (95% CI: 1.7–35.7), 7.6 (95% CI: 1.5–38.8), and 23.6 (95% CI: 5.5–100.7) for Categories 3, 4, and 5, respectively. There were too few liver‐related mortality or liver transplant events to detect differences across the five categories.

Table 1.

Prognostic significance of high peak ALT with normal or minimally elevated total bilirubin – DILIN

	Category 1	Category 2	Category 3	Category 4	Category 5
	Peak – ALT <500 and total bilirubin ≤1 mg up to peak ALT (N = 210)	Peak – ALT 500 to ≤1,000 but total bilirubin ≤1 mg up to peak ALT (N = 157)	Peak – ALT 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg up to peak ALT (N = 121)	Peak – ALT >1,000 but total bilirubin ≤1 mg up to peak ALT (N = 67)	Peak – ALT >1,000 but total bilirubin 1.1 to ≤2.5 mg up to peak ALT (N = 137)
Age (years), mean ± SD	48 ± 18	50 ± 15	53 ± 16	41 ± 17	48 ± 17
Females, %	70	69	69	69	61
BMI (kg/m2), mean ± SD	26.3 ± 6.3	27.5 ± 6.4	27.7 ± 6.3	27.3 ± 6.0	27.8 ± 6.9
Diabetes, %	23.8	23.6	28.1	13.4	20.4
Eosinophilia, %	15.5	9.7	13.3	20.3	16.3
Latency (days), median	68	55	47	50	46
Peak ALT (U/I), mean ± SD	308 ± 107	691 ± 134	700 ± 144	1,552 ± 1,042	2,038 ± 1,487
Latency from DILI onset to peak ALT (median, days)	9	7	6	3	3
AST at peak ALT (U/I), mean ± SD	250 ± 176	470 ± 246	488 ± 281	1,171 ± 1,521	1,687 ± 2,519
TBL at peak ALT (mg/dL), mean ± SD	0.6 ± 0.2	0.6 ± 0.2	1.5 ± 0.5	0.7 ± 0.2	1.6 ± 0.4
Alk P at peak ALT (U/I), mean ± SD	255 ± 311	186 ± 158	301 ± 273	190 ± 102	206 ± 143
Alk P >2 ULN at time of peak ALT n (%)	58 (27.6)	30 (19.1)	50 (41.3)	12 (17.9)	36 (26.3)
R‐value at peak ALT, mean ± SD	5.3 ± 3.5	12.3 ± 5.6	9.4 ± 5.8	26.6 ± 21.8	33.1 ± 29.2
nR‐Hy's Law, n (%)	0	2 (1.3)	11 (9.1)	6 (9.0)	32 (23.4)
nR‐Hy's Law, OR (95% CI)	No observations	1 (ref.)	7.8 (1.7–35.7)	7.6 (1.5–38.8)	23.6 (5.5–100.7)
Time from peak ALT to nR‐Hy's law (median, days)	NA	86	3	2	0
Liver‐related mortality, %	0.5	0	2.5	0	0.7
Liver transplant, %	0	0	0.8	1.5	0.7

ALT, alanine aminotransferase; AST, aspartate aminotransferase; AlkP, alkaline phosphatase; BMI, body mass index; CI, confidence interval; DILI, drug‐induced liver injury; DILIN, Drug Induced Liver Injury Network; OR, odds ratio; TBL, total bilirubin.

In the Spanish DILI Registry, there were 121 participants in Category 1, 33 participants in Category 2, 21 in Category 3, 16 in Category 4, and 21 in Category 5. The most frequent agent classes and single causative agents in this cohort are described in Table S2 . The rates of nR‐Hy's Law development across five Categories are shown in Table 2 . Similar to the DILIN cohort, Category 3 appeared to be the inflection point in terms of risk for subsequently meeting nR‐Hy's Law. However, no cases evolved into poor outcomes, that is, liver‐related death or liver transplantation.

Table 2.

