Eltrombopag Phase III Study In Chinese Chronic ITP Patients: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Eltrombopag Phase III Study In Chinese Chronic ITP Patients (NCT01762761), a randomized phase 3 study. The primary outcome is Number of Participants (Responders) Achieving a Platelet Count >=50×10^9/L After the First 6 Weeks of Stage 1.

2. Clinical Question

The registry record describes a Phase III study of eltrombopag efficacy, tolerance and safety in previously treated Chinese adults with chronic ITP. The central comparison was the number of participants meeting the platelet-response threshold by the end of the 6-week double-blind period. The trial also assessed other platelet-response definitions, bleeding, rescue treatment, time to response and longer-term outcomes. Those additional questions broaden the clinical scope, but the primary comparison remains the clearest prespecified measure of short-term platelet response posted in the record.

3. Trial Design

This was a randomized, parallel-group Phase III study comparing eltrombopag with placebo. The registry lists quadruple masking and describes both a double-blind period and an open-label phase. It records enrollment of 155 participants. The study status is completed; the recorded start date is 2013-02-18 and the primary completion date is 2014-06-05.

The brief description states that the randomized comparison assessed oral eltrombopag against placebo in Chinese adults with chronic ITP who had not responded to or had relapsed after prior treatment. The placebo-controlled primary assessment covered 6 weeks of Stage 1. Following the double-blind phase, the study included a 24-week open-label extension to evaluate longer-term efficacy and safety. The registry also describes the possibility of continued eltrombopag treatment, based on investigator discretion, until commercial launch in China for participants considered to benefit. A pharmacokinetic and pharmacodynamic analysis was embedded in the study.

The posted primary endpoint counts are 3 responders among 50 participants in the placebo group and 60 among 104 in the eltrombopag group. The primary analysis population is defined as randomized participants who received at least one dose and had at least one post-Baseline platelet count in Stage 1. The registry record does not report the original sample-size justification or the detailed allocation procedure beyond stating that allocation was randomized. It also does not report the operational details of masking or the specific procedures used to implement the open-label period.

The trial’s design supports a randomized comparison for the initial treatment period, while the open-label extension addresses continued treatment without the same placebo-controlled comparison. These phases therefore answer different kinds of questions. The primary result concerns the short-term randomized contrast; longer-term observations from the extension should not be interpreted as if they retained the same randomized comparison.

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: Participants diagnosed with primary immune thrombocytopenia (ITP) for at least 12 months prior to randomization, platelet count of \<30×10\^9/Liter (L) within 48 hours (hrs) prior to Day 1; no response or relapsed after splenectomy or if not splenectomised and either not responded to prior therapies or relapsed to prior therapy were enrolled.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intent-to-Treat (ITT) Population: all randomized participants who received at least one dose of study medication and with at least one platelet count post-Baseline in Stage 1.

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 (Responders) Achieving a Platelet Count >=50×10^9/L After the First 6 Weeks of Stage 1, measured over From the start of study treatment (Day 1) up to the end of Week 6 of Stage 1.

The primary endpoint counts participants whose platelet count reached at least 50×10^9/L after the first 6 weeks of Stage 1. A threshold-based outcome turns repeated platelet measurements into a responder classification: each participant either meets the defined criterion or does not. The registry states that platelet counts were assessed at Week 1, Week 2, Week 3, Week 4, Week 5 and Week 6.

This endpoint was used to compare the proportion of participants achieving a specified platelet response during the initial treatment period. A binary responder outcome is suited to logistic regression, the method identified for the primary analysis. The registry says the model adjusted for baseline ITP medication use, splenectomy status, baseline platelet count of 15×10^9/L or lower, and treatment. Adjustment accounts for those listed baseline factors in estimating the treatment comparison; it does not change the endpoint’s definition.

The response rule also depends on how missing or interrupted follow-up was handled. The registry states that participants who withdrew from Stage 1 or were emergently unblinded were classified as negative responders from that point onward and at subsequent visits. Intermittent missing data, except withdrawals, were treated as missing. If a participant died, subsequent assessments were considered missing. These rules mean that the endpoint is not simply a count of observed high platelet measurements: its classification incorporates the stated handling of withdrawal and missing assessments.

