Lurasidone HCl - A Long Term Phase 3 Study of Patients With Chronic Schizophrenia: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Lurasidone HCl - A Long Term Phase 3 Study of Patients With Chronic Schizophrenia (NCT00789698), a randomized phase 3 study. The primary outcome is Relapse of Psychotic Symptoms.

2. Clinical Question

3. Trial Design

This was a Phase 3 randomized, double-blind, parallel-group, active comparator-controlled study titled the PEARL 3 Extension Study. It studied lurasidone and quetiapine XR among outpatients with chronic schizophrenia. The registry describes it as a long-term extension following an acute study: the relapse-analysis population consisted of participants who had responded to 6 weeks of treatment in NCT00790192 and then entered the present study. The extension’s primary endpoint timeframe was 12 months.

The registry records an enrollment of 240. The posted primary-outcome groups are Lurasidone-Lurasidone, with 139 participants and 29 events, and Quetiapine-Quetiapine, with 79 participants and 21 events. These are the endpoint-specific counts; the registry does not explain the difference between these group counts and the overall enrollment. It also does not provide further detail about allocation procedures, treatment adherence, discontinuations, or participant flow.

The study started in 2008-12, had primary completion in 2011-07, and is listed as completed. Sumitomo Pharma America, Inc. is the sponsor. Three secondary outcomes are listed: change in computerized cognitive scores at Month 6, and change in PANSS and CGI-S scores at Month 12. The registry does not report the detailed methods for those secondary outcomes in the information available for this analysis. It also does not report the original sample-size justification or the trial’s full statistical analysis plan.

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: Patients transitioned from the acute phase study (D1050233 -NCT00790192) to the current study (D1050234) in a non-randomized fashion.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: The population for relapse analyses is the relapse population which consists of those subjects who are enrolled in the present study, demonstrated response to 6 weeks of treatment with either lurasidone or quetiapine XR in study D1050233-NCT 00790192, and who took at least one dose of study medication in the present study.

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 Relapse of Psychotic Symptoms, measured over 12 Months.

The primary endpoint is relapse of psychotic symptoms during the 12-month double-blind treatment period. The registry defines relapse as the earliest occurrence of one of several events: a specified degree of worsening in PANSS total score together with a specified CGI-S threshold, rehospitalization for worsening psychosis, or emergence of suicidal ideation, homicidal ideation, and/or risk of harm to self or others. The outcome is described as time to relapse, so the timing of the first qualifying event matters, not only whether a participant had an event.

Time-to-event outcomes are useful when participants can experience an event at different points during follow-up, or when follow-up does not end at the same point for everyone. A Cox proportional hazards model compares the event rates over follow-up while accounting for the timing of events and, in this registry analysis, includes country as a covariate. The model’s hazard ratio summarizes the relative event rate over time under its assumptions; it is not the same quantity as a simple difference in the proportions with an event.

The posted group counts also permit an unadjusted comparison of event proportions. That comparison is informative as a straightforward description of the observed counts, but it does not reproduce the registry’s time-to-event model. The registry does not report the detailed handling of censoring or the operational rules used to determine the date of relapse.

7. Results Posted in the Registry

• Quetiapine-Quetiapine: 26.6% (21/79) versus 20.9% (29/139) on control; risk difference 5.7 percentage points (95% CI -6.1 to 17.6), risk ratio 1.27 (0.78–2.08); unadjusted two-proportion z-test p = 0.334.

8. Statistical Interpretation

9. Statistical Methods Explained

The registry’s stated primary method is a Cox proportional hazards model. This is a time-to-event method: it uses the time until relapse, rather than reducing follow-up to a yes-or-no outcome alone. Participants without a recorded relapse may contribute information up to the point they leave observation or follow-up ends, depending on the analysis rules. Those rules, including the detailed censoring procedures, are not reported in the registry record.

What does the hazard ratio represent?

A hazard is the event rate at a particular point in follow-up among participants who have not yet had the event. A hazard ratio compares the hazards between groups over the follow-up period. It is not a risk ratio, a risk difference, or a direct statement about the proportion of participants who relapsed. The reported hazard ratio is 0.728, with a 95% confidence interval from 0.410 to 1.295. The interval includes the no-difference value, so the estimate is imprecise enough to be compatible with effects in either direction under this model.

The Cox model is commonly described as proportional hazards because it assumes that the relative hazard between groups is sufficiently stable over time for a single hazard ratio to be a useful summary. Whether this assumption was checked, and what diagnostics were used, are not reported in the registry record. The primary outcome description says country was included as a covariate. Other adjustment variables are not reported.

