OBJECTIVES:

Thiazolidinediones (TZD) are recommended to be used cautiously in diabetics with mild elevations in liver enzymes and not to be used in those with alanine aminotransferase >2.5 upper limit normal (ULN). However, studies are not adequate that evaluated the risk of TZD hepatotoxicity in diabetics with elevated liver enzymes. We conducted a study to test if diabetics with elevated liver enzymes have increased risk for hepatotoxicity from rosiglitazone (only TZD available on our formulary).

METHODS:

This study consisted of two cohorts of patients prescribed rosiglitazone since January 2000. Cohort 1: 210 diabetics with elevated baseline liver enzymes (aspartate aminotransferase (AST) >40 IU/L and/or alanine aminotransferase (ALT) >35 IU/L) who received rosiglitazone, and cohort 2: 628 diabetics with normal liver enzymes who received rosiglitazone. Elevations in liver biochemistries over a 12-month period after initiating rosiglitazone were characterized into mild to moderate or severe elevations and into "Hy's rule" based on published criteria.

RESULTS:

Compared to cohort 2, patients in cohort 1 did not have higher incidence of mild to moderate (10%vs 6.6%, p= 0.2) or severe elevations (0.9%vs 0.6%, p= 0.9) in liver biochemistries. Similarly, the incidence of liver biochemistry abnormalities meeting the Hy's Rule was statistically not different between the two cohorts (0%vs 0.3%, p= 0.9). The frequency of discontinuing rosiglitazone therapy during the follow-up was similar between cohorts 1 and 2 (8.6%vs 8.1%, p= 1.0).

CONCLUSIONS:

These results suggest that diabetics with elevated baseline liver enzymes do not have a higher risk of hepatotoxicity from rosiglitazone than those with normal enzymes.

INTRODUCTION

Selective peroxisomal proliferator-activated receptor- (PPAR-) agonists such as rosiglitazone and pioglitazone play an important role in the management of Type 2 diabetes (^1,2). Two PPAR- agonists approved by FDA for clinical use in the United States include rosiglitazone and pioglitazone. These agents are widely prescribed in the United States, and combined, more than 16 million prescriptions were written in 2003 (³). Unlike the troglitazone, which has been withdrawn from the market due to its hepatotoxicity, pioglitazone and rosiglitazone are generally considered to be very safe from a hepatotoxicity standpoint (^4,5,6). However, isolated case reports of severe liver damage following their use have appeared in the literature (^{7,8,9,10,11,12}). It is recommended by the manufacturers that liver enzymes should be checked in all patients prior to prescribing these agents, and they be not used in those with increased liver enzymes (alanine aminotransferase (ALT) >2.5 ULN) (¹³). It is further recommended that the initiation of therapy in patients with mild elevations in liver enzymes should be undertaken with caution (¹³). However, the studies are limited that addressed the specific question if diabetic patients with elevated baseline liver enzymes are at higher risk for hepatotoxicity from the available TZDs than those with normal baseline liver enzymes. This is an important issue because diabetics with elevated liver enzymes have high prevalence of nonalcoholic steatohepatitis (NASH), and several recent pilot studies have suggested that TZDs may improve liver histology in patients with NASH (^14,15,16).

We conducted a study to address some uncertainties related to the use of TZDs in diabetic patients with elevated liver enzymes. The main focus of our study was to evaluate if the risk of hepatotoxicity from TZDs is higher in diabetic patients with elevated liver enzymes than those with normal liver enzymes at baseline. Three secondary objectives of this study were (¹) to compare the frequency of TZD discontinuation between patients with elevated baseline liver enzymes and normal baseline liver enzymes within 12 months after their prescription, (²) to evaluate how often practicing physicians comply with the manufacturers' recommendations about TZD use in patients with elevated baseline liver enzymes, and (³) to evaluate if TZD therapy improves liver enzymes in diabetic patients with elevated liver enzymes.

