Improved nonalcoholic steatohepatitis after 48 weeks of treatment with the PPAR-y ligand rosiglitazone
Hepatology, October 2003, Volume 38, Number 4
Brent A. Neuschwander-Tetri1,2, Elizabeth M. Brunt1,3, Kent R. Wehmeier1,2, Dana Oliver1,2, Bruce R. Bacon1,2 From the 1Saint Louis University Liver Center, Departments of 2Internal Medicine and 3Pathology, Saint Louis University School of Medicine, St. Louis, MO.
Insulin resistance (IR) commonly is associated with nonalcoholic steatohepatitis (NASH), fatty liver. To establish whether IR causes NASH, this study was undertaken to determine if improving IR would improve the histologic features that define NASH.
Thirty adults with prior biopsy evidence of NASH were enrolled to receive rosiglitazone, 4 mg twice daily for 48 weeks. This dose was chosen because it is the highest dose currently recommended for the treatment of diabetes. Three patients were treated with hydroxy-methyl-glutaryl coenzyme A reductase (HMG-CoA) inhibitors (statins) before entry and 7 patients began treatment with statins during the trial. One patient was being treated with a fibrate before study entry and one patient began treatment with a fibrate during the study.
All patients were overweight (body mass index [BMI] > 25 kg/m2) and 23% were severely obese (BMI > 35 kg/m2); 50% had impaired glucose tolerance or diabetes. Liver biopsy specimens were obtained before beginning treatment and at treatment completion. Twenty-six patients had posttreatment biopsies; of these, 22 had initial protocol liver biopsies that met published criteria for NASH on subsequent blinded evaluation.
Within this initial NASH group, the mean global necroinflammatory score significantly improved with treatment and biopsies of 10 patients (45%) no longer met published criteria for NASH after treatment. Significant improvement in hepatocellular ballooning and zone 3 perisinusoidal fibrosis also occurred. Five patients withdrew early; the 25 patients completing 48 weeks of treatment had significantly improved insulin sensitivity and mean serum alanine aminotransferase (ALT) levels (104 initially, 42 U/L at the end of treatment). Adverse effects led to withdrawal of 3 patients (10%). Weight gain occurred in 67% of patients and the median weight increase was 7.3%. Within 6 months of completing treatment, liver enzyme levels had increased to near pretreatment levels.
In conclusion, improving insulin sensitivity with rosiglitazone resulted in improved histologic markers of NASH, an observation suggesting that insulin resistance contributes to its development and that improving insulin sensitivity may be important in treating this liver disease.
Background Nonalcoholic fatty liver disease may be a common hepatic manifestation of insulin resistance (IR).1 Nonalcoholic steatohepatitis (NASH) constitutes the subset of nonalcoholic fatty liver disease most associated with progressive liver disease and can cause cirrhosis, hepatocellular carcinoma, and liver-related death. If the prevalence of NASH is increasing in parallel with the increasing prevalence of obesity and insulin resistance, effective treatment is needed to prevent an epidemic of advanced chronic liver disease related to IR.
Because of the possible causal nature of IR, therapeutic interventions aimed at improving insulin sensitivity may be one approach to treating NASH. Weight reduction and lifestyle modifications that include increased physical exertion are effective methods of improving insulin resistance. Unfortunately, these changes have proved impossible to sustain for many individuals. Pharmacologic approaches to improving insulin sensitivity include the use of ligands for the peroxisomal proliferator activated receptor- (PPAR), a nuclear transcription factor that facilitates increased insulin responsiveness.
The primary aim of this study was to determine if improving insulin sensitivity with the PPAR ligand rosiglitazone would improve the characteristic histologic features that define NASH. Rosiglitazone was chosen because of its availability at the time the study was initiated and its potent effects on improving insulin sensitivity. A secondary aim was to determine if rosiglitazone is associated with hepatotoxicity in patients with preexisting liver disease.