Prognostic significance of high peak ALT with normal or minimally elevated total bilirubin—Spanish DILI Registry

	Category 1	Category 2	Category 3	Category 4	Category 5
	Peak – ALT <500 and total bilirubin ≤1 mg up to peak ALT (n = 121)	Peak – ALT 500 to ≤1,000 but total bilirubin ≤1 mg up to peak ALT (n = 33)	Peak – ALT 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg up to peak ALT (n = 21)	Peak – ALT >1,000 but total bilirubin ≤1 mg up to peak ALT (n = 16)	Peak – ALT >1,000 but total bilirubin 1.1 to ≤2.5 mg up to peak ALT (n = 22)
Age (years), mean ± SD	50 ± 15	51 ± 19	48 ± 19	42 ± 20	47 ± 16
Females, %	59	56	57	50	45
BMI (kg/m2), mean ± SD	26 ± 3.6	27 ± 4.1	25 ± 3.8	24 ± 4.5	26 ± 4.1
Diabetes, %	11.7	12.1	4.8	12.5	13.6
Eosinophilia, %	20.3	12.5	33.3	15.4	25.0
Latency (days), median	30	27	15	16	37
Peak ALT (U/I), mean ± SD	231 ± 104	698 ± 140	699 ± 168	1,908 ± 978	1,770 ± 726
Latency from DILI onset to peak ALT (days), mean ± SD	23 ± 125	5 ± 14	1 ± 3	1 ± 3	2 ± 10
AST at peak ALT (U/I), mean ± SD	146 ± 114	430 ± 302	504 ± 301	1,542 ± 1,194	1,316 ± 1,244
TBL at peak ALT (mg/dL), mean ± SD	0.6 ± 0.2	0.7 ± 0.2	1.6 ± 0.4	0.7 ± 0.2	1.7 ± 0.5
Alk P at peak ALT (U/I), mean ± SD	283 ± 366	201 ± 154	243 ± 117	383 ± 496	257 ± 137
Alk P> 2 ULN at the time of peak ALT, n (%)	22 (18.2)	5 (15.2)	5 (23.8)	4 (25.0)	1 (4.6)
R‐value at peak ALT, mean ± SD	7.6 ± 8.5	20.2 ± 12.3	13.4 ± 5.6	44.7 ± 39.5	36.8 ± 22.3
nR‐Hy's Law, n (%)	0 (0)	0 (0)	1 (4.8)	0 (0)	2 (9.1)
nR‐Hy's Law, OR (95% CI)	–	–	0.5 (0.04–5.7)	–	–
Time from peak ALT to nR‐Hy's Law (days), mean ± SD	–	–	7	–	26 ± 35
Liver‐related mortality, %	0	0	0	0	0
Liver transplant, %	0	0	0	0	0

ALT, alanine aminotransferase; AST, aspartate aminotransferase; AlkP, alkaline phosphatase; BMI, body mass index; CI, confidence interval; OR, odds ratio; TBL, total bilirubin; ULN, upper limit of normal.

Very high and towering peak AST with normal or near‐normal TBL

In the DILIN cohort, there were 307 participants in Category 1, 96 participants in Category 2, 104 in Category 3, 36 in Category 4, and 89 in Category 5. Their selected demographic and clinical characteristics are shown in Table 3 . Top 5 agent classes and top 10 single causative agents are described in Table S3 . Once again, Category 3 was the inflection point for the risk of subsequent nR‐Hy's Law. The incidence of meeting nR‐Hy's Law in Category 2 was not different from Category 1 (OR 0.8 [95% CI: 0.1–7.2]), whereas incidence in Category 3 (OR 9.0 [95% CI: 2.8–28.8]), Category 4 (OR 12.2 [95% CI: 3.1–47.9]), and Category 5 (OR 51.5 [95% CI: 17.6–150.5]) was significantly higher. Likewise, in the Spanish DILI Registry cases, we observed a similar inflection point at Category 3 (Table 4 ). The most frequent agent classes and single causative agents in this latter cohort are given in Table S4 .

Table 3.

Prognostic significance of high peak AST with normal or minimally elevated total bilirubin – DILIN

