Soy Protein/Effexor Hormone Therapy for Prostate Cancer: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Soy Protein/Effexor Hormone Therapy for Prostate Cancer (NCT00354432), a randomized phase 3 study. The primary outcome is Hot Flash Symptom Severity Score.

2. Clinical Question

3. Trial Design

This was a randomized phase III trial with a factorial model and quadruple masking. The registry lists enrollment of 120. The four arms crossed soy protein/isoflavones powder with venlafaxine: placebo in both forms, soy with a placebo pill, venlafaxine with placebo powder, and soy plus venlafaxine. Each posted primary-outcome arm summary lists 30 participants.

Participants in the placebo arm received placebo pill and powder once daily. In the soy arm, the pill was placebo and the powder contained soy protein/isoflavones. In the venlafaxine arm, the pill contained venlafaxine and the powder was placebo. The combination arm received venlafaxine pill and soy powder. The registry does not report additional details about the masking procedures or which roles were masked.

The trial started on 2007-02-01, had a primary completion date of 2010-08-01, and is listed as terminated. The reason for termination is not reported in the registry record. The registry identifies quality of life as a secondary outcome, with an outcome timeframe of 12 weeks. It also lists a mixed models analysis for the primary outcome and an ANCOVA for quality of life. Further details of the statistical plan, including the original sample-size justification and any planned interim analyses, are not reported in the registry record.

The four-arm structure permits comparisons of each active-treatment arrangement with placebo, as well as consideration of the two treatment components. However, the reported primary test is an overall test across the four groups. The registry does not report separate factorial main-effect or interaction estimates, so the available results should not be read as a complete estimate of each component's independent contribution.

4. Participant Flow and Accounting

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

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants were analyzed in a repeated measures mixed model. This allowed inclusion of all study 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 Hot Flash Symptom Severity Score, measured over 12 weeks.

The primary endpoint was the hot flash symptom severity score at 12 weeks relative to baseline, expressed as a percent of baseline score. The symptom diary score is defined as the sum of the number of hot flashes multiplied by their severity. Severity is coded from 0 for none to 3 for severe, and the number of hot flashes can be zero or greater. The endpoint has a range from 0 to infinity; lower values represent a better outcome.

Expressing the 12-week score relative to baseline places the follow-up measure in relation to each participant's starting symptom score. This can make a change in symptom burden easier to interpret than a follow-up score without baseline context. The registry reports least-squares means and standard errors for the arm summaries. Least-squares means are model-based averages rather than simple raw averages, and the registry says that all randomized participants were analyzed in a repeated measures mixed model to allow inclusion of all study participants.

The endpoint was selected to quantify both how often hot flashes occurred and their recorded severity, rather than frequency alone. The registry describes the score construction and the preferred direction but does not report further details about diary completion, handling of diary data, or the model's covariance structure.

7. Results Posted in the Registry

• Arm II - Soy: mean 77.20 versus 52.30 on control; difference 24.90 (95% CI -6.16 to 55.96), standardized effect d = 0.41; z-test p = 0.116.
• Arm III - Venlafaxine: mean 68.90 versus 52.30 on control; difference 16.60 (95% CI -15.19 to 48.39), standardized effect d = 0.26; z-test p = 0.306.
• Arm IV - Soy + Venlafaxine: mean 73.80 versus 52.30 on control; difference 21.50 (95% CI -10.00 to 53.00), standardized effect d = 0.35; z-test p = 0.181.

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.

Quality of Life

• 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:
  • Arm II - Soy: mean 121.50 versus 115.60 on control; difference 5.90 (95% CI -0.89 to 12.69), standardized effect d = 0.53; z-test p = 0.089.
  • Arm III - Venlafaxine: mean 114.30 versus 115.60 on control; difference -1.30 (95% CI -8.38 to 5.78), standardized effect d = -0.12; z-test p = 0.719.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm II - Soy and the control. The interval gives the range of effects that remain compatible with the data.
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between Arm III - Venlafaxine and the control. The interval gives the range of effects that remain compatible with the data.