Another thiazolidinedione, troglitazone, has been implicated causally in a number of incidents of severe hepatotoxicity, justifying concern that rosiglitazone also could cause liver injury, especially in those with ongoing necroinflammatory changes in the liver known to occur in NASH. A small pilot trial of troglitazone undertaken before the recognition of its hepatotoxicity suggested that this approach could be beneficial in patients with NASH. The initial response of this study cohort to rosiglitazone after 24 weeks of treatment and the details of one patient who withdrew from the study because of increasing aminotransferase levels during concomitant corticosteroid therapy has been detailed in an earlier report. The purpose of this report is to describe the histologic and biochemical changes at completion of 48 weeks of treatment and to describe the biochemical changes that occurred during a 6-month posttreatment follow-up evaluation.
Five patients (17%) did not complete 48 weeks of treatment. The reasons for withdrawal and durations of treatment were as follows: bad dreams (12 weeks), transportation difficulties (12 weeks), sensation of heavy legs (16 weeks), increased ALT levels (16 weeks), and disinterest in the second biopsy (36 weeks). Details of the possible adverse events occurring before the midpoint of the study are detailed elsewhere.9 No additional adverse events led to patient withdrawal.
Two patients had liver biopsies even though they withdrew after 16 and 17 weeks of treatment, respectively. One patient who completed 48 weeks did not have a biopsy at completion.
Changes in clinical measures of the 25 patients who received 48 weeks of treatment are reported. Body mass gradually increased over the treatment period in most patients. The mean increase in BMI was 6.5% (range, –5% to 18%). Five patients lost weight during treatment. After a 24-week posttreatment follow-up period, the 22 patients in whom weight data was available had a mean BMI of 34.6 kg/m2, indicating that the weight gained during treatment was not fully lost during the follow-up period. The serum levels of ALT, aspartate aminotransferase, alkaline phosphatase, and -glutamyl transpeptidase (GGT) all decreased significantly during treatment.
The enzyme level elevations at entry improved over the first 20 weeks of treatment and remained at reduced levels until treatment was discontinued. During 24 weeks of posttreatment follow-up evaluation, these levels increased toward pretreatment values (pretreatment vs. 24 weeks posttreatment [i.e., 72 weeks], P = .15, n = 24 patients). There were no significant changes in the cholesterol and triglyceride levels during treatment or during the 24-week follow-up period. Treatment of hyperlipidemia did not alter the treatment response.
NASH is defined by the pattern of injury and the constellation of histologic abnormalities. The histologic changes of NASH in the liver and the individual components that contribute to the diagnosis were tabulated before and after treatment from the blinded review.
Significant improvements among those meeting the criteria for NASH at the beginning of treatment were observed after treatment with respect to the global grade, amount of steatosis, features of necroinflammation, hepatocellular ballooning, and zone 3 perisinusoidal fibrosis.
As noted in Table 2, along with decreased amounts of steatosis, the intra-acinar localization of steatosis also shifted. Whereas 5 patients had non-zone 3 predominant steatosis pretreatment, 11 patients had this distribution posttreatment, 6 of which were nonzonal. A nonzonal distribution of steatosis is not observed typically in NASH. The serum alkaline phosphatase and GGT levels, enzymes typically thought to be increased in cholestatic injury, also improved significantly during treatment (Fig. 2). Although the GGT level has been reported to correlate with hepatic steatosis,11 a correlation between baseline liver fat and baseline GGT level or changes in liver fat and changes in GGT level were evaluated and not identified.
Interestingly, changes also were seen by a qualitative assessment in the amount of portal inflammation compared with lobular inflammation. Portal inflammation in NASH commonly is mild12; in the posttreatment biopsies there was a shift toward greater portal inflammation that was detected by the comparison of relative amounts. Whether this may be a feature of resolving steatohepatitis is not known.
As can be noted in Table 2, ballooning improved in 11 patients and did not change in another 11; no case showed worsening of that feature of liver cell injury. Mallory's hyaline, commonly associated with higher grade but not included in the previously proposed method of grading in NASH,10 showed a shift from 12 cases with either occasional or several, to 17 cases with none, 4 cases with occasional, and 1 with several. PAS-D Kupffer cells showed a shift toward greater numbers in the posttreatment biopsies, consistent with the understanding of this histologic feature as a marker of prior cell necrosis.