A threshold has an understandable clinical interpretation as a response criterion, but it does not describe every aspect of platelet history. The primary endpoint does not, by itself, summarize how long a participant remained above the threshold or the timing and pattern of all measurements.

7. Results Posted in the Registry

• Eltrombopag: 57.7% (60/104) versus 6.0% (3/50) on control; risk difference 51.7 percentage points (95% CI 40.1 to 63.2), risk ratio 9.62 (3.17–29.16); unadjusted two-proportion z-test p = <0.001.

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 Achieving a Platelet Count >=50×10^9/L at Least Once During the First 6 Weeks of Stage 1

• 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:
  • Eltrombopag: 76.9% (80/104) versus 18.0% (9/50) on control; risk difference 58.9 percentage points (95% CI 45.5 to 72.3), risk ratio 4.27 (2.34–7.79); unadjusted two-proportion z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the proportion with the event is statistically higher for Eltrombopag than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Time to Response

• Endpoint type: Time to event: compared as a hazard ratio.
• Test: Hazard ratio as reported by the sponsor (Log Rank); it was not recalculated.
• Result:
  • Placebo vs Eltrombopag: hazard ratio 6.12 (95% CI 4.01–9.34), reported p < 0.001 (as reported in the registry).
• Interpretation:
  • The 95% confidence interval excludes no difference (a ratio of 1), so the hazard of the event is statistically higher for Placebo vs Eltrombopag than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The secondary outcomes span several related but distinct aspects of response and safety: reaching platelet thresholds, bleeding and clinically significant bleeding, time to response, rescue treatment, the proportion of assessments above a threshold, and duration above that threshold. This range shows that the study was not limited to a single binary response definition. The registry lists 29 secondary outcomes, but the posted record does not provide enough numeric results in the available account to characterize every one of them.

For the secondary outcome of reaching at least 50×10^9/L at least once during the first 6 weeks, the registry reports 9 of 50 participants in the placebo group and 80 of 104 in the eltrombopag group. The sponsor-reported odds ratio is 23.80, with a 95% confidence interval from 8.54 to 66.33 and p < 0.001. This endpoint is closely related to the primary response outcome, though its wording specifies at least one threshold achievement during the period. The time-to-response analysis also addresses the timing dimension rather than simply whether a threshold was reached.

These selected secondary results are directionally consistent with the posted primary comparison in the sense that they also show a group contrast in platelet-response measures. That does not mean every secondary outcome had the same pattern or that all secondary findings are equally persuasive. Outcomes may differ in definition, follow-up period, statistical method, and completeness of reporting.

Secondary outcomes often involve multiple statistical tests. If many outcomes are tested, the chance of at least one apparently unusual result can increase unless multiplicity is addressed. The registry record does not report a multiplicity adjustment or a hierarchy for interpreting the secondary tests. Their p-values should therefore carry less confirmatory weight than the primary analysis, and results for unreported outcomes should not be inferred from the outcomes that are posted.

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

• Number of Participants Achieving a Platelet Count >=50×10^9/L at Least Once During the First 6 Weeks of Stage 1
• Number of Participants Achieving a Platelet Count >=30×10^9/L and at Least 2 Times the Baseline Platelet Count at Least Once During the 6 Weeks of Stage 1, the Whole Stage 2 and the Whole Stage 3
• Number of Participants With Bleeding as Assessed Using the World Health Organization (WHO) Bleeding Scale
• Number of Participants With Clinically Significant Bleeding as Assessed Using the World Health Organization (WHO) Bleeding Scale
• Time to Response
• Number of Participants Who Required Protocol-defined Rescue Treatment During the First 6 Weeks of Stage 1, Whole Stages 2 & 3
• Number of Participants With a Platelet Count >=50×10^9/L During at Least 75% of Their Platelet Count Assessments
• Total Duration of Time a Participant Had a Platelet Count >=50×10^9/L

9. Statistical Interpretation

10. Statistical Methods Explained

The primary analysis used logistic regression to compare responder status between treatment groups. Logistic regression is a general method for modeling a binary outcome, such as meeting or not meeting a response criterion. It estimates an odds ratio for the treatment contrast, conditional on the variables included in the model. In this study, the registry states that the model adjusted for baseline ITP medication use, splenectomy, baseline platelet count of 15×10^9/L or lower, and treatment. The posted sponsor analysis gives an odds ratio of 26.08 with a 95% confidence interval from 7.29 to 93.26 and p < 0.001.