The registry identifies quality of life as a secondary outcome and posts results at 12 weeks. The overall sponsor analysis is an ANCOVA with p = 0.208. The recalculated comparisons are available for soy and venlafaxine versus placebo, but the registry does not report a recalculated comparison for the combination arm. These secondary data provide additional context, but they do not alter the primary endpoint's overall mixed-model result.

The quality-of-life comparison estimates are less definitive than a simple ranking of arm means might suggest. In general, secondary outcomes can help describe whether a study's findings extend beyond its primary endpoint, but they are often less conclusive when several outcomes or comparisons are considered. The direction of a better quality-of-life score is not reported in the registry record, so numerical differences should not be characterized as favorable or unfavorable.

Unadjusted secondary comparisons carry less weight because they do not account for the chance of false-positive findings from examining multiple comparisons. The registry does not report a multiplicity adjustment or a prespecified hierarchy for secondary analyses. Quality-of-life results are therefore best treated as exploratory or supportive information, not as independent confirmation of a primary finding. The pattern does not provide a clear additional basis to resolve the uncertainty in the primary result.

The registry lists two outcome measures but only one posted secondary outcome. Further secondary endpoints and their results are not reported in the registry record. The relationship between quality of life and hot-flash symptom severity also cannot be established from these summaries alone.

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

• Quality of Life

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous: it is a numerical symptom score expressed as a percent of baseline. For a continuous endpoint, an arm mean describes its central level, while a standard error describes the estimated uncertainty in that mean. The registry reports least-squares means and standard errors, which are model-based summaries. They should not be interpreted as individual participant values or as standard deviations describing the spread of scores between participants.

What does the overall test ask?

The reported mixed models analysis gives p = .3455 for the four groups. In general, an overall test asks whether the observed group pattern is compatible with a model in which there is no group difference. A p-value describes how surprising data at least as incompatible with that model would be, if the model's assumptions held. It is not the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The registry does not report the overall test statistic, degrees of freedom, or a corresponding effect estimate.

A mixed model can account for repeated observations and can use available information across the repeated-measure structure. The registry states that the primary analysis included all randomized participants in a repeated measures mixed model, allowing inclusion of all study participants. Details such as the timing and number of repeated observations, covariance structure, missing-data assumptions, and model adjustment variables are not reported in the registry record. The phrase “all randomized participants” describes the stated analysis population; it does not establish that every participant contributed a fully observed outcome at every time point.

How do the pairwise comparisons differ?

The comparisons against placebo are recalculated, unadjusted mean differences based on the posted arm summaries. A mean difference expresses the distance between two group means in the endpoint's units. For these comparisons, the estimate is the active-treatment arm minus placebo. A positive estimate therefore corresponds to a higher score in the active arm, and the endpoint definition says that lower values represent a better outcome.

The recalculated intervals and p-values are not the sponsor's reported overall test. They address individual active-versus-placebo contrasts and use a simpler calculation from the posted summary statistics. They are unadjusted, so they do not correct for examining multiple contrasts. The registry reports that the standard errors were used to approximate pooled standard deviations for these calculations. Consequently, the recalculations are useful as a transparent comparison of arm summaries, but they do not reproduce all features of the repeated measures mixed model.

What does an interval convey?

In general, a confidence interval represents a range of effect values that are reasonably compatible with the data under the method used. Its width reflects precision: narrower intervals indicate more precise estimates, while wider intervals leave more uncertainty about the effect's size and direction. An interval that includes zero does not establish that the effect is zero. It indicates that a zero difference remains among the values compatible with the estimate and method. The registry does not state the confidence level for the recalculated intervals, so no confidence level is assigned to them in this interpretation.

What is the role of the factorial structure?

