Treatment of Nonalcoholic Fatty Liver Disease in Children (TONIC): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Treatment of Nonalcoholic Fatty Liver Disease in Children (TONIC) (NCT00063635), a randomized phase 3 study. The primary outcome is Number of Participants With Sustained Reduction in Alanine Aminotransferase (ALT) to Either 50% of Baseline Value or < 40 IU/L.

2. Clinical Question

TONIC asked whether modifying insulin resistance or oxidative stress could improve indicators of liver injury or quality of life in children with nonalcoholic fatty liver disease. The primary question focused on sustained reduction in alanine aminotransferase (ALT), a serum marker of liver injury. The registry record defines that outcome using ALT measurements across visits from 48 to 96 weeks of treatment.

The record describes the study purpose as determining whether therapeutic modification of insulin resistance or oxidative stress leads to improvement in serum or histologic indicators of liver injury or quality of life. The posted primary result concerns ALT reduction; it does not by itself answer every part of that broader purpose. Histologic changes and other outcomes were listed separately as secondary measures.

3. Trial Design

TONIC was a phase 3, randomized, triple-masked, parallel-group trial. The registry lists 173 enrolled participants and identifies the study as completed. The listed sponsor was the National Institute of Diabetes and Digestive and Kidney Diseases. The recorded start was 2005-09, and the primary completion was 2009-09.

Participants were assigned to metformin, vitamin E, or matching placebo. The registry specifies metformin at 500 mg twice daily and vitamin E at 400 IU twice daily. It does not give a separate placebo schedule or further details about the matching process. Triple masking indicates that masking involved multiple groups, but the identities of the masked parties and the procedures used to maintain masking are not reported in the registry record.

The parallel-group structure means the treatment groups were followed as separate groups rather than participants crossing from one assigned treatment to another. The primary outcome was assessed using ALT measurements from baseline through 96 weeks, with sustained reduction defined across visits from 48 to 96 weeks. Secondary outcomes included histologic scores assessed from baseline to 96 weeks, as well as other laboratory and body-size measures.

The registry does not report the original sample-size calculation, allocation ratio, randomization method, trial sites, eligibility criteria in detail, adherence measures, or the planned duration of follow-up beyond the endpoint timeframes. The trial's phase and masking labels describe its registry classification, but they do not supply these missing operational details.

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.

Overall Study

Analysis population for the primary outcome, as defined in the registry: All enrolled patients were included in analysis of the primary outcome, sustained reduction in ALT level. Patients missing a 96-week ALT measurement were imputed as not achieving a sustained reduction.

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 Number of Participants With Sustained Reduction in Alanine Aminotransferase (ALT) to Either 50% of Baseline Value or < 40 IU/L, measured over baseline and 96 weeks.

The primary endpoint counted participants who achieved a sustained reduction in ALT. The registry's outcome title describes reduction to either 50% of baseline or < 40 IU/L. Its detailed definition states that ALT had to be 50% or less of baseline, or 40 IU/L or less, at each visit from 48 to 96 weeks of treatment. Thus, the outcome required a qualifying result across the specified visits, rather than a single low ALT measurement.

ALT is a serum marker used in the trial as an indicator of liver injury. Requiring sustained reduction makes the endpoint more demanding than recording whether ALT fell at one measurement. The outcome is binary for each participant: the participant either met the sustained-reduction definition or did not. The registry included all enrolled patients in the primary-outcome analysis and specified that a missing 96-week ALT measurement was imputed as not achieving sustained reduction.

This endpoint captures a laboratory outcome, not every aspect of liver health. The trial also listed histologic outcomes based on liver biopsies, and other measures such as AST, body mass index, and serum vitamin E. Those are distinct outcomes and should not be treated as interchangeable with sustained ALT reduction.