METHODS

This study was reviewed and approved by the Institutional Review Board at Indiana University School of Medicine. Data were collected from a large academic medical practice located in Indianapolis, which uses the Regenstrief Medical Record System (RMRS). The details of RMRS are discussed extensively elsewhere (¹⁷). In brief, the RMRS captures patient information from three hospitals on the Indiana University Medical Center campus (Wishard Memorial Hospital, Indiana University Hospital, and Riley Hospital for Children) and from 30 clinics scattered around the inner city of Indianapolis (¹⁷). The RMRS links patients' prescriptions and their laboratory test results using a unique patient identifier. It uniquely allows for the identification of liver biochemistry abnormalities that occur after initiating therapy with specified medications.

To address the primary objective, using the RMRS we identified two cohorts of Type 2 diabetic patients who received rosiglitazone prescription at Wishard Memorial Hospital (400 bed inner-city academic hospital) and affiliated primary care clinics. Rosiglitazone is the only TZD on our hospital formulary since 2000.

Cohort 1 consisted of diabetic individuals with elevated baseline liver enzymes who were prescribed rosiglitazone and in whom liver enzymes were available within 6 months before and 12 months after rosiglitazone was first prescribed. Cohort 2 consisted of individuals with normal baseline liver enzymes who were prescribed rosiglitazone and in whom liver enzymes were available within 6 months before and 12 months after rosiglitazone was first prescribed.

The patients with serum AST >40 IU/L or ALT >35 IU/L were considered to have elevated transaminases as the ULN for AST and ALT at our laboratory were 40 IU/L and 35 IU/L, respectively. The absence of alcohol abuse was determined based on the clinic notes, discharge summaries, and other encounter transcripts. The presence of hepatitis C was determined by the presence of anti-HCV antibodies (by ELISA or RIBA) and the presence of hepatitis B by the positivity for hepatitis B surface antigen.

The elevations in liver biochemistries during the 12-month follow-up were defined and categorized depending on the patient's baseline levels of serum aminotransferases according to the published criteria (^18,19). "Mild to moderate" elevations in liver biochemistries were defined as elevations of AST and/or ALT up to 10 times ULN in patients with normal baseline enzymes or up to 10-fold elevations from their baseline values of AST and/or ALT in patients with elevated liver enzymes at baseline (^18,19). "Severe elevations" in liver biochemistries were defined as the development of serum bilirubin >3 mg/dl (regardless of their baseline transaminases) or elevations of AST and/or ALT greater than 10 times ULN in patients with normal baseline enzymes or >10-fold elevations from their baseline values of AST and/or ALT in patients with elevated liver enzymes at baseline. Additionally, we assessed the incidence of liver biochemistry elevations meeting the criteria of "Hy's rule," defined as the development of AST or ALT >3 ULN and bilirubin >2 ULN (²⁰).

Additionally, using the RMRS prescription records, we assessed the frequency of rosiglitazone discontinuation during the 12-month follow-up period. Subjects were considered to have discontinued the use of rosiglitazone if there was no evidence of a prescription refill after their supply of rosiglitazone ran out. It previously has been shown that most patients attending Wishard Memorial Hospital and the affiliated primary care clinics fill their medicine prescriptions exclusively at the Wishard Hospital pharmacies (²¹).

In order to evaluate how often practicing physicians comply with the manufacturers' recommendations about TZD use in patients with elevated baseline liver enzymes, we have examined the proportion of patients in whom AST or ALT levels were available within the prior 6 month period before prescribing rosiglitazone therapy. Additionally, we assessed if anybody with liver enzymes >2.5 ULN was prescribed rosiglitazone during the study period.

In order to examine if rosiglitazone therapy improves liver enzymes in diabetic patients with elevated liver enzymes, we have assessed the proportion of patients in cohort 1 who normalized liver enzymes during the follow-up and compared it to another liver disease control group. This liver disease control group consisted of one to four age, gender, race, and weight-matched diabetic patients seen during the same period with elevated liver enzymes but not started on rosiglitazone (n = 636).