Why is there also a recalculated risk difference?

The posted group counts allow a simple, unadjusted comparison of the proportions with an event. That recalculation gives a risk difference of 0.057, with a 95% confidence interval from -0.061 to 0.176 and p = 0.334. It is a transparent summary of the recorded counts, but it does not use event timing or adjust for country. It is therefore not a replacement for the sponsor’s stated Cox model. The registry’s model-based estimate takes priority when describing the trial’s reported primary analysis.

A confidence interval describes uncertainty around an estimate under the method used to calculate it. Wider intervals indicate less precision, while narrower intervals indicate greater precision. An interval that includes the no-difference value does not show that the groups are identical; it indicates that the data do not precisely distinguish the estimate from no difference under that analysis. Conversely, an interval excluding that value would not by itself establish clinical importance or satisfy an unreported non-inferiority margin.

Why can a crude count comparison differ from a time-to-event analysis?

Participants may have different lengths of observed follow-up, and the timing of events can affect a time-to-event estimate. A simple proportion gives equal weight to each participant’s event status as represented in the posted counts, while the Cox model uses the event-time information and may account for censoring. Adjustment for country can also affect an estimate. The registry does not report enough detail to identify how much each feature contributes to the difference between the summaries.

The outcome is labeled as a count of participants with an event in the posted group results, while its definition and sponsor analysis are explicitly time-to-event. This distinction is important: the counts are useful for a plain description of events, but they do not contain all the information used by the Cox model. The record does not provide the individual event times needed to reproduce that analysis.

What do the non-inferiority or equivalence label and missing p-value imply?

The registry classifies the reported analysis as non-inferiority or equivalence, but gives no margin, hypothesis specification, or decision rule. Such an analysis normally requires a prespecified boundary defining how much difference can be accepted for the intended conclusion. Without that boundary and its analysis details, the result cannot be independently interpreted as demonstrating non-inferiority or equivalence. The Cox model’s p-value is not reported; the model estimate and confidence interval should be described as recorded rather than supplemented with an inferred p-value.

The record also does not state whether there were interim analyses, how the analysis population was finalized, how missing data were treated, or whether sensitivity analyses were conducted. These omissions limit assessment of robustness. They do not invalidate the posted numbers, but they constrain what can be concluded from them.

10. Confidence Intervals and Precision

The sponsor’s Cox-model estimate is a hazard ratio of 0.728, with a 95% confidence interval from 0.410 to 1.295. The interval is broad and includes the no-difference value. It therefore allows a range of interpretations, including a lower hazard in the lurasidone group and a higher hazard, relative to the comparator, under the model. The estimate alone should not be read as a precise account of the size or direction of the treatment difference.

The unadjusted risk difference is 0.057, with a 95% confidence interval from -0.061 to 0.176. This interval also includes the no-difference value and extends to estimates in either direction. It indicates substantial uncertainty in the simple comparison of event proportions. The accompanying p-value is 0.334 for this recalculation, not for the sponsor’s Cox model.

Confidence intervals express sampling uncertainty under the method and assumptions used to calculate them; they do not include every possible source of uncertainty. For example, the registry does not report the detailed missing-data procedures, censoring rules, or checks of model assumptions. Those unreported elements can affect how confidently the estimates can be interpreted. The confidence interval should also not be confused with a non-inferiority boundary: no such margin is reported in the registry record.

11. 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 participants per arm for 80% power and participants per arm for 90% power. They are planning illustrations based on the observed effect, not the trial’s original sample-size calculation. The stated 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 the stated fraction of the observed effect. The original trial’s design assumptions are not reported in the registry record.

The 80% power column indicates the number of participants per arm estimated to give an 80% chance of detecting the assumed effect under those planning assumptions, if that effect is the one that truly applies. The 90% power column uses a higher target chance of detection and consequently requires more participants per arm for the same assumed effect. These are prospective planning quantities, not measures of what the completed trial achieved.

The rows consider planning for 100%, 75%, and 50% of the observed effect. Smaller effects need more participants to distinguish them from no difference with the same target power and significance level. The figures are based on the recalculated observed effect and should be understood as illustrations rather than a reconstruction of the study’s design. This is not observed power.

The registry’s assumptions distinguish time-to-event calculations, which give total numbers of events, from other calculations, which give participants per arm. The displayed planning figures for this comparison are expressed as participants per arm. The sample-size curve should be read as a visual expression of how the planning requirement changes as the assumed effect changes; it does not add information about the trial’s actual analysis or its original design rationale.