7. Results Posted in the Registry

• Metformin: 15.8% (9/57) versus 17.2% (10/58) on control; risk difference -1.5 percentage points (95% CI -15.0 to 12.1), risk ratio 0.92 (0.40–2.09); unadjusted two-proportion z-test p = 0.834.
• Vitamin E: 25.9% (15/58) versus 17.2% (10/58) on control; risk difference 8.6 percentage points (95% CI -6.3 to 23.5), risk ratio 1.50 (0.74–3.06); unadjusted two-proportion z-test p = 0.259.

8. Secondary Efficacy Endpoints

The registry also posts results for secondary outcomes. For each outcome below, the endpoint type, the test applied to the posted summary statistics, the result and its interpretation are given in turn. Secondary endpoints are usually not adjusted for multiple testing, so they support rather than confirm the primary finding.

Number of Participants With Improvement in Liver Fibrosis Score

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Metformin: 44.0% (22/50) versus 40.4% (19/47) on control; risk difference 3.6 percentage points (95% CI -16.1 to 23.2), risk ratio 1.09 (0.68–1.74); unadjusted two-proportion z-test p = 0.722.
  • Vitamin E: 36.0% (18/50) versus 40.4% (19/47) on control; risk difference -4.4 percentage points (95% CI -23.8 to 14.9), risk ratio 0.89 (0.54–1.48); unadjusted two-proportion z-test p = 0.654.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Vitamin E and the control. The interval gives the range of effects that remain compatible with the data.

Number of Participants With Improvement in Steatosis Score

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Metformin: 52.0% (26/50) versus 40.4% (19/47) on control; risk difference 11.6 percentage points (95% CI -8.1 to 31.3), risk ratio 1.29 (0.83–1.99); unadjusted two-proportion z-test p = 0.253.
  • Vitamin E: 54.0% (27/50) versus 40.4% (19/47) on control; risk difference 13.6 percentage points (95% CI -6.1 to 33.3), risk ratio 1.34 (0.87–2.06); unadjusted two-proportion z-test p = 0.181.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Vitamin E and the control. The interval gives the range of effects that remain compatible with the data.

Number of Participants With Improvement in Lobular Inflammation Score

• Endpoint type: Binary: the number of participants with the event, compared as proportions.
• Test: Unadjusted two-proportion z-test, with a Wald 95% confidence interval for the risk difference.
• Result:
  • Metformin: 46.0% (23/50) versus 42.6% (20/47) on control; risk difference 3.4 percentage points (95% CI -16.3 to 23.2), risk ratio 1.08 (0.69–1.69); unadjusted two-proportion z-test p = 0.733.
  • Vitamin E: 44.0% (22/50) versus 42.6% (20/47) on control; risk difference 1.4 percentage points (95% CI -18.3 to 21.2), risk ratio 1.03 (0.66–1.63); unadjusted two-proportion z-test p = 0.886.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Metformin and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Vitamin E and the control. The interval gives the range of effects that remain compatible with the data.

The registry lists secondary assessments spanning histologic features, serum measures, body mass index, and serum vitamin E levels. Ten secondary outcome measures are listed, while the full outcome set includes the primary measure as well. Among the histologic comparisons with posted results, the reported p-values do not indicate statistically significant differences from placebo. The observed counts vary by outcome and treatment group, but the broader pattern does not provide a clear, consistent separation across these histologic measures.

These outcomes add context to the primary ALT result, but they answer different questions. ALT is a serum marker assessed repeatedly for sustained reduction, whereas histologic scores reflect biopsy-based features assessed from baseline to 96 weeks. Findings on one measure do not automatically confirm or refute findings on another. Some listed secondary outcomes have no numerical results in the registry record, so their direction and precision cannot be described from this record.

The secondary findings carry less weight for confirmatory conclusions because many outcomes and comparisons were analyzed. When multiple tests are conducted, the chance of at least one apparently notable result can increase. The registry record does not report a multiplicity adjustment or a prespecified hierarchy for interpreting secondary tests. The secondary results are therefore most appropriately read as supportive or exploratory, rather than as independent confirmation of the primary endpoint.

Considered together, the posted secondary results do not overturn the primary finding of no statistically significant difference from placebo on sustained ALT reduction. Nor do they establish that all aspects of liver injury were unaffected. The evidence is limited by unreported results for some measures, uncertainty in the posted comparisons, and the multiplicity of outcomes.