What does an odds ratio describe?

Odds compare the chance of an event with the chance of no event. An odds ratio contrasts those odds between groups. It is not the same quantity as a ratio of probabilities, and it is not an absolute difference in probabilities. When the outcome is common, odds ratios can appear more extreme than risk ratios. For that reason, the adjusted odds ratio should be interpreted alongside the observed responder counts and, where useful, an absolute-scale estimate. The recalculated unadjusted risk difference of 0.517, with a 95% confidence interval from 0.401 to 0.632, gives an absolute-scale description of the posted group contrast. It does not use the same adjustment as the sponsor’s logistic model.

Why adjust for baseline factors?

Adjustment can account for the listed baseline variables when estimating the treatment contrast. The registry identifies the factors included, but does not report further details of the model specification, such as how continuous variables were represented or whether interaction terms were considered. No additional modeling choices should therefore be assumed. The adjusted analysis is the sponsor’s reported primary comparison, while the recalculated risk difference is a simple unadjusted calculation from the posted group summaries.

How should a p-value be read?

A p-value describes how incompatible the observed data are with a specified null model, under the assumptions of the analysis. It is not the probability that the null hypothesis is true, and it does not measure the probability that the result occurred by chance in a broad, assumption-free sense. The reported p-value of < 0.001 is evidence against the no-association null for the sponsor’s modeled comparison. Statistical significance does not by itself establish clinical importance, identify the effect’s exact size, or address every possible bias.

What do confidence intervals contribute?

A confidence interval presents a range of values compatible with the data and the method used, at the stated confidence level. It communicates precision as well as direction. The reported 95% interval for the odds ratio is wide, so the exact magnitude is not estimated narrowly. The recalculated risk-difference interval gives uncertainty on another scale. These intervals are not interchangeable because the estimates differ in both measure and adjustment.

How does the response-time analysis differ?

Time to response is a time-to-event outcome rather than a simple responder count. The registry reports a median of 3.14 weeks for eltrombopag, with a 95% confidence interval from 3.0 to 4.14; the placebo median is not reported in the registry record. The reported analysis used a log-rank method and gives a hazard ratio of 6.12, with a 95% confidence interval from 4.01 to 9.34 and p < 0.001. A hazard ratio compares event rates over follow-up under the analysis model; its direction depends on which group and event are used as the reference. The available record does not give all coding details needed to reconstruct that direction independently.

For the primary binary outcome, missing-data handling is partly described. Withdrawal or emergency unblinding led to negative-response classification from that time onward, while intermittent missing observations other than withdrawals were treated as missing. After death, subsequent assessments were considered missing. The registry record does not report further details of the statistical treatment of those missing values or the assumptions behind it.

11. Confidence Intervals and Precision

Precision concerns how narrowly the data estimate an effect, not whether the estimate is important. The sponsor-reported adjusted odds ratio for the primary endpoint is 26.08, with a 95% confidence interval from 7.29 to 93.26. The interval excludes the null value of 1, but spans a broad range of possible odds ratios. Thus the posted result supports a treatment-group difference in the modeled comparison while leaving substantial uncertainty about its exact magnitude.

The simple unadjusted risk difference is 0.517, with a 95% confidence interval from 0.401 to 0.632. This interval is on an absolute scale, unlike the odds-ratio interval. Its lower and upper limits both lie above zero, but the range still allows materially different estimates of the absolute contrast. The two intervals answer different questions and should not be compared by their numerical widths alone.