The progression of hepatic fibrosis underlies the evolution of liver disease that ultimately ends in cirrhosis and its complications. Significant changes in the pattern and character of fibrosis were observed with treatment in this study. The improvement found in zone 3 perisinusoidal fibrosis was caused primarily by a shift from dense, prominent collagen bands to more delicate collagen deposition in perisinusoidal spaces in zone 3 (found in biopsies of 7 patients). Eight patients with biopsy specimens lacking any zone 3 fibrosis generally remained unchanged; those without zone 3 fibrosis had stage 0 fibrosis (by definition). One biopsy specimen without initial zone 3 perisinusoidal fibrosis increased from none to delicate (overall stage changed from 0 to 2 because of additional periportal fibrosis) and one of the remaining specimens decreased from delicate to no zone 3 fibrosis (stage changed from 1 to 0).
Statistically significant improvement was not seen in the global fibrosis score, but 8 scores improved whereas 11 remained unchanged. Together, 19 of 22 posttreatment liver biopsy specimens showed improvement or stayed the same whereas 3 biopsy specimens worsened.
Noninvasive markers of inflammation and fibrosis have been unreliable. In contrast, a strong correlation was observed between the liver/spleen density ratio as a noninvasive measure of liver fat and the amount of fat estimated histologically in both pretreatment and posttreatment biopsies.
Spleen density did not change during treatment.
Several biopsy features were tabulated but because these findings showed no changes, they are not included in Table 2. These findings were Mallory body location and character, PAS-D-stained Kupffer cell location (although the number changed as noted earlier), glycogen nuclei location, the presence and zonality of megamitochondria (as detected by light microscopy of routinely stained biopsy specimens), and ductular changes.
No significant changes were documented when prior biopsies were compared with pretreatment biopsies in steatosis, necroinflammatory grade, or fibrosis stage. A trend toward improvement in most parameters studied was noted, however; this trend almost reached significance with respect to the prestudy changes in steatosis. It could represent improvement in NASH that may occur simply with establishing a diagnosis by a physician and application of the usual standard of care. Interpretation of these results was not altered by annualizing the histologic changes using a linear model (results not shown).
An unexpected observation was that on the blinded evaluation, 7 of the 30 enrolled patients' pretreatment study entry biopsies did not fulfill the strict criteria for NASH, the details of which will be discussed in a subsequent report (in preparation). These biopsy specimens lacked the pattern of injury and sufficient steatosis and inflammation on blinded review leading to this result. Because these patients' non-NASH entry biopsy specimens could not improve, they were excluded from the histologic analysis described earlier. These patients may fall into the category sometimes referred to as type 2 NAFL,13 or steatosis with inflammation, but with insufficient necroinflammatory activity to meet criteria proposed for a diagnosis of NASH.10 Biochemically, these 7 patients exhibited evidence of necroinflammation with a mean ALT level of 53 U/L (range, 34-106 U/L). Compared with the whole cohort, the patients in this group tended to be leaner (mean BMI of 35.0 kg/m2; range, 26-56), had less insulin resistance (mean quantitative insulin sensitivity check index [QUICKI], 0.306; range, 0.278-0.325), and less liver fat by imaging (mean liver/spleen density ratio, 0.93; range, 0.60-1.23). Importantly, 5 of 7 had fibrosis: 3 had stage 3 (bridging), and 1 each were stage 2 or portal fibrosis only; 4 of the 7 also had varying amounts of zone 3 perisinusoidal fibrosis. The significance of this observation is uncertain; it may emphasize a concern that hepatic fibrosis may occur in the absence of characteristic features of NASH, or, alternatively, these biopsy changes could reflect residual fibrosis after the resolution of active necroinflammatory lesions.
Measures of IR
Measures of IR and glycemic control improved during treatment. The homeostasis model assessment (HOMA)-IR and QUICKI, 2 measures of insulin sensitivity based on the product of the fasting insulin and fasting glucose levels, significantly improved (Table 1). Although the change in the QUICKI appears relatively minor (0.294-0.320), the full range of possible values, including normal and abnormal, typically spans from about 0.25 to about 0.45, with values greater than 0.357 being associated with normal insulin sensitivity. Therefore, the change seen may be relatively substantial within this narrow range, although only 2 patients had posttreatment values greater than 0.35. The improvement in the QUICKI at 48 weeks was similar in those with abnormal glucose tolerance (median, 0.029; range, –0.006 to 0.062) as it was in those with normal glucose tolerance (median, 0.032; range, –0.020 to 0.050). Changes in the QUICKI did not correlate with changes in histologic parameters.