Statistical Analyses

Data validity procedures, database management, and statistical analyses were performed using SAS software (SAS Institute, Cary, NC). Basic descriptive statistics, including means, standard deviations (SD), ranges, and percentages were used to characterize the study patients. Chi square tests for categorical variables and Student's t-tests for continuous variables were made for univariate comparisons between the groups. A p-value <0.05 was considered statistically significant.

RESULTS

There were 210 patients in cohort 1 and 628 patients in cohort 2. In general, the demographics and clinical characteristics were comparable between the two groups (Table 1). The baseline positivity for hepatitis B or C, reported alcohol use, lipid profiles, and hemoglobin A_1C were also comparable between the two groups (Table 1). As expected, based on the study design, individuals in cohort 1 had significantly higher AST and ALT values than the ones in cohort 2. The proportion of patients who were on insulin or statins at the initiation of rosiglitazone therapy was similar between the two groups.

Table 1 - Demographics and Selected Clinical Characteristics of Patients belonging to Cohorts 1 and 2.

Full table

The course of transaminases during the 12-month follow-up period in patients belonging to cohorts 1 and 2 is shown in Table 2. Compared to cohort 2, patients belonging to cohort 1 did not have higher incidence of mild to moderate elevations (10%vs 6.6%, p= 0.2) or severe elevations in liver biochemistries (0.9%vs 0.6%, p= 0.9). No one in cohort 1 developed liver test abnormalities to the extent of Hy's rule, whereas two patients in cohort 2 exhibited liver biochemistry abnormalities meeting the Hy's rule (0%vs 0.3%, p= 0.9).

Table 2 - The Course of Liver Biochemistries and the Incidence of Outcome Variables over a 12-Month Period.

Full table

The frequency of discontinuing rosiglitazone therapy was not significantly different between cohorts 1 and 2 (8.6%vs 8.1%, p= 1.0) (Table 2).The mean duration of rosiglitazone exposure in patients belonging to cohorts 1 and 2 was similar (357 400 and 423 423 days, respectively p= 0.3). The dose of rosiglitazone prescribed was similar between cohorts 1 and 2 (4.4 1.8 vs 4.4 2 mg/day, p= 0.8). The number of patients who received a new prescription for statins during the follow-up period was comparable between the two groups (11%vs 18%, p= 0.9). Similarly, the frequency of liver biochemistry monitoring during the rosiglitazone therapy was similar between the groups (2.9 2.3 vs 2.4 2.1, p= 0.4) (Table 1).

The proportion of patients in whom liver enzymes levels were available within 6 months before initiating rosiglitazone therapy was 50% in the year 2000, 54% in the year 2001, 57% in the year 2002, and 51% in the year 2003. Interestingly, the proportion of patients who developed bilirubin >3 mg/dl during the follow-up period (0.4%vs 0.5%, respectively, p= 0.9) or in whom the rosiglitazone therapy was discontinued (9.3%vs 7.4%, respectively, p= 0.8) was similar between patients with and without baseline liver enzymes available prior to initiating rosiglitazone therapy.

Of the 210 patients belonging to cohort 1, there were 15 individuals with baseline AST and/or ALT levels >2.5 ULN, a threshold beyond which the rosiglitazone is recommended not to be prescribed. None of these 15 patients developed mild to moderate or severe elevations, or elevations up to the Hy's rule during the 12-month follow-up.

The liver enzymes normalized in 61% of cohort 1 patients during the 12-month follow-up observation period and this proportion is significantly higher than in the liver disease control group (51%, p= 0.02).