Secondary outcomes registered for this trial, for reference (only those above were recalculated):

• Change in Serum Aspartate Aminotransferase (AST)
• Change in Nonalcoholic Fatty Liver Disease (NAFLD) Score (Histologic Feature Scores Determined by Standardized Scoring of Liver Biopsies) From Baseline at 96 Weeks of Treatment
• Number of Participants With Improvement in Liver Fibrosis Score
• Number of Participants With Improvement in Steatosis Score
• Number of Participants With Improvement in Lobular Inflammation Score
• Number of Participants With Improvement in Ballooning Degradation Score
• Change in Body Mass Index
• Change in Serum Vitamin E Levels

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is binary: each participant is classified as having achieved sustained ALT reduction or not. A comparison of two such groups can be expressed as a difference in event proportions, also called a risk difference. A risk difference of zero represents equal observed proportions; positive and negative values indicate which group had the higher observed proportion, according to the order of comparison. The recalculated estimates use the posted event counts and denominators without adjustment.

What does the confidence interval add?

A confidence interval gives a range of values compatible with the observed data under the method used to construct it. It conveys uncertainty around an estimate; it is not a range containing a known true effect with certainty. For the primary comparisons, the intervals cross zero. Thus the posted data are compatible with differences in either direction, including no difference. The intervals are broad enough that the point estimates should not be read as precise descriptions of the treatment effects.

How should the p-values be read?

A p-value measures how unusual data at least as discrepant as the observed data would be under a specified null model and its assumptions. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The registry reports p-values of 0.83 for metformin versus placebo and 0.26 for vitamin E versus placebo using Mantel-Haenszel analyses. The unadjusted recalculations report 0.834 and 0.259, respectively. These are close, but the two approaches are not identical: the recalculations are simple, unadjusted comparisons, while the registry names a Mantel-Haenszel method.

Why use a Mantel-Haenszel method?

As a general statistical explanation, the Mantel-Haenszel approach can combine comparisons across strata, often to account for a stratification factor while estimating an overall association. The registry does not specify the strata or other details of the sponsor's analysis, so it is not possible to say which factors were used or how they affected the result. The posted p-values should be understood as the registry's reported analyses, not as fully documented models.

What does the missing-data rule do?

The primary analysis included all enrolled patients. A participant missing the 96-week ALT measurement was treated as not achieving sustained reduction. This is a stated outcome rule, not an assumption that the missing measurement would necessarily have been unfavorable. It can affect the estimated event counts and the comparison if missing measurements differ across groups or relate to outcomes. The numbers of participants missing measurements, the timing and reasons for missingness, and sensitivity analyses using alternative assumptions are not reported in the registry record.

Why do secondary analyses need separate interpretation?

The registry lists multiple secondary outcomes and many analyses. Each additional test creates another opportunity to observe a nominally small p-value by chance, particularly when many outcomes or comparisons are considered. The record does not report a multiplicity adjustment or a hierarchy specifying which secondary results were confirmatory. Consequently, secondary p-values are best treated as exploratory unless further analysis-plan information establishes otherwise.

For the histologic secondary outcomes with posted comparisons, the registry names chi-squared tests. As a general explanation, a chi-squared test assesses whether the distribution of categorical outcomes differs between groups. The record provides the outcome counts and p-values, but does not give further model details. No adjusted estimates or covariate models are specified in the registry record.

These methods answer different questions. A point estimate describes the observed contrast, an interval expresses uncertainty, and a p-value measures compatibility with a null model under specified assumptions. None alone establishes a causal mechanism or resolves whether a difference would matter clinically. Randomization supports comparisons between assigned groups, but the strength of inference also depends on missing-data handling, endpoint definition, execution, and the completeness of the analysis record.