A factorial design crosses treatment components, in this case soy protein/isoflavones and venlafaxine, as reflected by the four combinations. Such a design can be used to study component effects and whether the combined effect differs from what would be expected from the components separately. Those questions require corresponding model terms and estimates. The registry reports an overall four-group mixed-model test, but does not report component main effects or an interaction estimate. The factorial label alone cannot establish that those effects were analyzed or that the study was powered for them.

The secondary quality-of-life outcome is also continuous, with least-squares means and standard errors. Its reported sponsor analysis is ANCOVA, while the posted pairwise recalculations use the arm summaries. In general, ANCOVA compares outcomes while accounting for covariates in a regression model. The covariates and details of the sponsor's model are not reported in the registry record. The reported methods therefore identify broad analysis approaches but do not provide enough detail to fully reconstruct either sponsor model.

11. Confidence Intervals and Precision

The primary recalculated intervals are broad. For soy versus placebo, the mean difference is 24.9, with an interval from -6.164 to 55.96. For venlafaxine versus placebo, it is 16.6, with an interval from -15.19 to 48.39. For soy plus venlafaxine versus placebo, it is 21.5, with an interval from -9.997 to 53.0. Each interval spans zero, so the posted summary statistics do not pinpoint whether the corresponding difference is positive or negative.

The range of compatible values also matters. The intervals include both values favoring placebo and values favoring the active arm, given that lower scores are better. Thus, the estimates are not precise enough to distinguish a possible difference from no difference or to settle the direction with confidence. Their breadth should be considered alongside the arm estimates rather than treating the point estimates as exact effects.

In general, a confidence interval communicates uncertainty about an estimated effect: its endpoints mark a range compatible with the data and calculation method. A wide interval signals limited precision, while a narrow one gives a more concentrated estimate. The confidence level for these recalculated intervals is not reported in the registry record, so the intervals are described without assigning a level. The reported least-squares means also have standard errors, which quantify uncertainty in those means, not the range of individual participant scores.

The sponsor's overall mixed-model analysis supplies a p-value but no overall effect estimate or interval in the registry record. The pairwise intervals are recalculations from summary statistics and are not replacements for intervals from the sponsor's model. Taken together, the available information supports caution about both the magnitude and direction of group differences.

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 illustrations report participants per arm for continuous-outcome comparisons; they are not total event counts. The calculations use a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The target effect is a stated fraction of the observed effect. The trial's own original design assumptions and sample-size justification are not reported in the registry record.

The 80% power column indicates the planned number of participants per arm under those assumptions to have an 80% chance of detecting the selected effect if that effect were the true one. The 90% power column uses a higher probability of detection under the same selected effect and assumptions, and therefore calls for more participants per arm. These are planning quantities, not a description of what happened in the trial.

The rows use 100%, 75%, and 50% of the observed effect as planning targets. A smaller target effect is harder to distinguish from random variation, so it requires more participants. The displayed sample sizes therefore grow as the assumed effect becomes smaller. Different comparisons have different observed effects and consequently different planning figures. The calculations are illustrative and depend on the stated normal-approximation and equal-allocation assumptions; they do not establish that the original study was designed using those assumptions.

This is not observed power. It is a forward-looking planning illustration based on the observed effect and specified assumptions, not a measure of the probability that the completed trial detected an effect.

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 = 24.900; pooled standard deviation σ = 61.384.
2. n = 2 × (1.96 + 0.842)² × 61.384² / 24.900² = 95.4, rounded up to 96 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 <- 24.9                      # observed difference in means
sigma <- 61.3840370128912                     # 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(354432)
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 = 96 (80% power), 128 (90% power)
 75% of observed effect: n per arm = 170 (80% power), 228 (90% power)
 50% of observed effect: n per arm = 382 (80% power), 511 (90% power)
power.t.test: 96.4 
Simulated power at n = 96 per arm: 0.796 (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 registry reports a mixed models analysis across all four primary-outcome groups, with p = .3455. It does not report an overall effect estimate or interval for that test. For quality of life, the sponsor's listed analysis is ANCOVA, with p = 0.208. Those are the registry's reported analyses.