The postglucose challenge insulin and glucose areas under the curve also significantly improved, indicating improved insulin responsiveness in the fed state as well. The significant improvement in hemoglobin A1c levels in the entire group was driven by a significant improvement in the 7 patients with diabetes (6.8% pretreatment vs. 5.6% posttreatment, P < .01), whereas no significant change occurred in the 6 patients meeting criteria for impaired glucose tolerance but relatively normal glycemia (5.4% vs. 5.3%) or normal glucose tolerance (5.2% vs. 5.1%, n = 11). The mean hemoglobin A1c at 72 weeks was 6.0%, representing a significant rise above posttreatment levels (P < .001). The largest increase was in the 7 patients with diabetes (5.6% posttreatment compared with 7.3% at 72 weeks, P < .01), although a trivial increase was observed in the 10 patients with normal glucose tolerance as well (5.2% after treatment vs. 5.3% at 72 weeks, P < .01).
Hemoglobin levels decreased during treatment (median decrease, 0.7 g/dL; range, 0.1-3.1 g/dL); the greatest decrease was observed in the first 4 weeks of therapy, after which the change was minimal. There was no correlation between the steady weight gain and the decrease in hemoglobin level, suggesting that the change in hemoglobin level was not a simple reflection of an increase in fluid retention and expanded intravascular volume. The other major side effect of treatment was weight gain, as described in the Results section. One patient withdrew early because of bad dreams, a possible but previously unreported side effect. Two patients withdrew because of what appeared to be personal reasons (transportation difficulties and fear of a second biopsy).
Predictors of response
Because not all patients responded to rosiglitazone, clinical predictors of response would be potentially useful to clinicians so that appropriate patients for this treatment could be identified, however, correlations between specific clinical parameters such as BMI, age, sex, degree of ALT level increase, or specifics of the entry biopsy and biochemical or histologic responses were not identified.
Although IR may be caused by chronic liver disease, it also may be a significant causal factor in the development of liver disease, specifically nonalcoholic fatty liver disease. This study was undertaken to determine if treating IR in patients with NASH would improve the histologic features that define NASH and its commonly associated clinical abnormalities. Earlier trials of troglitazone suggested that it might have been useful for this purpose, and results of small trials using the other clinically available thiazolidinedione, pioglitazone, have been reported recently in abstracts.
Endogenous ligands for PPAR remain largely unknown, as are the genes differentially regulated by this class of nuclear receptors that improve insulin sensitivity. Rosiglitazone is one pharmacologic ligand for PPAR and it has been found to increase insulin sensitivity in patients with diabetes. Similar to the observations in patients with diabetes, rosiglitazone improved insulin sensitivity in patients with NASH, although posttreatment measures approaching normal were rare. Insulin sensitivity appeared to be improved equally in those with normoglycemia compared with those with impaired glucose response. This observation suggests that unappreciated insulin resistance in NASH patients with normoglycemia may be equally responsive to insulin-sensitizing agents as is the known insulin resistance in those with impaired glucose tolerance.
NASH currently is defined by the presence and pattern of a constellation of liver biopsy abnormalities in the appropriate clinical setting. Using previously defined criteria,10 the overall necroinflammatory grade and amounts of steatosis and hepatocellular injury (ballooning) improved.