12. 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 = 0.209; treatment proportion p₁ = 0.266; average p̄ = 0.237.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6016 = 1.1791.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6002 = 0.5052.
4. n = (1.1791 + 0.5052)² / (0.0572)² = 867.3, rounded up to 868 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 <- 0.20863309352518                          # control proportion
p1 <- 0.265822784810127                          # 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(789698)
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 = 868 (80% power), 1,161 (90% power)
 75% of observed effect: n per arm = 1,511 (80% power), 2,022 (90% power)
 50% of observed effect: n per arm = 3,325 (80% power), 4,451 (90% power)
power.prop.test: 867.3 
Simulated power at n = 868 per arm: 0.798 (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.

13. Analyses Reported by the Sponsor

The sponsor’s reported primary analysis is the Cox proportional hazards model, with country as a covariate. It gives a hazard ratio of 0.728 and a 95% confidence interval from 0.410 to 1.295; no p-value is reported for this model. The analysis is labeled non-inferiority or equivalence, but the margin and decision rule are not reported in the registry record.

The recalculated result is an unadjusted risk difference based on the posted event counts: 0.057, with a 95% confidence interval from -0.061 to 0.176 and p = 0.334. It is not the same estimand or method as the Cox model. The recalculation does not account for event timing or country adjustment, and its p-value should not be attributed to the sponsor analysis.

The registry does not flag a discrepancy between the posted results and the recalculation. The sponsor’s model remains the appropriate result to cite as the trial’s reported primary analysis, while the unadjusted calculation is a supplementary description of the counts and their uncertainty.

14. Multiplicity, Missing Data and Analysis Populations

The registry lists four outcome measures in total and identifies three secondary outcomes: change in computerized cognitive scores at Month 6, and change in PANSS and CGI-S scores at Month 12. The primary outcome is time to relapse over 12 months. The record does not report whether the primary and secondary outcomes were treated as a single family for inference, whether multiplicity adjustments were made, or whether a hierarchy of testing was prespecified.

Multiplicity arises when several outcomes or comparisons are tested. Without an adjustment or an explicit testing strategy, the chance of at least one apparently notable result can increase across multiple tests. This general concern does not establish that the reported analyses were adjusted or unadjusted; that information is not reported in the registry record. The number of posted analyses is one, but the registry does not provide enough detail to determine the full set of analyses considered.

Missing-data methods are also not reported in the registry record. The relapse population definition says participants had to be enrolled, demonstrate response in the earlier study, and take at least one dose, but the description is truncated. The record does not specify which analysis population was used for the Cox model, how participants who discontinued were handled, or how outcome information that was unavailable was treated. For time-to-event analysis, details about censoring and follow-up are particularly relevant; those details are not available in the record.

The record does not report interim analyses, sensitivity analyses, protocol deviations, or the full statistical analysis plan. It also does not state whether the proportional hazards assumption was assessed. These are reporting gaps, not evidence that the procedures were or were not performed. They limit independent assessment of the analysis and of how robust the results are to choices about populations, missing information, or model assumptions.

15. 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 (10.6%), psychotic disorder (7.1%), agitation (2.4%), cholelithiasis (1.8%), completed suicide (1.4%).

The registry reports serious adverse event figures for four listed groups. For Lurasidone 80mg, 9 of 72 participants at risk had a serious adverse event; for Lurasidone 160 mg, 6 of 79; for Placebo-Lurasidone, 2 of 56; and for Quetiapine-Quetiapine, 17 of 85. These are group-level counts as recorded. The registry does not provide a death count for these groups.

The listed serious adverse event terms are Schizophrenia (10.6%), Psychotic Disorder (7.1%), Agitation (2.4%), Cholelithiasis (1.8%), Completed Suicide (1.4%), and Schizophrenia, paranoid type (1.4%). The record does not provide counts for each term alongside these percentages or explain whether terms can overlap within a participant. They should not be treated as mutually exclusive categories or summed to describe a total.

The safety groups do not match the two primary-outcome group labels in a way that the registry explains. The record does not describe how these safety groupings relate to the relapse population or to the endpoint counts. These figures describe reported serious adverse events; by themselves, they do not establish a comparative safety effect.

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

The posted data show relapse events in both primary-outcome groups. The unadjusted event-proportion comparison has a risk difference of 0.057, with a 95% confidence interval from -0.061 to 0.176. The sponsor’s time-to-event analysis reports a hazard ratio of 0.728, with a 95% confidence interval from 0.410 to 1.295. Both intervals include their respective no-difference values, and the registry does not report a p-value for the Cox model.

