A Safety Study Comparing LY2140023 to Atypical Antipsychotic Standard Treatment in Schizophrenic Patients: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Safety Study Comparing LY2140023 to Atypical Antipsychotic Standard Treatment in Schizophrenic Patients (NCT00845026), a randomized phase 2 study. The primary outcome is Time to Discontinuation Due to Adverse Event (AE).

2. Clinical Question

This is a tolerability-focused question, not a direct test of symptom improvement. Discontinuation may reflect whether participants remain on assigned treatment, but the primary endpoint as defined does not capture every aspect of safety, effectiveness, or treatment experience. The trial also included safety and other measures, including electrocardiogram, neurological examination, blood pressure, weight, pulse, glucose, and lipid outcomes. The registry describes an optional extension for participants who completed active treatment, but the primary question concerns the initial 24-week period.

3. Trial Design

The registry describes a phase 2, randomized, parallel-group comparison with enrollment of 261 participants. It lists schizophrenia as the condition and records the study as completed. The official title describes a long-term, multicenter, randomized, open-label comparative safety study. The registry also lists single masking; it does not identify which party was masked, and the relation between that listing and the open-label description is not reported in the registry record.

Participants were assigned to LY2140023 or atypical antipsychotic standard-of-care treatment. LY2140023 was administered orally twice daily at 80 mg/day. The standard-of-care medications were aripiprazole at 20 mg/day, olanzapine at 15 mg/day, or risperidone at 4 mg/day. The listed active treatment period was 24 weeks. The primary endpoint population is all randomized participants, and the posted arm summaries are 130 for LY2140023 and 131 for standard-of-care treatment.

The primary objective was to assess time to discontinuation due to lack of tolerability, defined in the registry as discontinuation due to adverse events. The primary timeframe was baseline through 24 weeks. Participants who completed active treatment were eligible to continue into an optional 28 weeks of treatment extension, which assessed key safety and efficacy measures. The extension was optional, so its outcomes concern a continuation group rather than necessarily the full randomized population.

The registry records a start of 2009-03 and primary completion of 2010-05. The trial’s original sample-size justification, planned interim analyses, detailed randomization procedures, and the identity of any masked party are not reported in the registry record. These omissions limit how fully the design and analysis conduct can be reconstructed from the registry information.

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: The study included an antipsychotic drug taper/discontinuation (Study Period I) and a 3-day, single-blind placebo lead in (Study Period II). Study Period III was a 24-week active treatment phase. Participants who completed Study Period III were eligible to continue in an optional, 28-week, active treatment extension phase (Study Period IV).

Study Period III

Reasons for not completing this period:

Study Period IV

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants.

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 Time to Discontinuation Due to Adverse Event (AE), measured over Baseline through 24 weeks.

The primary endpoint is time to discontinuation due to an adverse event, measured from baseline through 24 weeks. In this study, discontinuation attributed to an adverse event was the operational definition of lack of tolerability. The endpoint therefore focuses on how long participants remained on treatment before stopping for an adverse-event reason. It does not count every adverse event as a discontinuation, and it does not represent all reasons for stopping treatment unless they were specifically due to an adverse event.

Time-to-event outcomes are useful when both whether an event occurs and when it occurs matter. They can account for differences in event timing and, when handled with appropriate methods, can use information from participants whose event status is not observed through the entire follow-up period. The registry’s primary summary also presents mean time in days for each arm: 93.02 days for LY2140023 and 116.9 days for standard-of-care treatment, with standard errors of 5.95 and 5.73, respectively. Those summaries provide an arm-level description, but a mean alone does not show the distribution of event times or explain how censoring was treated.

The sponsor’s reported primary analysis uses a log-rank method, which compares event-time experience across groups rather than simply comparing arm means. The posted numerical summaries permit a separate unadjusted comparison of means, but that is not the same analysis as the sponsor’s time-to-event test. The registry does not report the event-time curves, the number of discontinuations due to adverse events, or enough detail to reproduce the sponsor’s analysis. The endpoint is therefore best read alongside the sponsor’s reported result and with attention to the difference between a mean-based recalculation and a time-to-event analysis.

7. Results Posted in the Registry

• LY2140023: mean 93.02 versus 116.86 on control; difference -23.84 (95% CI -40.03 to -7.65), standardized effect d = -0.36; z-test p = 0.004.

