Rosiglitazone Effect on Mitochondria and Lipoatrophy: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Rosiglitazone Effect on Mitochondria and Lipoatrophy (NCT00367744), a randomized phase 2 study. The primary outcome is Change in Limb Fat at 48 Weeks.

2. Clinical Question

The study also identifies change in common carotid artery intima-media thickness as a secondary outcome. The registry record does not provide a result for that outcome in the posted comparison. The main result therefore addresses the reported limb-fat endpoint, not every outcome examined in the study.

The clinical question is comparative: it concerns the distribution of change in limb fat between the randomized groups. The registry reports group medians and a Wilcoxon (Mann-Whitney) test. That test compares ranks across groups; it does not directly estimate a difference between medians. The posted result can therefore describe the group summaries and the reported rank-test evidence, but it should not be read as a direct estimate of how much an individual participant’s limb fat would change because of treatment.

3. Trial Design

This was a phase 2, randomized, quadruple-masked, parallel-group study with rosiglitazone and placebo arms. The registry lists enrollment as 71 and identifies the study as completed. The sponsor was the National Institute of Allergy and Infectious Diseases (NIAID). The registry gives a study start of 2006-07 and primary completion of 2008-12.

Participants were described as having HIV-1 infection and receiving stable antiretroviral therapy that did not include stavudine (d4T) or zidovudine (AZT). The rosiglitazone arm received active rosiglitazone 4 mg BID; the control arm received matching placebo BID. The parallel design means the comparison was between groups assigned to different study treatments, rather than a crossover in which each participant received both.

Randomization supports a group comparison by allocating participants to study arms through a chance-based process. Masking is intended to reduce the influence of treatment knowledge on care, participant reporting, and outcome assessment. The registry calls the masking quadruple but does not specify which roles were masked. Details of the randomization procedure, allocation concealment, and the exact masking arrangements are not reported in the registry record.

The primary endpoint was change in limb fat at 48 weeks, measured using dual-energy x-ray absorptiometry (DEXA). The registry describes the primary analysis population as intention to treat, with last observation carried forward if week 48 limb-fat data were missing and post-baseline limb-fat data were available. Other protocol details, including the planned enrollment rationale and statistical analysis plan, are not reported in the registry record.

4. Participant Flow and Accounting

The registry reports how participants moved through each stage of the study. All counts below are exactly as posted; missing counts are not inferred from other stages.

Before assignment: A total of 71 subjects were randomized. Results are reported for 71 subjects.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intention to treat analysis with last observation carried forward if week 48 limb fat data was missing and post-baseline limb fat was availble.

5. Baseline Characteristics

The table reproduces the baseline characteristics posted in the registry, which describe the participants analysed at baseline and help show whether the randomized groups were comparable.

6. Primary Endpoint

The registry lists the primary outcome as Change in Limb Fat at 48 Weeks, measured over 48 weeks.

The primary endpoint is change in limb fat, in grams, from baseline to week 48. The registry defines change as the week 48 measurement minus the baseline measurement. Limb fat was measured at both time points using dual-energy x-ray absorptiometry (DEXA), an imaging method used to estimate body composition.

Change from baseline focuses on how the measured quantity moved during follow-up rather than only comparing values at the final visit. The endpoint is relevant to the study’s stated focus on HIV lipoatrophy and limb fat. A positive change represents an increase relative to baseline under the registry’s definition; a negative change represents a decrease. These are general interpretations of the direction of a change score, not claims about the cause of any individual participant’s result.

The registry summarizes the endpoint with medians and interquartile ranges. That choice describes the centre and spread of the observed change scores without requiring the distribution to be represented by a mean and standard deviation. The posted group medians are 448 grams for rosiglitazone and 153 grams for placebo. The corresponding interquartile ranges are 138 to 1670 grams and -100 to 682 grams. The endpoint captures limb-fat change; it does not, by itself, establish what caused that change or describe other health outcomes.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary result combines descriptive summaries of limb-fat change with a rank-based comparison. Understanding what each statistic answers helps keep the group summaries, the hypothesis test, and the uncertainty about the contrast distinct. General statistical explanations follow, alongside the details reported in the registry record.

What do the medians describe?

A median is the middle value when observations are ordered. It describes the centre of a group’s observed distribution and is less affected by unusually large or small values than an arithmetic mean can be. The reported medians are 448 grams for rosiglitazone and 153 grams for placebo. They summarize the centre of each arm’s change scores; they do not by themselves estimate the difference between arms or show how precisely that difference is known.