A confidence interval is generated by a statistical procedure that, under its assumptions, has a stated long-run coverage property. It is not a range that assigns a probability to the true value after the data have been observed. It also does not account for every source of uncertainty, such as limitations in trial conduct, outcome measurement, or incomplete reporting. The posted intervals reflect the analysis methods used and the available outcome data.

The response-time estimate has its own uncertainty: the eltrombopag median is 3.14 weeks, with a 95% confidence interval from 3.0 to 4.14. The placebo median is not reported in the registry record. The hazard-ratio interval is from 4.01 to 9.34. As with the primary result, the interval conveys a range of estimates compatible with the method, while the incomplete details of event coding limit independent interpretation of direction.

Precision is also distinct from statistical significance. A small p-value can coexist with uncertainty about the exact magnitude, especially when an estimate has a wide interval. The intervals are therefore essential context for the point estimates.

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 planning figures are illustrations based on the observed effect, not the original trial design assumptions. The original sample-size justification and assumptions are not reported in the registry record. For the binary outcomes, the figures are participants per arm. For the time-to-event outcome, the figures are total events, not participants per arm.

The illustrations assume a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is set at the stated fraction of the observed effect. For each assumed effect, the 80% power column gives the number of participants per arm, or total events for the time-to-event calculation, associated with an 80% chance of detecting that assumed effect under the planning assumptions if it is present. The 90% power column uses a higher target chance of detection and consequently lists larger requirements.

The entries labeled 100% of the observed effect use the full observed effect as the planning target. The 75% and 50% rows consider smaller effects relative to that observed estimate. Smaller effects are harder to distinguish from random variation, so they require more participants or, for time-to-event planning, more events under otherwise similar assumptions. These calculations are not observed power. They do not report the probability that the completed trial detected its true effect, and they should not be read as evidence that the trial was adequately or inadequately designed.

Planning estimates depend on the chosen assumptions and may differ from the actual design process. In particular, using an observed effect as a planning anchor does not establish that the same effect would be expected in a new study. The illustrations are best read as conditional examples of how the required information changes with the target effect and desired power, rather than as a reconstruction of the sponsor’s original planning.

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 = 0.060; treatment proportion p₁ = 0.577; average p̄ = 0.318.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6589 = 1.2913.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.5482 = 0.4613.
4. n = (1.2913 + 0.4613)² / (0.5169)² = 11.5, rounded up to 12 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 <- 0.06                          # control proportion
p1 <- 0.576923076923077                          # 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(1762761)
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 = 12 (80% power), 15 (90% power)
 75% of observed effect: n per arm = 19 (80% power), 25 (90% power)
 50% of observed effect: n per arm = 35 (80% power), 47 (90% power)
power.prop.test: 11.5 
Simulated power at n = 12 per arm: 0.845 (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 sponsor’s primary analysis reports a logistic-regression odds ratio of 26.08, with a 95% confidence interval from 7.29 to 93.26 and p < 0.001. The recalculated estimate is an unadjusted risk difference of 0.517, with a 95% confidence interval from 0.401 to 0.632 and p < 0.001. The registry record does not flag a discrepancy between the sponsor analysis and the recalculation.

The two estimates are not competing versions of the same number. The sponsor’s odds ratio comes from a model adjusted for baseline ITP medication use, splenectomy, baseline platelet count of 15×10^9/L or lower, and treatment. The recalculated risk difference compares the posted group summaries without that adjustment. One is a relative contrast in odds; the other is an absolute contrast in response proportions. Different scales and adjustment strategies naturally yield different numerical estimates.

The sponsor’s model-based result is the reported primary analysis and should be distinguished from the simple recalculation. The recalculation is useful as an independent descriptive view of the posted counts, but it does not replace the sponsor’s specified analysis. Both are consistent with a statistically significant group contrast by their reported p-values, while their confidence intervals express uncertainty on their respective scales.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies a primary endpoint and lists a broad set of secondary outcomes. Testing multiple endpoints creates a multiplicity issue: when many hypotheses are examined, the chance of obtaining one or more small p-values under null conditions can increase. A prespecified hierarchy or adjustment can help control how confirmatory claims are made. The registry record does not report whether a multiplicity strategy was used, how secondary tests were ordered, or whether particular outcomes were considered exploratory. The posted secondary p-values should therefore be interpreted with more caution than the primary result.