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 Potentially Clinically Significant Changes in QT Intervals Electrocardiograms (ECGs)

• 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:
  • LY2140023: 2.5% (3/118) versus 3.1% (4/127) on control; risk difference -0.6 percentage points (95% CI -4.8 to 3.6), risk ratio 0.81 (0.18–3.53); unadjusted two-proportion z-test p = 0.776.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between LY2140023 and the control. The interval gives the range of effects that remain compatible with the data.

Change From Baseline in Weight at 52 Weeks Endpoint

• Endpoint type: Continuous: a mean value per participant, compared as a difference in means.
• Test: Unadjusted two-sample z-test for the difference in means, using the posted standard errors.
• Result:
  • LY2140023: mean -4.76 versus 3.88 on control; difference -8.64 (95% CI -12.33 to -4.95), standardized effect d = -0.58; z-test p = <0.001.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for LY2140023 than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

Number of Participants With Potentially Clinical Significant Change in Fasting Glucose Level

• 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:
  • LY2140023: 11.3% (8/71) versus 16.7% (13/78) on control; risk difference -5.4 percentage points (95% CI -16.5 to 5.7), risk ratio 0.68 (0.30–1.53); unadjusted two-proportion z-test p = 0.344.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between LY2140023 and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes address several different safety and monitoring domains rather than one unified hypothesis. The registry lists 25 secondary outcomes, including electrocardiogram measures, neurological examination, blood pressure, weight, pulse, fasting glucose, and lipids. Their timeframes are not uniform: some are described through 24 weeks, while others extend to 52 weeks. Results across these domains therefore do not form a single direct replication of the primary time-to-discontinuation comparison.

The detailed posted comparisons illustrate a mixed pattern rather than a consistent signal across all outcomes. Some estimates are close to no difference and have intervals compatible with differences in either direction; another continuous comparison has an interval excluding zero. These outcomes measure different quantities and use different denominators, so a direction in one measure cannot be generalized to the full safety profile or treated as an explanation for discontinuation. The registry’s outcome list also includes measures for which detailed comparative estimates are not reported in the registry record.

Secondary analyses generally carry less evidential weight than a prespecified primary analysis, particularly when many outcomes are examined. A multiplicity adjustment would address the increased chance of chance findings across repeated tests, but whether one was planned or used is not reported in the registry record. The unadjusted secondary p-values and intervals should consequently be viewed as exploratory descriptions rather than as independent confirmation of the primary finding.

These outcomes can provide context about what was monitored and where estimates may warrant further study. They do not overturn or settle the primary result, especially because the primary sponsor analysis is a log-rank test and the secondary outcomes concern different endpoints. Broad conclusions require attention to the individual outcome definitions, available denominators, precision, follow-up, and analysis methods rather than counting how many estimates point in one direction.

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

• Number of Participants With Shift From Baseline to Maximum Post-Baseline Grading in Electroencephalograms (EEGs)
• Number of Participants With Potentially Clinically Significant Changes in QT Intervals Electrocardiograms (ECGs)
• Number of Participants With Treatment-Emergent Change in Neurological Examination
• Change From Baseline in Blood Pressure (BP) at 52 Weeks Endpoint
• Change From Baseline in Weight at 52 Weeks Endpoint
• Change From Baseline in Pulse Rate at 52 Weeks Endpoint
• Number of Participants With Potentially Clinical Significant Change in Fasting Glucose Level
• Number of Participants With Potentially Clinical Significant Change in Lipids Level

9. Statistical Interpretation

10. Statistical Methods Explained

What does a mean comparison describe?

A mean summarizes the average value in an arm. For the primary endpoint, the registry gives mean time to discontinuation in days and a standard error for each arm. The recalculated comparison subtracts the standard-of-care mean from the LY2140023 mean and reports an unadjusted mean difference. A negative difference indicates that the posted LY2140023 mean is lower. This simple summary can be useful for describing the posted values, but it does not preserve all the information in individual event times.

The standard error describes the estimated uncertainty in an arm mean. It is not the spread of participant-level observations. A standard error is generally smaller when the estimate is more precise, while the variability among individual participants can be much larger. The registry reports standard errors for the primary means and the weight means, but does not provide participant-level observations or the full distribution of those outcomes.