The interquartile range describes the span from the lower to the upper quartile, covering the central half of observations under the usual definition. The registry reports ranges of 138 to 1670 grams for rosiglitazone and -100 to 682 grams for placebo. These ranges help convey within-group variability. They are not uncertainty intervals for the group medians or for a treatment contrast. Overlap between interquartile ranges does not replace a formal comparison, and their endpoints should not be treated as confidence limits.

What does the Wilcoxon (Mann-Whitney) test assess?

The reported test is a nonparametric, rank-based method for comparing two independent groups. In general, it orders the observations and evaluates whether ranks tend to be higher in one group than the other. It can be useful when a comparison is not based on a normal-distribution model. The test does not require the outcome to be reduced to a difference between means.

A rank-test result is not automatically a test specifically of medians. If the groups’ distributions have similar shapes, a rank contrast may sometimes be interpreted as evidence of a location shift. The registry record does not report distributional diagnostics or state such an assumption. It is therefore more precise to say that the test compared ranks than to describe it as directly testing whether the two medians differ.

The reported p-value is 0.02. In general, a p-value is calculated under a null model and describes how incompatible the observed test statistic, or a more extreme one, would be with that model. It is not the probability that the null hypothesis is true, nor does it quantify the probability that the result occurred by chance. It also does not measure effect size, clinical importance, or the probability that the finding will recur. The registry does not report a confidence interval or a numerical effect estimate for the rank comparison, so the p-value alone cannot communicate the range of plausible contrast sizes.

How should change scores be read?

The endpoint is defined as week 48 limb fat minus baseline limb fat. A change score retains the direction of movement: positive values indicate an increase, while negative values indicate a decrease. This is a general reading of the arithmetic definition. The registry reports the endpoint in grams and says that DEXA was used at baseline and week 48. The measurement description does not establish how measurement error was handled or whether any additional adjustments were made; such details are not reported in the registry record.

The primary outcome is summarized as a median and interquartile range rather than a mean and standard deviation. In general, a median-based summary can be informative when data are skewed or contain extreme values, while a mean-based summary answers a different question about an arithmetic average. The registry does not state why this summary was selected, and no distributional plots or diagnostic information are reported in the registry record.

What does the missing-data rule imply?

The analysis is described as intention to treat, with last observation carried forward if the week 48 limb-fat measurement was missing and post-baseline limb-fat data were available. In general, last observation carried forward substitutes a participant’s most recent available post-baseline observation for a missing later measurement. This produces an analysis value for some participants without a week 48 observation, but it assumes that carrying that earlier value forward is an acceptable representation of the unobserved week 48 value. If the outcome would have continued to change, the substitution can misrepresent that trajectory.

The registry does not state how many measurements were missing, how many participants were handled by this rule, or whether the method was assessed against alternatives. It also does not describe the treatment of participants without post-baseline limb-fat measurements. These details matter when assessing the impact of missingness and cannot be inferred from the reported medians or p-value.

What information is absent from the analysis report?

The registry reports one primary-outcome analysis using the Wilcoxon (Mann-Whitney) method. It does not report a confidence interval for the between-group contrast, an adjusted model, or additional analyses testing the robustness of the result. It does not report the statistical analysis plan, whether any interim analyses occurred, or how multiplicity was handled. General statistical practice treats those details as relevant because analysis choices and testing across multiple outcomes can affect how evidence is interpreted. Their absence does not establish that any particular procedure was or was not used; it means the procedure is not reported in the registry record.

10. Analyses Reported by the Sponsor

The registry reports the sponsor’s primary comparison as a Wilcoxon (Mann-Whitney) test, with p = 0.02. It also reports the arm-level medians and interquartile ranges for change in limb fat. The test result should be read as the reported rank-based comparison, not as a separately stated estimate of the difference between medians.

No separate recalculated comparison or confidence interval is reported in the registry record. Consequently, there is no second estimate to set against the sponsor’s analysis. The descriptive medians and ranges can be reviewed alongside the reported test, but they do not independently reproduce that test or establish its uncertainty. Any new calculation from summary values alone would not necessarily match a test based on participant-level ranks.

The registry does not provide enough detail to independently verify the analysis population, the handling of every missing observation, or other model and testing choices beyond the stated intention-to-treat approach and last-observation-carried-forward rule. Those limits should be distinguished from a discrepancy: the registry does not report a conflicting recalculated result. The sponsor’s stated analysis remains the available inferential result, with the omissions above limiting independent assessment.