11. Confidence Intervals and Precision

Precision concerns how tightly the data locate the treatment-group difference. The primary risk-difference intervals are wide and extend to both negative and positive values. Because zero lies within each interval, the results are compatible with no difference as well as differences in either direction. The intervals therefore do not support a precise estimate of the effect for either active-treatment comparison.

The event counts were 9 of 57 for metformin, 15 of 58 for vitamin E, and 10 of 58 for placebo. With these group sizes and event counts, the uncertainty around a difference is substantial. The point estimate for vitamin E is numerically higher than placebo, while the metformin point estimate is numerically lower; the intervals caution against treating either direction as established.

A confidence interval is not a guarantee that the true effect lies within its endpoints, and interval width does not indicate clinical importance. It is a measure of statistical uncertainty under the calculation's assumptions. In this setting, the broad ranges mean that the data do not finely distinguish small differences from larger differences in either direction.

The record does not provide the original design assumptions or a registry-reported confidence interval estimate for every outcome. The posted interval calculations are unadjusted comparisons. The registry's reported primary analyses provide p-values but do not include a corresponding effect estimate or interval in the analysis record.

12. Planning a Similar Trial: Sample Size

This is a planning illustration, not a re-evaluation of the original trial. It asks: if a new study wanted to detect a difference like the one observed, or a smaller one, how large would it need to be? It is deliberately not “observed power” of the finished trial, which is not a meaningful quantity.

The sample-size illustrations translate an observed effect into planning requirements under stated assumptions. For the binary outcomes shown, the numbers are participants per arm, not total events. The calculations assume a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. They examine detecting the full observed effect, or 75% or 50% of that observed effect.

The 80% power column gives the estimated participants per arm for a design that, under those assumptions and for the specified effect, would have an 80% chance of rejecting the null hypothesis when that effect is true. The 90% power column gives the corresponding planning estimate for a 90% chance. These are design probabilities conditional on the assumptions, not guarantees that a future study will reach a particular result.

Smaller effects require more participants per arm to distinguish them from random variation at the same significance level and power target. The values can therefore rise sharply when the planning effect is reduced from the observed effect. They are illustrative calculations based on the observed trial effect, rather than the trial's own sample-size justification. The original design assumptions and sample-size justification are not reported in the registry record.

This is not observed power. It does not calculate the probability that the completed trial detected an effect after the results were known. The displayed quantities are forward-looking planning illustrations, conditional on their assumptions, and should not be used to reinterpret the trial's observed p-values or confidence intervals.

13. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.172; treatment proportion p₁ = 0.158; average p̄ = 0.165.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.5251 = 1.0292.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5250 = 0.4419.
4. n = (1.0292 + 0.4419)² / (-0.0145)² = 10,266.1, rounded up to 10,267 participants per arm for 80% power.

The smaller-effect rows in the table above repeat the same calculation with the effect multiplied by 0.75 and 0.5 (for ratios, applied to the logarithm of the ratio).

R code

alpha <- 0.05
z_a   <- qnorm(1 - alpha / 2)          # two-sided
p0 <- 0.172413793103448                          # control proportion
p1 <- 0.157894736842105                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^2, 6))
}
for (f in c(1, 0.75, 0.5))
  cat(sprintf("%3.0f%% of observed effect: n per arm = %s (80%% power), %s (90%% power)\n", 100 * f,
              format(n_per_arm(0.80, f), big.mark = ","), format(n_per_arm(0.90, f), big.mark = ",")))

# Check against R's built-in function (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(63635)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 10,267 (80% power), 13,743 (90% power)
 75% of observed effect: n per arm = 18,413 (80% power), 24,649 (90% power)
 50% of observed effect: n per arm = 41,789 (80% power), 55,943 (90% power)
power.prop.test: 10266.1 
Simulated power at n = 10,267 per arm: 0.801 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The registry reports Mantel-Haenszel analyses for the primary comparisons: p = 0.26 for vitamin E versus placebo and p = 0.83 for metformin versus placebo. The unadjusted recalculations from the posted counts give p = 0.259 and p = 0.834, respectively. The close values are consistent with the same broad interpretation: neither comparison is statistically significant at the conventional 0.05 level.