DISCUSSION

In this study, we have shown that diabetic patients with elevated baseline liver enzymes do not have higher risk of hepatotoxicity from rosiglitazone than those with normal baseline liver enzymes. To our knowledge, there is only one other study published in the literature that addressed this question (¹⁸). Lebovitz et al. analyzed the data from 5,006 diabetics who participated in phase 2/3 clinical trials with rosiglitazone (²²). Out of these 5,006 patients, 295 individuals had elevated liver enzymes at baseline (ALT between 1 and <2.5 fold ULN) whereas 4,711 diabetics had normal liver enzymes (²¹). The incidence of on-therapy ALT >3 times ULN was 1.4% in patients with elevated baseline liver enzymes, and it appeared to be significantly higher than those with normal baseline liver enzymes (0.25%, p= 0.01). The discrepancy in the results between our study and that of Lebovitz et al. is likely due to the difference in the definition of rosiglitazone hepatotoxicity. In the study by Lebovitz et al., on-therapy ALT >3 ULN was considered to be significant, regardless of baseline liver enzymes. However, in our study, the definition of hepatotoxicity depended on whether or not liver enzymes were elevated prior to starting the rosiglitazone. We believe that the baseline level of liver enzymes (rather than ULN) should be used as the reference value for defining hepatotoxicity in patients with elevated liver enzymes (¹⁸). This concept has been adopted by the NIH-funded Drug Induced Liver Injury Network (DILIN) for defining hepatotoxicity in patients with underlying liver disease (personal knowledge of the DILIN proceedings) (²³).

It is interesting that baseline liver enzymes were not checked prior to prescribing rosiglitazone in nearly half of the patients, and the risk of hepatotoxicity in this group did not appear to be different from those who had their liver enzymes checked prior to starting rosiglitazone. Contrary to the manufacturers' recommendation, rosiglitazone was prescribed for 15 patients who had baseline liver enzymes >2.5 times ULN, and they all appeared to have tolerated the rosiglitazone therapy with no evidence of hepatotoxicity. These observations suggest that further studies are needed to address the validity of the manufacturers' recommendations about rosiglitazone use in patients with elevated liver enzymes.

Our study suggests that elevated liver enzymes in diabetic patients may normalize with rosiglitazone therapy. This finding is consistent with prior observations by Lebovitz et al., who showed that liver enzymes normalized with rosiglitazone in nearly 70% of their patients with elevated baseline liver enzymes (²¹). Elevated liver enzymes in patients with diabetes are likely to represent the underlying nonalcoholic fatty liver disease, and it recently has been shown that TZDs improve hepatic histology in patients with nonalcoholic steatohepatitis (^14,15,16).

This retrospective study of electronic medical record system carries several limitations that deserve further discussion. First, our study examined the course of liver biochemistries within 12 months after starting rosiglitazone and thus it does not provide information about what happens to liver tests after that interval. However, hepatotoxicity from rosiglitazone appears to be an early rather than late event (^7,8,9). Second, we defined the rosiglitazone discontinuation based on the prescription records rather than actual documentation that it was discontinued because of side effects. As the majority of the patients receiving care at our facility fill their prescriptions exclusively at our institutional pharmacy, the data derived from the pharmacy records are generally reflective of their prescription medicinal use. Third, our sample size may not have been adequate enough to answer this question definitively. As our study was exploratory in nature and as there existed no prior data to make sample size calculations, our power analysis was not performed a priori. However, based on our study results, a minimum of 4,600 patients (2,300 patients in each group) would be required to demonstrate a difference in the incidence of severe hepatotoxicity with 80% power at a 5% significance level, and a minimum of 2,700 patients (1,350 patients in each group) would be required to demonstrate a difference in the incidence of mild to moderate hepatotoxicity. Such studies are obviously difficult to conduct and may require multiple sites and multiple databases. Finally, as rosiglitazone is the only TZD available on our formulary, we were only able to evaluate its potential for hepatotoxicity in patients with elevated baseline liver enzymes. Thus, the findings from this study may not be applicable to other TZDs available in this country. Despite these limitations, our study provides important information about the safety of rosiglitazone in diabetic patients with elevated baseline liver enzymes.