11. Multiplicity, Missing Data and Analysis Populations

The registry identifies a primary endpoint of change in limb fat at 48 weeks and lists change in common carotid artery intima-media thickness as a secondary outcome. It reports one analysis for the primary endpoint. A result for the secondary outcome is not available in the posted comparison; the registry record indicates that values were missing or not estimable for a two-arm comparison for another endpoint.

When a study evaluates multiple outcomes or performs multiple statistical tests, multiplicity can increase the chance of observing at least one apparently notable result under a collection of null hypotheses. General approaches include designating a primary outcome, adjusting significance thresholds, or interpreting secondary findings as exploratory. The registry record does not report whether multiplicity adjustments were planned or used, how secondary outcomes were tested, or whether other analyses were conducted. It is therefore not possible to characterize the reported p-value as adjusted or unadjusted on the basis of the record.

The primary analysis is described as intention to treat, with last observation carried forward if week 48 limb-fat data were missing and post-baseline limb-fat data were available. This is a specific missing-data rule, but the record does not report the number of missing observations, their distribution between arms, the reasons for missingness, or sensitivity analyses. General statistical interpretation depends on whether missing outcomes are related to observed or unobserved values. The record does not establish those relationships.

The registry does not report the protocol’s detailed statistical plan, sample-size justification, interim-analysis procedures, or a fuller description of the analysis populations. These items should be treated as unreported rather than assumed absent from the study conduct. The available description supports reading the main result as an intention-to-treat analysis using the stated imputation rule, while leaving important questions about multiplicity and missing-data robustness unresolved.

12. Safety Results

The registry reports serious adverse events as affected participants divided by participants at risk.

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 0% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: elevated creatinine (grade 2) (20.6%), elevated cholesterol (grade 2 or 3) (16.2%), elevated blood pressure (grade 2 or 3) (14.7%), elevated triglycerides (grade 2) (11.8%), low phosphate (grade 2 or 3) (11.8%).

The registry reports serious adverse events in 15 of 34 participants in the rosiglitazone group and 13 of 37 in the placebo group. It reports other adverse events in 6 of 34 participants in the rosiglitazone group and 7 of 37 in the placebo group. These are counts of participants reported as affected, with the corresponding at-risk group counts. They describe events recorded during the study and do not, by themselves, establish that a study treatment caused an event.

The registry lists the following serious adverse-event terms with their reported percentages: Elevated Creatinine (Grade 2), 20.6%; Elevated Cholesterol (Grade 2 or 3), 16.2%; Elevated Blood Pressure (Grade 2 or 3), 14.7%; Elevated Triglycerides (Grade 2), 11.8%; Low Phosphate (Grade 2 or 3), 11.8%; and Elevated LDL Cholesterol (Grade 2 or 3), 10.8%. The record does not provide further detail in this summary about the denominator or whether participants could contribute to more than one listed term, so these percentages should be read as registry-reported figures rather than reconstructed event rates.

Deaths are not reported in the registry record. The adverse-event summaries do not provide enough information to assess timing, duration, severity beyond the listed grades, or causal attribution for each event. They should be interpreted as descriptive safety reporting, not as a complete comparison of treatment-related risk.

13. What the Result Does, and Does Not, Mean

The result shows that the registry reported a higher median change in limb fat in the rosiglitazone group than in the placebo group at 48 weeks. The reported medians were 448 grams and 153 grams, respectively, and the Wilcoxon (Mann-Whitney) comparison had p = 0.02. The interquartile ranges indicate variation in change scores within both groups. These registry entries describe the posted primary outcome and its reported statistical test.

The result does not provide a confidence interval for the group contrast, an estimate of its precision, or a direct numerical estimate of the difference between groups. A p-value does not supply those quantities. The reported test is rank-based, so it should not be presented as a direct test of the median difference without suitable assumptions about the outcome distributions. The registry does not state whether such assumptions were examined.

The result does not establish what happened for every participant: group medians do not show individual trajectories. Nor does the primary endpoint alone establish a change in other measures of body composition, symptoms, functioning, or longer-term health. Common carotid artery intima-media thickness is listed as a secondary outcome, but a result for it is not reported in the posted comparison.

The study’s population and treatment context also matter. The registry describes people with HIV-1 infection receiving stable antiretroviral therapy that did not include stavudine or zidovudine. The result should not automatically be generalized to other populations or treatment contexts. The reported missing-data approach, last observation carried forward under the stated condition, is another consideration; the amount of missing data and sensitivity of the result to alternative methods are not reported.