The recalculations are simple comparisons of the posted counts and denominators. The registry identifies the sponsor analyses as Mantel-Haenszel, but does not report enough detail to reproduce their full specification, including any strata used. The small differences between the reported p-values and the recalculated values are compatible with differences in method and rounding; the registry record does not flag a discrepancy.

The recalculated risk differences and intervals are useful for describing the direction and uncertainty of the group contrasts, but they do not replace the registry's own analysis. The registry's test remains the reported sponsor analysis, while the unadjusted figures are supplemental descriptions based on summary data. Neither the p-values nor the recalculated estimates demonstrate equivalence or establish that a treatment did or did not cause a difference.

15. Multiplicity, Missing Data and Analysis Populations

The registry reports a large set of outcomes and analyses, including the primary endpoint, multiple secondary measures, and 22 analyses overall. Each statistical test is subject to random variation. When many tests are examined, some results may appear more noteworthy by chance, even if there is no underlying difference for the corresponding outcome. This is why confirmatory trials often specify how multiple comparisons will be handled.

A multiplicity adjustment, a prespecified testing sequence, or a clear distinction between confirmatory and exploratory analyses can help control interpretation across multiple tests. Such an adjustment or hierarchy is not reported in the registry record. The absence of this detail does not mean that no adjustment was used; it means the record does not establish whether one was planned or applied. Secondary p-values should therefore be interpreted cautiously and not treated as though each were the sole planned test.

The primary population is described as all enrolled patients. For a missing 96-week ALT measurement, the registry specifies classification as not achieving sustained reduction. The number of missing measurements, reasons for missingness, missingness by group, and sensitivity analyses are not reported in the registry record. As a result, the impact of this rule on the primary comparison cannot be assessed from the posted information alone.

The record also does not report the original statistical analysis plan, details of interim analyses, or whether any outcome definitions changed over time. These omissions limit assessment of whether the analyses followed a prespecified plan and whether all relevant outcomes and analyses are represented. They do not negate the posted findings, but they constrain how confidently the results can be interpreted as confirmatory.

16. 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: hepatobiliary/pancreas (8.6%), hepatotoxicity (3.5%), depression (3.4%), appendicitis (1.8%), constipation (1.8%).

The adverse-event record reports serious adverse events for 7 of 57 participants in the metformin group, 6 of 58 in the vitamin E group, and 14 of 58 in the placebo group. It also lists other adverse events for 17 of 57, 17 of 58, and 18 of 58 participants, respectively. These are descriptive counts; the registry does not provide comparative statistical tests for these figures in the record.

The listed serious-event terms include hepatobiliary/pancreas (8.6%), hepatotoxicity (3.5%), depression (3.4%), appendicitis (1.8%), constipation (1.8%), and elevated serum glucose (1.8%). The record does not make clear whether these term percentages are pooled across groups or how they relate to individual participants, so they should not be assigned to a treatment group.

These data do not establish that a treatment caused an event or that the groups differed in safety. Death counts are not reported in the registry record. Duration, severity details beyond the serious-event classification, and further event-level information are also not reported.

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

The posted primary results describe the proportion of participants meeting a specific sustained ALT-reduction definition during treatment. The registry analysis did not find a statistically significant difference between metformin and placebo or between vitamin E and placebo. The associated estimates remain uncertain, with intervals spanning no difference and possible effects in either direction.

That finding does not establish equivalence, prove that the treatments have no effect, or demonstrate that they improve liver outcomes. A non-significant result can arise when a real difference is absent, but it can also occur when the study data are too imprecise to distinguish a difference from random variation. The broad intervals and group sizes make precision an important part of interpretation.

The primary endpoint is a biochemical measure defined by repeated ALT values. It is not a direct measure of every clinical or histologic aspect of fatty liver disease. The registry separately lists biopsy-based outcomes and other measures, but the posted secondary findings do not provide a consistent basis for extending the primary result to all possible outcomes. Some secondary results are not numerically available in the record.