The primary analysis population is defined as randomized participants who received at least one dose of study medication and had at least one post-Baseline platelet count in Stage 1. This definition is narrower than all randomized participants because it requires treatment exposure and a post-Baseline assessment. The registry calls it the Intent-to-Treat population, but the record does not report how many randomized participants were excluded from that population or provide a separate accounting of analysis populations by outcome.

The primary endpoint description gives specific rules for some missing observations. Participants who withdrew from Stage 1 or were emergently unblinded were treated as negative responders from the time of withdrawal or unblinding and at subsequent visits. Intermittent missing data, apart from withdrawals, were treated as missing. If a participant died, later assessments were considered missing. The record does not report additional methods for handling missing data, sensitivity analyses, or the assumptions used for intermittent missing measurements.

These choices matter because a binary responder result depends on both observed platelet counts and rules for observations that are not available. Treating withdrawal as nonresponse can differ from excluding missing observations or imputing them in another way. A sensitivity analysis can assess how robust a finding is to alternative assumptions, but such analyses are not reported in the registry record. Nor is the original statistical analysis plan reported in sufficient detail to establish whether all analysis choices followed a prespecified plan.

16. Safety Results

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

Deaths

Other (non-serious) adverse events

Participants with at least one non-serious event at or above the registry's 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: platelet count decreased (7.8%), thrombocytopenia (3.9%), acute myocardial infarction (3.9%), death (3.9%), cataract (3.8%).

The adverse-event record lists serious events and other adverse events for the treatment groups. Serious adverse events were recorded among 17 participants in the placebo group, with 51 at risk, and 33 participants in the eltrombopag group, with 104 at risk. Other adverse events were recorded among 48 participants in the placebo group, with 51 at risk, and 90 participants in the eltrombopag group, with 104 at risk. The registry lists 3 deaths for placebo and 0 for eltrombopag.

The at-risk counts are those recorded in the adverse-event data; they differ from the primary endpoint group count for placebo. The registry record does not explain that difference, so the adverse-event denominator should not be silently substituted for the primary analysis population. The listed figures count participants with events and do not establish that treatment caused an event. They also do not convey event timing, severity beyond the serious-event classification, duration, recurrence, or individual clinical context.

The registry lists platelet count decreased at 7.8%, thrombocytopenia at 3.9%, acute myocardial infarction at 3.9%, death at 3.9%, cataract at 3.8%, and cerebral infarction at 3.8% among the top serious-event terms. These percentages are registry entries; the record excerpt does not provide enough context to infer comparative risk from those terms alone. Safety counts should be considered alongside the trial’s size, duration, event definitions and the limitations of the posted reporting.

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

The posted primary result shows that more participants in the eltrombopag group than in the placebo group met the platelet-response criterion after the first 6 weeks of Stage 1: 60 of 104 versus 3 of 50. The sponsor’s adjusted logistic-regression comparison and the unadjusted recalculation both report p < 0.001, with confidence intervals that describe the uncertainty around their respective estimates. This is evidence of a difference in the measured response outcome within the trial population and study period.

The result does not, by itself, establish the exact size of the treatment effect. The adjusted odds-ratio interval is wide, and the unadjusted risk-difference interval is also a range rather than a single certain value. Neither p-values nor point estimates describe every participant’s response or resolve uncertainty about outcomes that were not measured or not reported.

The primary outcome is a threshold-based platelet measure. It does not alone answer questions about how long platelet counts remained above the threshold, bleeding outcomes, rescue treatment, longer-term effects, or overall safety. Those are separate outcomes, and not all are fully characterized by the record excerpt. The open-label extension also differs from the randomized placebo-controlled period, so observations from it should not be treated as an equivalent randomized comparison.

The findings apply most directly to the enrolled population described in the registry: Chinese adults with chronic ITP previously treated for the condition. The registry record does not establish how well the results generalize to populations with different characteristics or treatment histories. Nor does a registry result alone resolve all questions about trial conduct, missing data, multiplicity, or the interpretation of every secondary endpoint.