These results do not establish that the groups are identical. The intervals permit a range of possible differences, and the estimates are not precise enough to identify a clear direction from the posted information. Nor do the results, on their own, establish non-inferiority or equivalence. The registry labels the analysis as non-inferiority or equivalence, but the required margin and decision rule are not reported. A label is not a substitute for those details.

The analysis methods also answer different questions. The Cox model uses time to relapse and includes country as a covariate. The recalculated risk difference compares event proportions without adjustment. The latter is useful for describing the counts but does not replace the model-based primary analysis. The absence of a reported p-value for the Cox model should remain explicit rather than being filled in from the recalculation.

The registry’s results should be interpreted within the defined relapse population, which involved prior response to treatment in the earlier study and at least one dose in the extension. The eligibility description is incomplete, and the handling of missing information, censoring, and analysis populations is not reported. The posted data therefore support a cautious description of the observed outcomes and their uncertainty, not a broader claim about treatment effects beyond this trial population and its recorded methods.

17. Why This Trial Matters Statistically

This trial illustrates why a clinical outcome’s definition and analysis method need to be read together. Relapse is defined using several possible clinical events, and the primary question concerns time to the first qualifying event. The Cox model is therefore aligned with event timing, whereas a comparison of the posted counts compresses that information into whether an event was recorded. Both summaries can be useful, but they are not interchangeable.

It also demonstrates how the choice of estimand affects interpretation. The sponsor’s hazard ratio comes from a model that includes country as a covariate; the recalculated risk difference is unadjusted. Reporting both makes the distinction visible and gives readers a direct description of the event counts alongside the registered model result. The confidence intervals are essential: point estimates alone would give an incomplete impression of the uncertainty.

The non-inferiority or equivalence label raises a further statistical issue. Interpretation of that type of analysis depends on a prespecified margin and decision rule. Because those details are not reported in the registry record, the label cannot be used to infer that the criterion was satisfied. This is a practical reminder that analysis labels need the underlying hypotheses and boundaries to be interpretable.

Finally, the record identifies a defined extension-study population and several secondary outcomes, but leaves important analysis details unreported. The limits around missing data, censoring, multiplicity, and analysis populations are relevant to independent evaluation. Statistical interpretation is strongest when it separates observed counts, model-based estimates, uncertainty, and the boundaries of what the registry documents.

18. 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 outcome groups have different posted participant counts, and the registry does not explain the relationship between those counts and the overall enrollment of 240. It does not provide a participant flow account, reasons for exclusion, discontinuation information, or endpoint-specific follow-up details. The relapse population description is truncated, limiting assessment of exactly who was included.

The registry gives a Cox-model estimate and confidence interval but does not report the model p-value, detailed censoring rules, checks of the proportional hazards assumption, or additional covariates beyond country in the primary outcome description. It labels the analysis non-inferiority or equivalence but does not report the margin, hypothesis specification, or decision rule. These omissions prevent a full independent evaluation of that claim.

The unadjusted recalculation uses posted counts and does not reproduce the time-to-event analysis. It should not be interpreted as a substitute for the sponsor’s model. The record does not describe the original sample-size rationale, interim analyses, analysis populations, missing-data procedures, sensitivity analyses, or multiplicity strategy. The planning illustrations are based on the observed effect and stated planning assumptions, not on the original trial design.

Safety reporting is presented in groupings that differ from the primary-outcome labels, without an explanation of how the populations align. Death counts are not reported in the registry record, and event-term counts are not given alongside the listed percentages. The record also does not clarify whether serious adverse event terms overlap. These limitations constrain interpretation of the safety information as a comparison between the primary endpoint groups.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: Lurasidone HCl - A Long Term Phase 3 Study of Patients With Chronic Schizophrenia — NCT00789698.
• Registry record last updated 2015-06-12; status last verified 2015-05.
• Results first posted 2012-08-21.
• 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.

22. Record Summary

NCT00789698 is a completed Phase 3 randomized, double-blind, parallel-group active comparator-controlled extension study of lurasidone and quetiapine XR in chronic schizophrenia. The registry lists enrollment of 240 and a 12-month primary endpoint of time to relapse of psychotic symptoms. The sponsor is Sumitomo Pharma America, Inc.

The posted primary analysis is a Cox proportional hazards model with country as a covariate. It reports a hazard ratio of 0.728 and a 95% confidence interval from 0.410 to 1.295; the model p-value is not reported. An unadjusted recalculation from the posted event counts gives a risk difference of 0.057, with a 95% confidence interval from -0.061 to 0.176 and p = 0.334. The registry does not report the non-inferiority or equivalence margin, missing-data procedures, or full analysis plan.