The pairwise values are recalculated, unadjusted comparisons from the posted arm summaries. They compare each listed active-treatment group with placebo and do not reproduce the sponsor's four-arm mixed model. The pairwise estimates, intervals, and p-values answer narrower questions than an overall test across all groups. The primary recalculations are consistent with the broad pattern of the reported overall test in that they do not establish a clear difference, but the analyses are not interchangeable.

The registry does not report further model details needed to reproduce the sponsor's analyses. Since the recorded discrepancy flag is false, no disagreement between the posted summaries and recalculations is identified. The recalculated results should still be understood as simple supplementary comparisons rather than as sponsor-reported tests.

15. Multiplicity, Missing Data and Analysis Populations

The trial has four arms and the posted primary summaries include three comparisons with placebo, in addition to the sponsor's overall test. Considering several contrasts increases the opportunity for at least one apparently unusual result to arise by chance. The recalculated pairwise p-values are unadjusted; they do not account for the number of comparisons. No multiplicity adjustment for the primary comparisons is reported in the registry record.

The registry's overall mixed-model test addresses the four-group question, while the recalculated contrasts address individual active-versus-placebo questions. These are distinct statistical questions. A significant overall test, if present, would not by itself identify which arms differ; conversely, a collection of unadjusted pairwise tests should not replace the overall test. In this record, the overall test is not statistically compelling, and the recalculated intervals all include zero. No claim of equivalence follows from that pattern.

For quality of life, the sponsor reports an ANCOVA p-value and two recalculated pairwise comparisons. The registry does not report how the secondary outcome was prioritized, whether it was confirmatory, or whether adjustments were planned across outcomes and contrasts. The full multiplicity strategy is not reported in the registry record.

The primary analysis population is described as all randomized participants in a repeated measures mixed model, intended to allow inclusion of all study participants. The registry does not provide participant flow, the number with observed primary outcomes at each time, reasons for missing observations, or the model's assumptions about missing data. A mixed model can use incomplete repeated measurements under assumptions, but the record does not state which assumptions were used or assess their plausibility. The exact handling of missing quality-of-life data is also not reported.

Although the registry lists 30 participants in each primary-outcome arm, the quality-of-life summaries have smaller arm counts. The reason for the difference in posted counts is not reported in the registry record. The registry also does not report a detailed protocol, analysis plan, interim analyses, or the original sample-size justification. These omissions limit assessment of prespecification, analysis choices, and the potential effects of missing data and multiple testing.

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: erectile dysfunction (14.8%), pain - other (6.7%), insomnia (3.6%), pain - chest nos (3.6%), vascular - other (3.6%).

The registry lists serious adverse events among 30 participants at risk in the placebo arm, 27 in the soy arm, 29 in the venlafaxine arm, and 28 in the combination arm. The numbers affected by at least one serious adverse event are 4, 5, 5, and 5, respectively. These counts describe the posted safety groups and should not be read as estimates of treatment causation.

The registry also lists other adverse events affecting 28 of 30 participants in the placebo arm, 23 of 27 in the soy arm, 29 of 29 in the venlafaxine arm, and 27 of 28 in the combination arm. These figures are counts of affected participants, not a count of every event. The registry does not give enough detail to interpret event duration, severity gradations, or whether a participant experienced multiple events.

Among the listed serious-event terms, erectile dysfunction is reported as 14.8%, pain - other as 6.7%, insomnia as 3.6%, pain - chest NOS as 3.6%, vascular - other as 3.6%, and infection - rectum as 3.4%. The registry does not provide arm-specific counts for these terms in the posted summary. Death counts are not reported in the registry record. Safety summaries should be interpreted descriptively, especially in a trial of this size, and do not alone establish whether an event was caused by a treatment.