These statistics describe study findings; they are not individualized medical guidance. The distinction between a measured response, an estimated group contrast and a broader conclusion about treatment is important. The trial’s posted data support interpretation of its defined outcomes, while leaving other clinical and methodological questions open.

18. Why This Trial Matters Statistically

This trial is statistically informative because it combines randomization with a clearly defined binary primary outcome and a reported adjusted comparison. Random allocation is intended to support a fair treatment-group comparison by reducing systematic differences at baseline. The endpoint definition, analysis population and adjustment variables are described in the registry, allowing readers to understand important parts of the primary analysis rather than relying only on a headline estimate.

It also illustrates why effect measures need context. The sponsor reports an adjusted odds ratio, while an unadjusted calculation from the group summaries expresses the contrast as a risk difference. Both can be useful, but they answer different questions and are on different scales. Reading the responder counts, adjusted estimate, absolute estimate and confidence intervals together gives a fuller statistical account than relying on a p-value alone.

The study also demonstrates how endpoint definitions shape interpretation. A response after the 6-week period, at least one response during that period, time to response, and duration above a threshold are related but not interchangeable outcomes. Their differences matter for both analysis and clinical interpretation. The registry’s stated handling of withdrawals and missing assessments further shows that outcome classification can depend on analysis rules, not only observed measurements.

Finally, the record highlights limits of registry reporting. The original sample-size justification, multiplicity strategy, detailed missing-data methods and some analysis specifications are not reported. Those omissions constrain assessment of planning and secondary findings. Statistical interpretation therefore depends not just on the numerical estimates, but also on distinguishing what the record documents from what it leaves uncertain.

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.

The registry record does not report the original sample-size justification or the assumptions used to plan enrollment. Planning illustrations based on the observed effect cannot reconstruct those original decisions. The record also does not report a multiplicity strategy for the many secondary outcomes, so the confirmatory status of individual secondary findings is unclear.

Some aspects of the primary analysis are described, including adjustment factors and missing-data rules, but further statistical details are not reported in the registry record. These include the full model specification, sensitivity analyses for missing data, and a detailed account of participants included or excluded from each analysis. The primary analysis population requires treatment exposure and a post-Baseline platelet assessment, and the number of randomized participants outside that population is not reported.

The adverse-event at-risk count for placebo is 51, whereas the primary endpoint count is 50. The registry record does not explain the difference. Safety and efficacy denominators should therefore be kept distinct rather than assumed to refer to identical populations. The posted adverse-event counts also do not give detailed timing, duration, recurrence, or participant-level context.

The record lists a 24-week open-label extension and possible continued treatment based on investigator discretion. Without a concurrent randomized placebo comparison for that period, longer-term observations cannot be interpreted in the same way as the initial randomized result. In addition, the registry record does not provide enough details to independently reconstruct the event coding for the time-to-response hazard ratio. Finally, registry results alone do not resolve all questions about generalizability beyond the enrolled Chinese adults with previously treated chronic ITP.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Eltrombopag Phase III Study In Chinese Chronic ITP Patients — NCT01762761.
• Registry record last updated 2019-12-03; status last verified 2019-11.
• Results first posted 2015-03-09.
• 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

The registry describes a completed, randomized, parallel-group Phase III study of eltrombopag versus placebo in Chinese adults with chronic ITP. It records enrollment of 155 participants, quadruple masking, a 6-week primary comparison and an open-label extension. The primary outcome counted participants reaching a platelet count of at least 50×10^9/L after the first 6 weeks of Stage 1.

The sponsor-reported adjusted odds ratio was 26.08, with a 95% confidence interval from 7.29 to 93.26 and p < 0.001. The unadjusted recalculated risk difference was 0.517, with a 95% confidence interval from 0.401 to 0.632 and p < 0.001. These estimates use different measures and adjustment strategies. The registry also posts selected secondary response and time-to-response results and adverse-event counts. The original sample-size justification, multiplicity approach and additional missing-data analyses are not reported in the registry record.