Why is a log-rank test relevant?

A log-rank test is a standard method for comparing time-to-event patterns between groups. It uses the ordering of observed event times and can incorporate participants whose follow-up ends before an event is recorded, subject to the analysis assumptions. That makes it different from comparing the group means as if each participant had a fully observed event time. The registry reports a log-rank p-value for the primary endpoint, but does not report the underlying event counts, a hazard ratio, an interval for the treatment contrast, or details about censoring. It is therefore not possible to recreate or fully audit that result from the posted summaries.

In general, a time-to-event test evaluates evidence about group differences in event experience; it does not directly communicate the size of a difference in days. Its p-value also does not show the probability that a treatment is better or worse. The registry’s sponsor-reported p = 0.184 is the relevant primary test result, while the recalculated p = 0.004 pertains to an unadjusted comparison of means. Because the analysis methods and estimands differ, those p-values answer different statistical questions.

How should intervals and p-values be understood?

A confidence interval describes a range of values compatible with the data and the method used to calculate it. Its width conveys precision: a wider interval allows a broader range of effects, while a narrower interval indicates greater precision under the model. For a difference in means, an interval spanning zero is compatible with no mean difference under the specified calculation. For a risk difference, an interval spanning zero is likewise compatible with no difference in risk. An interval that excludes the null value is not, by itself, proof of clinical importance or freedom from bias.

A p-value measures how unusual the data or a more extreme result would be under a specified null hypothesis and statistical model. It is not the chance that the null hypothesis is true, and it does not measure the size or importance of an effect. The recalculated comparisons are labelled unadjusted. They do not account for the number of outcomes examined or other possible analysis choices. The sponsor’s primary p-value is a log-rank result, which should be interpreted as the test reported for the primary time-to-event question.

What do the binary comparisons estimate?

For binary secondary outcomes, the posted arm values are numbers of participants meeting the outcome among those with an outcome denominator. A risk difference contrasts the proportions in the two groups; negative values in these comparisons mean the observed proportion is lower in the LY2140023 group. The accompanying interval indicates how uncertain that difference is. When an interval includes zero, the data are compatible with differences in either direction under the calculation. Small event counts or limited denominators can produce imprecise estimates.

For continuous secondary outcomes, the comparison is expressed as a difference between arm summaries. The weight outcome is reported as a least-squares mean change with standard errors. Least-squares means are model-based adjusted summaries, but the model details and covariates are not reported in the registry record. The recalculated comparison is therefore a simple comparison based on the posted summaries, not an independent reproduction of an adjusted model.

Why can over-dispersion matter?

General statistical explanation: over-dispersion occurs when observed variability is greater than a model assumes. If a model underestimates variability, standard errors can be too small and intervals too narrow, making evidence appear more precise than it is. Conversely, accounting for extra variability can widen uncertainty. Whether over-dispersion is relevant depends on the outcome and model. The registry does not report a model specification that would allow an assessment of over-dispersion for the posted endpoints.

The recalculated values are unadjusted calculations from posted summary statistics. They are not a replacement for an analysis that uses participant-level data, incorporates the original analysis plan, or accounts for covariates and missing observations. In particular, a recalculated mean difference cannot stand in for a time-to-event analysis when the registry’s primary endpoint is defined by event timing. The distinction is important for interpreting both the magnitude and the evidential strength of the results.

11. Confidence Intervals and Precision

The recalculated primary mean difference is -23.84 days, with an interval from -40.03 to -7.65. This interval excludes zero under the unadjusted mean-comparison calculation and indicates that the posted mean-based estimate is more precise than a result whose interval spans both negative and positive differences. It still allows a range of possible mean differences; it should not be read as a guaranteed effect or as the sponsor’s time-to-event estimate.

The sponsor reports p = 0.184 from a log-rank analysis, but the registry does not report an effect estimate or interval for that analysis. Without those quantities, the precision of the sponsor’s primary comparison cannot be judged from the p-value alone. The absence of an interval is especially relevant because a p-value does not state which effect sizes remain compatible with the data.