These findings are a statistical description of the registry’s reported analysis, not medical advice. They do not establish an individual outcome or settle the clinical importance of the observed group summaries. The registry record leaves important uncertainties about precision, missing data, secondary outcomes, and the broader analysis plan.

14. Why This Trial Matters Statistically

This trial illustrates why the statistical method needs to be read alongside the outcome summary. The registry reports medians and interquartile ranges for limb-fat change and uses a Wilcoxon (Mann-Whitney) test for the group comparison. The summaries describe each group’s observed distribution, while the rank test addresses how observations compare across groups. Neither should be substituted for the other, and the test is not automatically a direct test of the difference between medians.

The result also shows why a p-value is not a complete account of evidence. The reported p = 0.02 conveys the result of the stated test, but the registry does not report a confidence interval or a rank-based effect estimate. Readers therefore have a test result and descriptive summaries without an interval that expresses uncertainty about the between-group contrast. This limits assessment of its precision and plausible size.

The endpoint’s interpretation depends on its measurement and definition: DEXA-based limb-fat measurement at baseline and week 48, summarized as week 48 minus baseline. The intention-to-treat description and last-observation-carried-forward rule are also important, because missing data handling can shape an endpoint analysis. The registry’s lack of missingness counts and sensitivity analyses means the influence of that rule cannot be assessed from the posted details.

Finally, the record distinguishes the primary limb-fat result from a listed secondary outcome for which no two-arm result is posted. The absence of detail on multiplicity and the broader analysis plan is relevant to how a single reported test is interpreted among the study’s outcomes. Together, these points make the trial useful for illustrating careful reading of outcome summaries, rank tests, missing-data rules, and limits in registry reporting.

15. Limitations

• The posted summary statistics are the only data these results rest on; patient-level data, full protocols and statistical analysis plans are not available.
• Recalculated results are simple, unadjusted calculations from posted summaries. They do not replace the sponsor's prespecified analysis.

The reported comparison has no confidence interval or stated effect estimate for the between-group contrast. The group medians and interquartile ranges provide descriptive information, but they do not quantify the precision of that contrast. The registry reports p = 0.02 from a Wilcoxon (Mann-Whitney) test, but does not provide additional details such as a rank-based effect measure or distributional assessment.

The primary analysis used last observation carried forward for missing week 48 limb-fat data when post-baseline data were available. The registry does not report how much data were missing, whether missingness differed between groups, why measurements were missing, or whether alternative missing-data methods were examined. Consequently, the robustness of the primary result to the stated imputation rule cannot be evaluated from the registry record.

The record lists common carotid artery intima-media thickness as a secondary outcome but does not report a result for it in the posted two-arm comparison. It also does not report how multiplicity was addressed. The full statistical analysis plan, sample-size justification, interim analyses, and allocation-concealment details are not reported in the registry record. These omissions limit assessment of the prespecification and context of the reported test; they do not establish that particular methods were or were not used.

The adverse-event information is also limited in the registry summary. Counts are reported for participants affected by serious and other adverse events, but deaths are not reported in the registry record, and details on timing and causal attribution are not given. The listed serious-event percentages lack accompanying denominator details in the record summary. These limits call for cautious interpretation of safety reporting as descriptive rather than definitive evidence of comparative risk.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Rosiglitazone Effect on Mitochondria and Lipoatrophy — NCT00367744.
• Registry record last updated 2017-02-06; status last verified 2006-08.
• Results first posted 2017-02-06.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

19. Record Summary

The registry describes a completed, randomized, quadruple-masked, parallel phase 2 study sponsored by the National Institute of Allergy and Infectious Diseases (NIAID). It lists enrollment as 71, a start of 2006-07, and primary completion of 2008-12. The study compared rosiglitazone active 4 mg BID with matching placebo BID in people with HIV-1 infection and lipoatrophy receiving stable antiretroviral therapy that did not contain stavudine or zidovudine.

The primary endpoint was change in DEXA-measured limb fat from baseline to 48 weeks. The registry reports medians of 448 grams for rosiglitazone and 153 grams for placebo, with interquartile ranges of 138 to 1670 grams and -100 to 682 grams. The reported Wilcoxon (Mann-Whitney) test gave p = 0.02. A confidence interval for the between-group comparison is not reported in the registry record.

The analysis is described as intention to treat with last observation carried forward under the stated missing-data condition. Change in common carotid artery intima-media thickness is listed as a secondary outcome, but its result is not reported in the posted comparison.