However, if improvement in NASH should occur with effective intervention, how this is best measured presently is not known because the timing and nature of improvement in the histologic features need further evaluation. Treatment with rosiglitazone was associated with several histologic findings interpreted as improvement or resolution of NASH. Some of the observations have not been described previously. These included shifts in parenchymal localization of steatosis and changes in the relative amounts of portal inflammation compared with intra-acinar inflammation. These changed such that many posttreatment biopsies no longer showed characteristic morphology of NASH. Specifically, treatment was associated with a shift of inflammation from being predominantly lobular to more portal based. Whether this is a feature of improvement in NASH as it is in other forms of liver injury is not known, but emphasis on the distribution of inflammation before and after treatment may deserve further attention. Another observation that resulted from the blinded histology review was that although overall fibrosis scores for stage did not show significant improvement, the zone 3 perisinusoidal component of that score often shifted from dense deposition to more delicate deposition. This change in the nature of zone 3 fibrosis quality may represent dissolution of the collagen that was present. Whether longer treatment durations would be associated with resolution of the zone 3 fibrosis or improvement in the overall fibrosis stage cannot be determined by this study and may be a focus of subsequent studies. Similarly, the nature of zone 3 fibrosis might be worthy of detailed examination in future treatment trials. Scoring methods that do not address these changes may not adequately describe important changes associated with the resolution of NASH.
Treatment was associated clearly with improved standard noninvasive markers of necroinflammation and steatosis. The aminotransferase levels improved over the first 20 weeks of treatment and remained at lower levels until treatment was stopped, after which the ALT and aspartate aminotransferase levels increased toward pretreatment levels (Fig. 2). Surprisingly, the alkaline phosphatase and GGT levels also improved during therapy. Whether the changes in the cholestatic liver enzyme levels will prove useful in identifying patients with NASH or measuring their response to therapy will need further study.
Although this study showed improved insulin sensitivity with biochemical and histologic markers, an alternative explanation of the observed improvement is that rosiglitazone may have favorably influenced the necroinflammatory changes and fibrogenesis directly, resulting in diminished NASH by direct down-regulation of the inflammatory cascade and fibrosis. Precedent for this mechanism has been suggested by studies in animals and human cells in culture.20-25
Correlations between histologic changes and clinical or biochemical changes were sought but the small sample size prevented extensive statistical analysis. No significant correlation was found between changes in ALT level and changes in ballooning or changes in the global necroinflammatory score. This may be unexpected if increases in the ALT level are thought to directly reflect hepatocellular injury. Unfortunately, other factors are likely to determine the steady-state serum levels of the hepatocellular enzymes, a point made clear by this data. Correlations also were sought between a variety of clinical parameters and none were found. Specifically, the changes were not predicted by changes in the BMI, changes in insulin sensitivity, or changes in glycemic control. The lack of correlations could reflect the relatively small cohort size of this study or possibly the absence of any such correlation. Larger clinical trials sufficiently powered to identify clinical predictors of response will be needed to address these issues.
An important aspect of this trial was measurements of clinical parameters during the 6-month follow-up period. The liver enzyme levels and glycemic control reverted to pretreatment values during this period, indicating that the insulin-sensitizing effects of treatment were not sustained. This finding would suggest that any benefit conferred by using thiazolidinediones may last only as long as drug intervention is used.
Treatment of NASH with a PPAR-y ligand was not without untoward effects. Weight gain is a known effect of this class of drugs and was observed in this cohort of patients. The weight gained during the use of these drugs tends to be peripheral fat rather than central fat16 and therefore it may not be associated with increased risks associated with the metabolic syndrome. Nonetheless, weight gain is quite disheartening in this patient population and the counterproductive psychologic effects associated with drug-induced weight gain cannot be ignored.
In conclusion, this study provided an opportunity to evaluate the relationship of insulin resistance and the histologic features of NASH by treatment with rosiglitazone. In a blinded histologic review of prior, pretreatment, and posttreatment biopsies, steatosis amount and localization in the acini, hepatocellular ballooning, numbers of Kupffer cells, and the global necroinflammatory grade improved during treatment with rosiglitazone. Moreover, the characteristic zone 3 fibrosis of NASH appeared to have improved. Finally, the pattern of inflammation also changed from lobular to portal predominance.
This trial provides preliminary evidence that the use of a PPAR ligand should be examined as a therapeutic alternative within the context of a larger, placebo-controlled trial. It also provides proof of principal that addressing insulin resistance may be an appropriate treatment option for individuals with NASH. Exercise remains the most effective nonsurgical means of improving insulin sensitivity and should be recommended to all patients with NASH. Despite the risk for developing cirrhosis and its complications, adopting and maintaining an altered lifestyle that includes increased physical exertion is difficult or impossible for many obese people.