	Category 1	Category 2	Category 3	Category 4	Category 5
	Peak– AST <500 and total bilirubin ≤1 mg at any time up to peak AST (N = 307)	Peak – AST 500 to ≤1,000 but total bilirubin ≤1 mg at any time up to peak AST (n = 96)	Peak – AST 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg at any time up to peak AST (n = 104)	Peak – AST >1,000 but total bilirubin ≤1 mg at any time up to peak AST (n = 36)	Peak – AST >1,000 but total bilirubin from 1.1 to ≤2.5 mg at any time up to peak AST (n = 89)
Age (years), mean ± SD	48 ± 17	49.6 (16.59)	53.3 (15.81)	42.6 (16.87)	45.8 (17.18)
Females, %	68	70	66	81	66
BMI (kg/m2), mean ± SD	26.9 ± 6.2	27.2 ± 6.8	28.0 ± 6.6	27.2 ± 6.3	26.7 ± 6.9
Diabetes, %	22.1	26.0	21.2	22.2	24.7
Eosinophilia, %	15.2	14.9	17.6	11.4	10.2
Latency (days), median	60	51	57	32	41
Peak AST (U/I), mean ± SD	463 ± 249	882 ± 436	1,005 ± 522	1,685 ± 1,466	2,393 ± 1,781
Latency from DILI onset to peak AST (days), median	9	3	6	1	2
ALT at peak AST (U/I), mean ± SD	446 ± 244	848 ± 430	959 ± 495	1,634 ± 1,463	2,247 ± 1,678
TBL at peak AST (mg/dL), mean ± SD	0.6 ± 0.2	0.6 ± 0.2	1.5 ± 0.4	0.7 ± 0.3	1.7 ± 0.4
Alk P at peak AST (U/I), mean ± SD	226 ± 269	205 ± 194	241 ± 206	208 ± 139	218 ± 151
Alk P >2 ULN at the time of peak AST, n (%)	72 (23.5)	21 (21.9)	36 (34.6)	8 (22.2)	25 (28.1)
R‐value at peak AST, mean ± SD	8.2 ± 6.1	14.4 ± 10.1	15.3 ± 12.2	26.9 ± 30.2	35.6 ± 32.6
nR‐Hy's Law, n (%)	4 (1.3)	1 (1.0)	11 (10.6)	5 (13.9)	36 (40.4)
nR‐Hy's Law, OR (95% CI)	1 (ref.)	0.8 (0.09‐7.22)	8.96 (2.79‐28.8)	12.22 (3.1‐47.9)	51.453 (17.59‐150.5)
Time from peak AST to nR‐Hy's Law (days), mean ± SD	86	113	0	3	1
Liver‐related mortality, %	0.3	0	2.9	0	2.3
Liver transplant, %	0	0	0	0	1.1

AST, aspartate aminotransferase; AlkP, alkaline phosphatase; ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; DILI, drug‐induced liver injury; DILIN, Drug Induced Liver Injury Network; OR, odds ratio; TBL, total bilirubin.

Table 4.

Prognostic significance of high peak AST with normal or minimally elevated total bilirubin – Spanish DILI Registry

	Category 1	Category 2	Category 3	Category 4	Category 5
	Peak – AST <500 and total bilirubin ≤1 mg up to peak AST (n = 141)	Peak – AST 500 to ≤1,000 but total bilirubin ≤1 mg up to peak AST (n = 14)	Peak – AST 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg up to peak AST (n = 13)	Peak – AST >1,000 but total bilirubin ≤1 mg up to peak AST (n = 12)	Peak – AST >1,000 but total bilirubin from 1.1 to ≤2.5 mg up to peak AST (n = 15)
Age (years), mean ± SD	50 ± 16	51 ± 19	57 ± 21	43 ± 22	46 ± 17
Females, %	57	62	31	58	33
BMI (kg/m2), mean ± SD	26 ± 3.7	27 ± 5.1	30 ± 4.4	25 ± 4.3	26 ± 4.3
Diabetes, %	11.4	7.1	23.1	16.7	0
Eosinophilia, %	16.8	21.4	38.5	22.2	41.7
Latency (days), median	30	30	37	12	11
Peak AST (U/I), mean ± SD	166 ± 106	682 ± 145	740 ± 137	1,949 ± 1,155	1,813 ± 1,266
Latency from DILI onset to peak AST (days), mean ± SD	7 ± 30	2 ± 7	0 ± 0	0 ± 0	3 ± 13
ALT at peak AST (U/I), mean ± SD	309 ± 232	782 ± 305	1,065 ± 615	2,083 ± 1,079	1,726 ± 998
TBL at peak AST (mg/dL), mean ± SD	0.6 ± 0.2	0.8 ± 0.2	1.6 ± 0.4	0.7 ± 0.2	1.8 ± 0.5
Alk P at peak AST (U/I), mean ± SD	262 ± 336	211 ± 212	360 ± 434	512 ± 566	319 ± 234
Alk P > 2 ULN at the time of peak AST, n (%)	25 (17.7)	1 (7.1)	3 (23.1)	4 (33.3)	3 (20.0)
R‐value at peak AST, mean ± SD	9.6 ± 10.2	22.0 ± 11.9	23.6 ± 19.0	52.1 ± 45.4	34.2 ± 29.1
nR‐Hy's Law, n (%)	0 (0)	0 (0)	1 (7.7)	0 (0)	0 (0)
nR‐Hy's Law, OR (95% CI)	–	–	–	–	–
Time from peak AST to nR‐Hy's Law (days), mean ± SD	–	–	–	–	–
Liver‐related mortality, %	0	0	0	0	6.7a
Liver transplant, %	0	0	0	0	0