For the secondary comparisons, precision varies. The QT interval risk-difference interval extends from -0.048 to 0.035, and the fasting-glucose risk-difference interval extends from -0.165 to 0.057; both include zero and allow differences in either direction. The weight mean-difference interval extends from -12.33 to -4.953. These intervals arise from different outcomes and summary methods, so their widths should not be compared as if they were on a common scale.

Precision is affected by the amount of usable information, outcome frequency, variability, follow-up, and analysis method. The denominators for secondary outcomes differ from those in the primary summaries, and the registry does not explain all reasons for those differences. The confidence intervals do not account for every possible source of uncertainty, such as unreported missing-data procedures, model choices, or multiplicity. Accordingly, they describe statistical uncertainty under the calculations used, not the full uncertainty about the study’s conclusions.

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 table and curve translate an observed effect into planning illustrations under stated assumptions. For this continuous primary endpoint, the figures are participants per arm, not total events. Time-to-event planning displays use a total number of events, while the other planning displays use participants per arm. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect to detect is a stated fraction of the observed effect.

The 80% power column gives the illustrative number of participants per arm under which the calculation would have an 80% chance of detecting the specified effect if the assumptions and effect were true. The 90% power column describes the corresponding planning target for a 90% chance. These are planning concepts, not guarantees; actual results depend on the true effect, variability, adherence, outcome availability, and analysis method. The illustrations are based on the observed effect and are not the original trial’s design justification. The original sample-size assumptions are not reported in the registry record.

The primary planning illustration gives 123 participants per arm for 80% power and 165 participants per arm for 90% power when planning for 100% of the observed effect. Planning for 75% of the observed effect gives 219 participants per arm for 80% power and 293 participants per arm for 90% power. Planning for 50% of the observed effect gives 492 participants per arm for 80% power and 659 participants per arm for 90% power. Smaller effects need more participants because they are harder to distinguish from random variation.

This is not observed power. It is a forward-looking illustration based on an effect estimate after the trial, and it does not measure the probability that the study’s finding is true or that a future study will reproduce it. The primary discrepancy between the sponsor’s log-rank test and the recalculated mean comparison is another reason to avoid treating these figures as a definitive design prescription.

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, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -23.840; pooled standard deviation σ = 66.717.
2. n = 2 × (1.96 + 0.842)² × 66.717² / -23.840² = 122.9, rounded up to 123 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
delta <- -23.84                      # observed difference in means
sigma <- 66.7168341100502                     # pooled standard deviation
n_per_arm <- function(power, frac = 1) {
  ceiling(round(2 * (z_a + qnorm(power))^2 * sigma^2 / (delta * frac)^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 (t-based, so very slightly larger)
cat("power.t.test:", round(power.t.test(delta = delta, sd = sigma, power = 0.80)$n, 1), "\n")

# Simulation: exact two-sample t-test using simulated means and variances
set.seed(845026)
n    <- n_per_arm(0.80)
reps <- 20000
m_diff <- rnorm(reps, delta, sigma * sqrt(2 / n))             # difference in sample means
s2     <- sigma^2 * rchisq(reps, 2 * (n - 1)) / (2 * (n - 1))   # pooled sample variance
tstat  <- m_diff / sqrt(s2 * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(tstat) > qt(0.975, 2 * (n - 1))), reps))

Output from running the code

100% of observed effect: n per arm = 123 (80% power), 165 (90% power)
 75% of observed effect: n per arm = 219 (80% power), 293 (90% power)
 50% of observed effect: n per arm = 492 (80% power), 659 (90% power)
power.t.test: 123.9 
Simulated power at n = 123 per arm: 0.797 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-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 reported primary analysis is a log-rank test with p = 0.184. The recalculation from posted arm means gives a mean difference of -23.84 days, an interval from -40.03 to -7.65, and p = 0.004. These results differ because the log-rank test and mean comparison use different summaries and methods. The sponsor’s analysis takes precedence as the trial’s own analysis of the primary time-to-event endpoint.

The mean-based calculation is useful as a transparent check on the posted summaries, but it is not a reproduction of the sponsor’s time-to-event analysis. The registry does not provide the sponsor’s effect estimate, interval, event counts, event-time curve, or detailed handling of censoring. Those omissions prevent a more complete reconciliation of the two analyses. The difference should therefore be stated plainly rather than resolved by choosing the result that appears more statistically persuasive.

