Honest Criticisms and Limitations of the PIVENS Trial

At a glance

| Field | Detail | |---|---| | N | 247 (randomized) | | Intervention | Pioglitazone 30 mg/day or vitamin E 800 IU/day | | Comparator | Placebo | | Duration | 96 weeks | | Primary endpoint | Improvement in NASH histology (composite score requiring ≥2-point drop in NAS without worsening fibrosis) | | Key result | Vitamin E met the primary endpoint vs placebo (42% vs 19%, p = 0.001); pioglitazone did not (34% vs 19%, p = 0.04 against a pre-specified threshold of p <0.025) |

The Population Problem: Who Was Actually Studied

The single most consequential design decision in PIVENS was the exclusion of patients with diabetes. NASH and type 2 diabetes overlap heavily. Estimates from large cohort studies place the prevalence of NASH among type 2 diabetics at 37% or higher, and diabetics carry the greatest risk of fibrosis progression. By excluding this group entirely, the trial studied a population that does not reflect the majority of patients whom hepatologists and endocrinologists treat for NASH in clinic.

This was not an oversight. The investigators reasoned that pioglitazone's glucose-lowering effects would confound metabolic endpoints in diabetics and complicate blinding. The logic holds from a trial-design standpoint. From a clinical-translation standpoint, it means that every prescription of pioglitazone or vitamin E for a diabetic NASH patient relies on extrapolation rather than direct evidence from this trial.

Additional enrollment filters narrowed the study population further. Patients with cirrhosis were excluded. So were those consuming more than 20 g of alcohol daily for women or 30 g for men. The result was a relatively young cohort (mean age ~46) with moderate disease severity and minimal comorbidity, a group that may respond differently to treatment than older, sicker patients with overlapping metabolic conditions.

Statistical Design: Why Pioglitazone "Failed"

Pioglitazone's result deserves careful reading. The drug achieved a 34% response rate on the primary endpoint compared to 19% for placebo, yielding a nominal p-value of 0.04. In most trials, p <0.05 would be reported as statistically significant. In PIVENS, it was not.

The trial used a Hochberg adjustment to control the family-wise error rate across two active comparisons against a single placebo arm. The pre-specified significance threshold for each comparison was p <0.025. Pioglitazone cleared 0.05 but not 0.025.

This is a legitimate statistical safeguard against false positives when running multiple comparisons. It is also the reason many clinicians misunderstand the PIVENS result. Pioglitazone did not fail because it showed no effect. It failed because the trial was not powered to detect its effect size at the stricter alpha level. The AASLD practice guidance from 2023 acknowledges this nuance, recommending pioglitazone for biopsy-proven NASH regardless of diabetes status, partly because of supportive secondary endpoints from PIVENS and partly because of confirmating data from other trials.

Pioglitazone actually outperformed vitamin E on resolution of NASH (47% vs 36%), a key secondary endpoint. The composite primary endpoint weighted ballooning, lobular inflammation, and steatosis equally in the NAS score, a weighting that does not necessarily reflect clinical importance. Fibrosis, the histological feature most strongly associated with liver-related mortality, was not part of the primary endpoint at all.

The Composite Endpoint and What It Misses

The NAS-based composite endpoint requires a ≥2-point improvement without worsening of fibrosis. Several problems follow from this design.

First, the NAS was developed as a tool for grading NASH severity in research settings, not as a treatment-response metric. Using it as a primary endpoint assumes that changes in NAS translate to meaningful clinical benefit, an assumption that was not validated at the time of trial design and remains debated.

Second, the "without worsening fibrosis" guardrail sounds protective but sets a low bar. A patient could show a 2-point NAS improvement while fibrosis remained completely unchanged, and this would count as a success. The PIVENS publication itself reported that neither vitamin E nor pioglitazone produced statistically significant fibrosis improvement over placebo. For a disease whose morbidity is driven primarily by fibrosis progression to cirrhosis, this gap matters.

Third, the trial captured no hard clinical outcomes: no hepatic decompensation events, no hepatocellular carcinoma incidence, no liver-related mortality, and no all-cause mortality differences. At 96 weeks with 247 patients, it was never designed to detect these. The absence of such data means that the clinical significance of NAS improvement in PIVENS remains inferential.

Duration: 96 Weeks Is Not Long Enough

NASH is a disease measured in decades. Fibrosis progresses slowly. Cirrhosis develops over 10 to 20 years in many patients. The PIVENS trial ran for 96 weeks, just under two years, followed by a 24-week post-treatment observation period.

During the post-treatment phase, ALT levels rose in both active-treatment groups, suggesting that biochemical improvements were treatment-dependent rather than durable. This is particularly relevant for vitamin E, which lacks a clear mechanistic reason to produce lasting histological change after discontinuation. The trial did not include follow-up biopsies after washout, so the durability of histological gains remains unknown.

For pioglitazone, the weight gain observed during treatment (mean +4.7 kg) raises the question of whether long-term use is sustainable in a population already burdened by metabolic syndrome. The trial did not follow patients long enough to determine whether weight gain would plateau, continue, or reverse the metabolic benefits of insulin sensitization.