AST, aspartate aminotransferase; AlkP, alkaline phosphatase; ALT, alanine aminotransferase; BMI, body mass index; CI, confidence interval; OR, odds ratio; TBL, total bilirubin; ULN, upper limit of normal.

^a Liver‐related death with cholestatic damage.

Very high and towering ALT or AST at onset with normal or near‐normal TBL

The prognostic significance of very high and towering ALT or AST with normal or near‐normal TBL at DILI onset in the DILIN Prospective Study is shown in Table 5 . Similar to our observation with peak ALT or AST, there once again appeared to be a transition point from Category 3 for the risk of meeting nR‐Hy's Law subsequently. The incidence of nR‐Hy's Law cases in Category 2 was not statistically different from Category 1, either for ALT or AST based assessments. However, the risk of nR‐Hy's Law cases for Categories 3–5 was significantly higher for both ALT and AST assessments. The corresponding data from the Spanish Registry is given in Table 6 . The Spanish DILI Registry had numerical resemblance to the DILIN in terms of increasingly higher risk moving from lower Categories (1 or 2) to higher Categories (3 and 5), but not the same statistical significance likely due to smaller sample size.

Table 5.

Prognostic significance of high onset ALT and AST with normal or minimally elevated total bilirubin – DILIN

	Category 1	Category 2	Category 3	Category 4	Category 5
	Onset – ALT < 500 and total bilirubin ≤1 mg (n = 375)	Onset – ALT 500 to ≤1,000 but total bilirubin ≤1 mg (n = 145)	Onset – ALT 500 to ≤1,000 but total bilirubin from 1.1. to ≤2.5 mg (n = 98)	Onset – ALT >1,000 but total bilirubin ≤1 mg (n = 54)	Onset – ALT >1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 95)
nR‐Hy's Law, n (%)	50 (13.3)	12 (8.3)	21 (21.4)	14 (25.9)	50 (52.6)
nR‐Hy's Law OR (95% CI)	1 (ref.)	0.6 (0.3–1.1)	1.8 (1.0–3.1)	2.3 (1.2–4.5)	7.2 (4.4–11.9)

	Category 1	Category 2	Category 3	Category 4	Category 5
	Onset – AST < 500 and total bilirubin ≤1 mg (n = 461)	Onset – AST 500 to ≤1,000 but total bilirubin ≤1 mg (n = 85)	Onset – AST 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 89)	Onset – AST >1,000 but total bilirubin ≤1 mg (n = 30)	Onset – AST >1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 64)
nR‐Hy's Law, n (%)	56 (12.2)	13 (15.3)	34 (38.2)	8 (26.7)	34 (53.1)
nR‐Hy's Law OR (95% CI)	1 (ref.)	1.3 (0.7–2.5)	4.5 (2.7–7.5)	2.6 (1.1–6.2)	8.2 (4.7–14.4)

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CI, confidence interval; DILIN, Drug Induced Liver Injury Network; OR, odds ratio.

Table 6.

Prognostic significance of high onset ALT and AST with normal or minimally elevated total bilirubin – Spanish DILI Registry

	Category 1	Category 2	Category 3	Category 4	Category 5
	Onset – ALT < 500 and total bilirubin ≤1 mg (n = 142)	Onset – ALT 500 to ≤1,000 but total bilirubin ≤1 mg (n = 36)	Onset – ALT 500 to ≤1,000 but total bilirubin from 1.1. to ≤2.5 mg (n = 22)	Onset – ALT >1,000 but total bilirubin ≤1 mg (n = 17)	Onset – ALT >1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 24)
nR‐Hy's Law, n (%)	2 (1.4)	2 (5.6)	2 (9.1)	1 (5.9)	3 (12.5)
nR‐Hy's Law OR (95% CI)	1 (ref.)	4.1 (0.6–30.3)	7.0 (0.9–52.5)	4.4 (0.4–51.0)	10.0 (1.6–63.4)