The sponsor’s reported p-value does not provide evidence of a statistically clear separation in event-time experience under the reported log-rank test. It also does not establish that the groups are equivalent or that no meaningful difference exists. Without an effect estimate and interval for the sponsor’s analysis, the plausible magnitude of differences remains difficult to assess. The recalculated mean comparison should remain secondary to the sponsor’s own analysis when interpreting the registered primary endpoint.

15. Multiplicity, Missing Data and Analysis Populations

The registry lists 26 outcome measures and 25 secondary outcomes. Examining multiple outcomes creates a general multiplicity concern: even if no real differences exist, repeated testing can produce apparently unusual results by chance. Statistical procedures can control this issue, but they may affect thresholds or interval interpretation. Whether the trial planned or applied a multiplicity adjustment is not reported in the registry record. The secondary comparisons should therefore be treated cautiously, particularly when considered individually.

The registry identifies the primary population as all randomized participants. It does not provide the complete analysis-set definitions for the secondary outcomes, nor does it state how missing observations were handled. The posted denominators vary across outcomes. Those differences may reflect outcome availability or analysis-specific eligibility, but their reasons are not reported in the registry record. It is not possible to infer that missingness was balanced, unrelated to outcome, or adequately addressed.

For a time-to-event endpoint, censoring occurs when an event is not observed during a participant’s available follow-up. Appropriate handling depends on the timing and reasons for the end of follow-up. The registry’s reported log-rank method indicates a time-to-event comparison, but detailed censoring rules and follow-up completeness are not reported in the registry record. The posted mean summaries do not supply that information.

The original statistical analysis plan, sample-size justification, interim-analysis plans, and detailed model specifications are also not reported in the registry record. Without these details, readers cannot verify which analyses were prespecified, whether adjustments were used, or how decisions about missing data were made. These are documentation limits, not evidence that a particular procedure was or was not followed. They matter because analysis choices can affect both estimates and uncertainty, especially when there are multiple outcomes and incomplete follow-up.

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 5% reporting threshold.

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: schizophrenia (8.3%), psychotic disorder (2.9%), abnormal behaviour (2.1%), hallucination (2.1%), suicidal ideation (2.1%).

The registry reports serious adverse event figures by treatment group. For LY2140023, 22 participants were affected among 130 at risk. The separately listed standard-of-care groups report 6 among 48 for olanzapine, 4 among 48 for aripiprazole, and 2 among 35 for risperidone. These are counts of participants affected, not counts of every serious event, and the denominators differ across the listed groups. The registry does not provide a combined serious-event count for the standard-of-care group in these entries.

The registry also lists the most frequent serious adverse event terms with percentages: Schizophrenia at 8.3%, Psychotic disorder at 2.9%, Abnormal behaviour at 2.1%, Hallucination at 2.1%, Suicidal ideation at 2.1%, and Pneumonia at 1.5%. These terms describe reported serious adverse events; the listing does not establish that treatment caused them. The registry does not report death counts in these entries, so deaths should not be interpreted as absent.

Comparisons of these counts require caution because group sizes and exposure may differ, and the posted figures alone do not adjust for follow-up duration or other factors. Serious adverse events are one part of safety reporting and do not summarize all adverse events or overall tolerability. The primary endpoint separately concerns discontinuation attributed to adverse events.

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

The primary endpoint is intended to describe treatment continuation when stopping is attributed to an adverse event. The sponsor’s reported log-rank analysis gives p = 0.184, while the separate recalculated comparison of posted means gives a difference of -23.84 days with an interval from -40.03 to -7.65 and p = 0.004. Because these analyses differ and the registry flags a discrepancy, the sponsor’s own analysis takes precedence. The registry does not provide enough detail to quantify the sponsor’s estimated time-to-event contrast.

The result does not show that LY2140023 is safer or less safe overall than standard-of-care treatment. Discontinuation for adverse events is a specific outcome, and the other safety measures cover different aspects of health and monitoring. Nor does a lack of a statistically clear result in the sponsor’s primary test prove that the groups are equivalent. A non-significant test can coexist with effects that are uncertain in size, especially when the analysis estimate and interval are not reported.