References

1. Meriden, T. Progress with thiazolidinediones in the management of Type 2 diabetes mellitus. Clin Ther 2004;26: 177–190.
2. Reusch, JE, Gadsby, R. Thiazolidinedione therapy: The benefits of aggressive and early use in type 2 diabetes. Diabetes Techno Ther 2003;5: 685–693.
3. http://www.drugtopics.com (accessed on 6/23/04).
4. Isley, WL. Hepatotoxicity of thiazolidinediones. Expert Opin Drug Safety 2003;2: 581–586.
5. Scheen, AJ. Hepatotoxicity with thiazolidinediones: Is it a class effect? Drug Safety 2001;24: 873–888.
6. Tolman, KG, Chandramouli, J. Hepatotoxicity of the thiazolidinediones. Clin Liver Dis 2003;7: 369–379.
7. Bonkovsky, HL, Azar, R, Bird, S, et al. Severe cholestatic hepatitis caused by thiazolidinediones: Risks associated with substituting rosiglitazone for troglitazone. Dig Dis Sci 2002;47: 1632–1637.
8. Forman, LM, Simmons, DA, Diamond, RH. Hepatic failure in a patient taking rosiglitazone. Ann Intern Med 2000;132: 118–121.
9. Al-Salman, J, Heider, A, Kemp, D et al. Hepatocellular injury in a patient receiving rosiglitazone: A case report. Ann Intern Med 2000;132: 121–124.
10. May, LD, Lefkowitch, JH, Kram, MT et al. Mixed hepatocellular-cholestatic liver injury after pioglitazone therapy. Ann Intern Med 2002;136: 449–452.
11. Pinto, AG, Cummings, OW, Chalasani, N. Severe but reversible cholestatic liver injury after pioglitazone therapy. [Letter]. Ann Intern Med 2002;137: 857.
12. Maeda, K. Hepatocellular injury in a patient receiving pioglitazone [Letter]. Ann Intern Med 2001;135: 306. | PubMed | ChemPort |
13. Physician Desk Reference (PDR), 58th Edition, 2004.
14. Sanyal, AJ, Contos, AJ, Sargeant, C et al. A randomized controlled pilot study of pioglitazone and vitamin E versus vitamin E for nonalcoholic steatohepatitis. [Abstract]. Hepatology 2002;26: A382.
15. Neuschwander-Tetri, BA, Brunt, EM, Wehmeier, KR et al. Improved nonalcoholic steatohepatitis after 48 weeks of treatment with the PPAR-gamma ligand rosiglitazone. Hepatology 2003;38: 1008–1017. | Article | PubMed | ISI | ChemPort |
16. Promrat, K, Lutchman, G, Uwaifo, GI et al. A pilot study of pioglitazone treatment for nonalcoholic steatohepatitis. Hepatology 2004;39: 188–196. | Article | PubMed | ISI | ChemPort |
17. McDonald, CJ, Overhage, JM, Tierney, WM et al. The Regenstrief Medical Record System: A quarter century experience. Int J Med Inform 1999;54: 225–253.
18. Chalasani, N, Aljadhey, H, Kesterson, J et al. Patients with elevated liver enzymes are not at higher risk for statin hepatotoxicity. Gastroenterology 2004;126: 1287–1292. | Article | PubMed | ChemPort |
19. Vuppalanchi, R, Teal, E, Chalasani, N. Patients with elevated baseline liver enzymes do not have higher frequency of hepatotoxicity from lovastatin than those with normal baseline liver enzymes. Am J Med Sci 2005;39: 237–242.
20. Senior, JR. Regulatory perspective. In Kaplowitz N, DeLeve LD (eds.) Drug Induced Liver Disease. Marcel Dekker, Inc: New York NY., , ;2003: 739–754.
21. Murray, MD, Loos, B, Tu, W et al. Work patterns of ambulatory care pharmacists with access to electronic guideline-based treatment suggestions. Am J Health Syst Pharm 1999;56(3):225–232.
22. Lebovitz, HE, Kreider, M, Freed, MI. Evaluation of liver function in Type 2 diabetic patients during clinical trials. Evidence that rosiglitazone does not cause hepatic dysfunction. Diabetes Care 2002;25: 815–821. | Article | PubMed | ISI | ChemPort |
23. https://dilincentral.dcri.duke.edu, accessed on December 16, 2004.

Acknowledgements

This work is in part supported through an NIH grant (UO-1 DK 065211 to N.C).