	Category 1	Category 2	Category 3	Category 4	Category 5
	Onset – AST < 500 and total bilirubin ≤1 mg (n = 167)	Onset – AST 500 to ≤1,000 but total bilirubin ≤1 mg (n = 15)	Onset – AST 500 to ≤1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 16)	Onset – AST >1,000 but total bilirubin ≤1 mg (n = 13)	Onset – AST >1,000 but total bilirubin from 1.1 to ≤2.5 mg (n = 16)
nR‐Hy's Law, n (%)	3 (1.8)	2 (13.3)	4 (25.0)	0 (0)	0 (0)
nR‐Hy's Law OR (95% CI)	1 (ref.)	8.4 (1.3–54.9)	18.2 (3.7–90.9)	NA	NA

ALT, alanine aminotransferase; AST, aspartate aminotransferase; CI, confidence interval; NA, not available (odds ratio could not be calculated); OR, odds ratio.

DISCUSSION

It is well‐known that DILI patients with hepatocellular injury (predominant aminotransferase elevation) and jaundice have nearly 10% risk of acute liver failure, ⁵ , ⁶ , ⁷ , ⁸ , ⁹ but the risk of such adverse liver outcomes in DILI patients presenting with high aminotransferases but no jaundice is not well understood. Our study addresses this knowledge gap. Our main observations are that patients with DILI and very high ALT or AST (500 to 1,000 U/L) tolerate their liver injury if their TBL is within normal range. However, if their TBL is even minimally elevated, then they are at higher risk for poor liver outcomes. In addition, patients with DILI and towering ALT or AST (>1,000 U/L) are at higher risk for poor liver outcomes, regardless of TBL level. Within the subgroup of patients with towering peak ALT or AST, not unexpectedly, participants with minimal elevation in TBL exhibited higher risk for meeting nR‐Hy's Law than those with normal TBL. Collectively, our study expands our understanding of the importance of markedly high aminotransferases in the setting of normal or modestly elevated TBL.

Our observations are important for regulators, drug developers, and clinicians. Our analyses based on peak ALT or AST are more relevant for the regulators since they generally receive the reports of suspected DILI cases after their occurrence. Since the clinical trials are not statistically powered for detecting rare adverse drug reactions (ADRs) such as DILI, the regulators and drug developers search for hepatic safety signals using surrogates such as Hy's Law. In this study, we identified two other surrogates, towering aminotransferases regardless of TBL levels or very high aminotransferases with minimally elevated TBL, for estimating poor liver outcomes. Our observations from the onset ALT or AST analyses are useful for clinicians and clinical trialists who evaluate DILI in real time. They also suggest that very high aminotransferases, when associated with even minimally elevated TBL or towering aminotransferases regardless of TBL, should raise concern for potential poor outcomes and thus should be followed more closely.

In terms of relative value of AST vs. ALT to identify nR‐based Hy's Law cases, we found a higher value for ALT compared with AST. For example, in the DILIN cohort, out of 51 nR‐based Hy's Law cases, 45 were picked up by ALT, with only an additional six cases identified by including AST in the definition. However, in the same cohort, out of 51 nR‐based Hy's Law cases, 38 were picked up by AST, with 13 more cases identified by including ALT in the definition.

Our study has notable strengths but is also unable to assess several issues. Our analysis addresses an important knowledge gap in predicting poor outcomes in DILI and is based on two well‐known prospective cohorts of DILI with liver‐related outcomes carefully collected. The observations we made on the DILIN cohort were similar in the Spanish DILI Registry and thus providing confidence to their reliability and generalizability. However, while we had nearly 1,000 DILI patients with two cohorts combined, we were unable to evaluate the relationship between very high and towering aminotransferase levels and more critical liver outcomes such as liver‐related death or liver transplantation because such events were very rare in these populations. Nonetheless, we examined their relationship with subsequent risk of meeting nR‐Hy's Law which is a validated predictor of poor liver outcomes in hepatocellular DILI. Another omission is the lack of finer TBL subcategorization, for example, ULN to <1.5 times ULN, 1.5–2 times ULN, because it would have increased analysis complexity and decreased statistical power. Finally, our study is not able to offer any mechanistic clues to why only some individuals with towering aminotransferases develop nR‐based Hy's Law whereas others tolerate the liver injury well without consequences. Notwithstanding these limitations, we believe our study adds important knowledge to our understanding of predictors of poor liver outcomes in patients with DILI.