Biopsy Timing and Ascertainment Bias

Liver biopsy is an imperfect reference standard. Sampling error is well documented: a single 15-gauge core captures roughly 1/50,000th of the liver. Intra-observer and inter-observer variability in NAS scoring add further noise.

PIVENS mitigated this with centralized pathology reads by a single expert. This reduces inter-observer variability but introduces a different risk: the entire dataset's histological assessment depends on one reader's interpretive framework. If that reader's threshold for scoring ballooning or inflammation shifted even subtly over the course of reading 494 biopsies (two per patient), it could systematically affect results.

The 24-week window for the post-randomization biopsy also introduces timing variability. Two patients assigned to the same arm might have their endpoint biopsies months apart, during which disease activity could fluctuate independently of treatment.

The Vitamin E Safety Signal No One Wanted to Talk About

In 2011, a year after PIVENS was published, the SELECT trial reported that vitamin E supplementation at 400 IU/day was associated with a statistically significant increase in prostate cancer risk among healthy men (HR 1.17 to 99% CI 1.004-1.36). PIVENS used 800 IU/day, double the SELECT dose.

The PIVENS cohort included both men and women, and the trial was not designed or powered to detect cancer outcomes. No excess cancer signal was observed during the study period. But the SELECT finding cast a shadow over long-term vitamin E use, particularly in male NASH patients, and contributed to the cautious language in subsequent AASLD guidance, which restricts its recommendation to nondiabetic patients with biopsy-proven NASH and notes the uncertain long-term safety profile.

Hemorrhagic stroke risk at high-dose vitamin E supplementation, identified in meta-analyses, adds another layer of concern that the 96-week PIVENS window could not address.

Generalizability Gaps Beyond Diabetes Exclusion

Several additional factors limit how broadly PIVENS results can be applied.

Ethnicity and race. The cohort was predominantly white (67%) and Hispanic (23%), with limited representation of Black and Asian patients. NASH prevalence and histological patterns differ across ethnic groups, and treatment response may as well.

Fibrosis stage. Patients with cirrhosis (stage 4 fibrosis) were excluded. Patients with bridging fibrosis (stage 3) were included but represented a minority. The population skewed toward earlier-stage disease, where the clinical urgency for treatment is lower.

BMI range. Mean BMI was approximately 34. Patients at the extremes of BMI, the very obese and the lean NASH phenotype, were underrepresented.

Concomitant medications. Statin use, which affects liver enzymes and may independently influence NASH histology, was permitted but not controlled for as a stratification variable.

Post-Publication Commentary

Letters to the editor following the PIVENS publication raised several points that did not receive adequate attention in the original manuscript.

One recurring criticism targeted the clinical relevance of a 2-point NAS improvement. Correspondents noted that a patient could move from NAS 5 to NAS 3, still meeting diagnostic criteria for NASH, and be classified as a treatment success. The endpoint measured relative improvement, not resolution or clinically meaningful disease modification.

Others questioned whether the placebo response rate of 19% reflected true histological improvement or regression to the mean combined with biopsy sampling variability. A 19% spontaneous improvement rate is high enough to raise this concern, especially given the known limitations of single-core liver biopsy.

The NIDDK-funded NASH CRN, which conducted PIVENS, has been transparent about these limitations in subsequent publications. The group's later FLINT trial of obeticholic acid used a similar composite endpoint but added fibrosis improvement as a co-primary, reflecting lessons learned from PIVENS about the inadequacy of NAS alone.

What This Means for Practice Today

PIVENS remains one of the most cited NASH trials despite its limitations, largely because no pharmacotherapy was approved for NASH until resmetirom (Rezdiffra) received accelerated FDA approval in March 2024. For 14 years, vitamin E and pioglitazone were the only options with randomized evidence, however imperfect.

Clinicians prescribing based on PIVENS should acknowledge three things: the evidence applies most directly to nondiabetic patients with moderate NASH and without cirrhosis; the benefit signal is histological rather than clinical; and the durability of treatment effect beyond two years is unknown. These are not reasons to withhold treatment from appropriate patients. They are reasons to set honest expectations and monitor actively.

Frequently asked questions

›

›

›

›

›

›

›

›

›

›

References

1. Sanyal AJ, Chalasani N, Kowdley KV, et al. Pioglitazone, vitamin E, or placebo for nonalcoholic steatohepatitis. N Engl J Med. 2010;362(18):1675-1685. PubMed
2. Klein EA, Thompson IM Jr, Tangen CM, et al. Vitamin E and the risk of prostate cancer: the Selenium and Vitamin E Cancer Prevention Trial (SELECT). JAMA. 2011;306(14):1549-1556. PubMed
3. Rinella ME, Neuschwander-Tetri BA, Siddiqui MS, et al. AASLD Practice Guidance on the clinical assessment and management of nonalcoholic fatty liver disease. Hepatology. 2023;77(5):1797-1835. PubMed
4. Rezdiffra (resmetirom) prescribing information. U.S. Food and Drug Administration. 2024. FDA Label
5. Neuschwander-Tetri BA, Loomba R, Sanyal AJ, et al. Farnesoid X nuclear receptor ligand obeticholic acid for non-cirrhotic, non-alcoholic steatohepatitis (FLINT): a multicentre, randomised, placebo-controlled trial. Lancet. 2015;385(9972):956-965. PubMed