The posted secondary findings also do not establish a broad comparative safety conclusion. They span distinct endpoints and follow-up periods, have varying denominators, and are not reported with a multiplicity adjustment. Some estimates are imprecise, and the registry does not report detailed results for every listed secondary outcome. Serious adverse-event counts describe reported events in the listed groups but do not, by themselves, identify treatment causation or account for all differences in exposure.

These findings are therefore best understood as registry-level statistical evidence about a defined tolerability endpoint and a set of additional monitored outcomes. They do not determine how any individual should be treated, and they do not substitute for a complete clinical assessment. Their interpretation is limited by the missing details on event timing, censoring, missing data, analysis plans, and the sponsor’s effect estimate for the primary comparison.

18. Why This Trial Matters Statistically

This trial illustrates why the definition and analysis of an endpoint matter as much as its label. The primary outcome is described as time to discontinuation due to adverse events, which is a time-to-event question. A comparison of arm means in days is understandable, but it does not necessarily answer the same question as a log-rank test that compares event-time experience. The different results make clear that statistical summaries are not interchangeable simply because they concern the same endpoint.

The registry also shows why effect estimates and intervals are important alongside p-values. The sponsor’s reported analysis provides a p-value without an effect estimate or interval, leaving the size and precision of the primary time-to-event comparison unclear. By contrast, the recalculated mean comparison includes an estimated difference and interval, but it is not the sponsor’s primary method. Readers need both the right estimand and enough information to judge uncertainty.

The secondary outcomes further demonstrate how safety trials can involve many domains, varying follow-up periods, and different denominators. Such breadth can be useful for surveillance and description, but it increases the need for clear prespecification and careful treatment of multiplicity. A single nominally notable secondary result should not be promoted as decisive evidence when adjustment and analysis plans are not reported.

Finally, planning illustrations based on an observed effect are not a substitute for the original design assumptions or evidence of achieved power. They can help convey how study size relates to the effect a future study aims to detect, while remaining conditional on assumptions. The registry record highlights the importance of reporting the analysis plan, event counts, uncertainty, missing-data procedures, and outcome-level denominators so that statistical results can be interpreted with appropriate confidence.

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 primary result is difficult to fully evaluate because the registry gives the sponsor’s log-rank p-value but not its effect estimate, interval, event count, event-time curve, or detailed censoring rules. The available arm means permit an unadjusted recalculation, but this is not a replacement for the sponsor’s time-to-event analysis. The discrepancy between those results limits any simple summary of the primary finding.

The registry’s design descriptions are not fully aligned: the official title says open-label, while the masking field lists single masking. Which party, if any, was masked is not reported in the registry record. Detailed randomization procedures and the original analysis plan are also not reported. The registry does not explain whether or how any interim analyses were planned.

Information on missing observations and analysis populations is limited. The outcome denominators vary, but reasons for the differences are not reported. For the optional extension, the registry says participants who completed active treatment were eligible to continue, which means extension outcomes may reflect a selected subset; the exact extension analysis population and missing-data methods are not reported.

There are many outcomes, but multiplicity procedures are not reported. The secondary endpoints use different measures and timeframes, making broad comparisons difficult. The safety entries provide counts but do not establish causality, and death counts are not reported in those entries. These limitations concern the available registry documentation and do not establish that a particular design or analysis procedure was absent from the conduct of the trial.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Safety Study Comparing LY2140023 to Atypical Antipsychotic Standard Treatment in Schizophrenic Patients — NCT00845026.
• Registry record last updated 2022-11-08; status last verified 2022-10.
• Results first posted 2022-09-01.
• 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 phase 2 randomized parallel-group study of LY2140023 versus atypical antipsychotic standard-of-care treatment in participants with schizophrenia. Enrollment was 261. LY2140023 was given orally twice daily at 80 mg/day, while standard-of-care treatment involved aripiprazole, olanzapine, or risperidone. The primary outcome was time to discontinuation due to adverse events through 24 weeks.

The sponsor reported a log-rank p-value of 0.184. A separate unadjusted recalculation from posted means gives a difference of -23.84 days, with an interval from -40.03 to -7.65 and p = 0.004. Because the methods differ and the registry identifies a discrepancy, the sponsor’s analysis takes precedence. The record also lists secondary safety outcomes and serious adverse-event counts, but multiplicity, missing-data procedures, and several details needed to fully evaluate the primary analysis are not reported in the registry record.