In summary, patients with DILI who have very high aminotransferase levels, but normal TBL appear not to be at excessive risk for poor liver outcomes. However, very high aminotransferases when accompanied by even minimally increased TBL, or towering aminotransferases regardless of TBL levels, are potentially associated with a higher risk of poor liver outcomes.

FUNDING

Research reported in this publication was supported by the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) of the National Institutes of Health under award numbers U01DK065211 (Indiana University), U01DK065184 (University of Michigan), U01DK065201 (University of North Carolina‐Chapel Hill), U01DK083020 (University of Southern California), U01DK083027 (Thomas Jefferson University/Albert Einstein Medical Center), U01DK100928 (Icahn School of Medicine at Mount Sinai), and U24DK065176 (Duke University). Additional support is provided by the intramural programs of the NIDDK and National Cancer Institute. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This study was also funded by grants of Instituto de Salud Carlos III (ISCIII), cofounded by Fondo Europeo de Desarrollo Regional—FEDER, cofounded by European Union (PID2022‐140169OB‐C21, PT23/00137, PPRO‐CTS649‐G‐2023, PPRO‐CTS1032‐G‐2023), by CIBERehd (grant number EHD25PI01), and by the Agencia Española de Medicamentos y Productos Sanitarios. CIBERehd and Plataforma de Investigación Clinica are funded by ISCIII. IAA holds a Miguel Servet contract funded by ISCIII (CP25/00033).

CONFLICT OF INTEREST

Dr Chalasani reports research grants to his institution from the National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), National Institute on Alcohol Abuse and Alcoholism (NIAAA), National Cancer Institute (NCI), Exact Sciences, Boehringer Ingelheim, and Madrigal Pharmaceuticals, for which he serves as principal investigator. He reports current participation on advisory boards or data safety monitoring boards for Madrigal, GSK, Zydus, Biomea Fusion, Chugai, and Insitro related to metabolic liver disease and drug safety. He reports prior advisory roles (preceding 3 years) with Eccogene, Altimmune, Ipsen, Pfizer, and Merck, which are no longer active. He reports equity interests in Heligenics, a drug discovery start‐up company in which his family has invested and for which he serves on the board, and in Avant Santé, a clinical research organization start‐up company for which he serves on the board and may receive stock options. Dr Vuppalanchi reports research grants to his institution from Eli Lilly, GSK, Takeda, and Zydus Therapeutics; prior institutional research support from Galectin Therapeutics, Kowa, and Gilead, which are no longer active; and consulting fees from Agios and Takeda. He reports participation on advisory boards or data safety monitoring boards for Fortrea, Medpace, Thermo Fisher, Madrigal Pharmaceuticals, Regeneron, 89Bio, GSK, and Cogent Biosciences, with payments made to him; and prior advisory roles with Intercept Pharmaceuticals and Akero Therapeutics, which are no longer active. He reports stock ownership in Viking Therapeutics and Pfizer. Dr Ghabril reports research support to his institution from the NIH Drug‐Induced Liver Injury Network and the NIH Functional Assessment in Liver Transplantation Study, with salary support for protected research effort. He reports participation on advisory boards for BioCryst, Zydus, Gilead, CymaBay, and Akero, with payments made to him, and unpaid participation with the Alzheimer's Clinical Trial Consortium. He reports a leadership role as Chair of the American Board of Internal Medicine Transplant Hepatology Approval Committee, for which he receives compensation. No other disclosures were reported. Drs Hayashi, Barnhart, Kamusiime, Ismael Alvarez‐Alvarez, M. Isabel Lucena, and Hao Niu declare no conflicts of interest for this paper. Dr Andrade reports current participation on adjudication committees or data safety monitoring boards for Abbott, Bayer, BMS, Boehringer Ingelheim, ENYO, GSK, iQure, Inventiva, Johnson & Johnson, Kainos Medicine, Roche, Ryvu Therapeutics, Sanofi. The content of this publication is solely the responsibility of the authors and does not necessarily represent the official views of the Food and Drug Administration.

AUTHOR CONTRIBUTIONS

Conceptualization: NC, PH, HB, RJA, MIL. Data generation: NC, HB, RK, RJA, MIL. Data analysis & interpretation: NC, HB, RK, RJA, MIL, PH, IAA, HN. Manuscript preparation: NC, PH, HB, RJA, MIL, IAA, HN. Manuscript critical review: All authors.

Supporting information

Table S1.