The findings apply to the enrolled trial population, the assigned treatments and doses, and the follow-up periods described in the registry. The record does not provide enough detail to assess generalizability across different patient groups or settings. The results should be read as evidence about the registered comparisons and endpoints, not as a complete account of every potential effect or safety outcome.

Statistically, the clearest conclusion is limited: the registry's primary analyses did not show a significant difference from placebo in sustained ALT reduction. They leave uncertainty about the size and direction of any underlying difference. They do not answer whether a different endpoint, population, or treatment schedule would yield another result.

18. Why This Trial Matters Statistically

This trial illustrates why a treatment comparison should not be reduced to whether a p-value crosses a threshold. The primary outcome counts and point estimates give a direction for the observed group contrasts, while the confidence intervals show that the estimates are imprecise. The registry's non-significant p-values and the intervals' inclusion of no difference are related but distinct parts of the evidence.

It also shows the importance of defining an outcome precisely. Sustained ALT reduction required qualifying measurements across visits, and missing 96-week measurements were classified as failures to achieve the endpoint. A different endpoint definition or missing-data rule could produce a different analysis, so those choices belong in interpretation rather than being treated as technical details.

The record makes visible the distinction between a sponsor-reported analysis and an unadjusted recalculation from summary counts. Their p-values are close, but the named methods differ and the registry does not provide every analysis detail. Transparent reporting of methods, effect estimates, intervals, and missing-data patterns helps readers understand what can and cannot be concluded.

Finally, the multiple secondary outcomes demonstrate why multiplicity matters. Secondary measures can provide useful context, but without a reported testing strategy they do not have the same confirmatory standing as a prespecified primary comparison. The trial is therefore a useful example of reading estimates, uncertainty, analysis methods, and outcome scope together rather than treating any single statistic as a complete answer.

19. 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.

Several details needed to fully assess the trial are not reported in the registry record. These include the eligibility criteria and age range, the randomization procedure and allocation ratio, the parties masked, and how masking was maintained. The record identifies a randomized, triple-masked, parallel design but does not supply these implementation details.

The original sample-size justification, target effect, and design assumptions are not reported. The planning illustrations based on observed effects cannot be substituted for the protocol's original calculation. The registry also does not report whether interim analyses occurred or provide a full statistical analysis plan.

For the primary outcome, the missing-96-week rule is specified, but the amount and distribution of missing data and sensitivity analyses are not reported. The registry does not provide enough detail to assess whether missingness differed across groups or whether other assumptions would alter the result. In addition, the reported primary analysis gives p-values without a registry-reported effect estimate and confidence interval; the interval calculations discussed in interpretation are unadjusted recalculations from summary data.

The record lists many outcomes and analyses without reporting a multiplicity strategy. Several secondary measures are listed without numerical results, limiting assessment of the broader outcome pattern. Adverse-event counts are descriptive, and group-specific term details and deaths are not reported. These limitations restrict precision and completeness of interpretation but do not change what the registry reports for the primary comparisons.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Treatment of Nonalcoholic Fatty Liver Disease in Children (TONIC) — NCT00063635.
• Registry record last updated 2012-09-27; status last verified 2012-08.
• Results first posted 2012-09-27.
• 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.

23. Record Summary

TONIC was a completed, phase 3, randomized, triple-masked, parallel-group trial with 173 enrolled participants. It compared metformin, vitamin E, and matching placebo in children with fatty liver. The primary outcome was sustained ALT reduction assessed across visits from 48 to 96 weeks.

The registry reports p = 0.83 for metformin versus placebo and p = 0.26 for vitamin E versus placebo. The posted counts were 9 of 57 for metformin, 15 of 58 for vitamin E, and 10 of 58 for placebo. Neither primary comparison was statistically significant. The estimates are imprecise, and the registry record does not establish equivalence or rule out effects in either direction.

Secondary outcomes covered histology and other measures, but the record does not provide numerical results for every listed outcome or a multiplicity strategy. The findings are best interpreted as limited to the registered endpoints, analyses, and follow-up.