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

The posted primary outcome summaries show the 12-week hot flash score relative to baseline for four randomized groups. The registry's overall mixed-model test has p = .3455. The unadjusted recalculated comparisons against placebo have intervals that include zero, and their estimates are imprecise. Taken together, these results do not establish a clear difference between the groups on the primary endpoint.

This is not evidence that the treatments have identical effects or that meaningful differences are impossible. The intervals permit a range of effects, including differences in either direction. The available results also do not establish that any treatment reduces symptoms, nor do they support a medical recommendation. The endpoint is a symptom measure, and the posted summaries do not supply enough precision to settle the treatment contrasts.

The reported overall test concerns the four groups as a whole, while the recalculated comparisons concern selected pairwise contrasts with placebo. Neither should be treated as a test of factorial main effects or interaction, which are not reported. The primary result also does not answer every question about quality of life or safety. The secondary quality-of-life analysis is not a clear independent resolution of the primary uncertainty, and the safety counts are descriptive.

The trial is listed as terminated, but the reason for termination is not reported in the registry record. The record does not report the original design assumptions, interim analyses, detailed missing-data handling, or a complete multiplicity plan. Those omissions limit how fully the statistical conduct can be assessed. The findings are most accurately described as inconclusive for clear between-group differences in the posted primary summaries, rather than as proof of no effect.

18. Why This Trial Matters Statistically

This trial illustrates why a four-arm randomized design and its analysis need to be read at more than one level. The overall mixed-model test addresses whether the groups differ collectively, while the pairwise recalculations describe specific contrasts against placebo. Those questions are related but not interchangeable, and the distinction matters when interpreting a factorial trial.

It also shows why effect estimates and intervals are important alongside p-values. The overall p-value alone does not communicate the magnitude or direction of a group difference. The recalculated intervals make the uncertainty in each pairwise estimate visible: each includes zero and spans differences in either direction. This helps prevent a point estimate or arm ranking from being mistaken for a precise result.

The record also provides a practical example of the limits of summary-data recalculation. Simple comparisons can make posted results more interpretable, but they do not reconstruct a repeated measures mixed model or account for multiplicity. The sponsor's overall analysis remains a different statistical result, with limited model detail available in the registry.

Finally, the record separates statistical interpretation from claims about clinical importance. It defines the endpoint's direction but does not establish a threshold for a meaningful change. The quality-of-life outcome and adverse-event summaries add context, yet do not eliminate uncertainty in the primary outcome. Interpreting these elements together is more informative than relying on one p-value or one comparison in isolation.

19. Limitations

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

Several trial-specific limitations constrain interpretation. The trial is listed as terminated, but the reason is not reported in the registry record. The posted primary summaries list 30 participants per arm, while the quality-of-life summaries list smaller numbers, and the reasons for those differences are not reported. The registry describes the primary population as all randomized participants in a repeated measures mixed model, but does not provide participant flow, missingness by visit, or the assumptions used to handle missing data.

The primary comparison summaries do not include sponsor-reported pairwise effect estimates. The registry gives an overall mixed-model p-value without a corresponding overall estimate or interval. The recalculated comparisons are unadjusted and approximate, relying on posted arm summary statistics rather than the complete model output. Their intervals should not be mistaken for intervals from the sponsor's mixed model.

Although the design is factorial, the registry does not report main-effect or interaction estimates. The record also does not report a detailed multiplicity plan, prespecified hierarchy for secondary outcomes, model covariates, covariance structure, or original sample-size justification. These gaps limit assessment of how the analyses were planned and how much precision the study was intended to achieve.

There is no reported threshold for a clinically important change in the hot flash score, and the preferred direction for the quality-of-life scale is not reported. The safety summary does not provide detailed arm-specific breakdowns of the listed serious-event terms, and death counts are not reported in the registry record. These limitations make the results informative as posted summaries, but not a complete account of all design, analysis, and safety details.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: Soy Protein/Effexor Hormone Therapy for Prostate Cancer — NCT00354432.
• Registry record last updated 2021-09-28; status last verified 2021-09.
• Results first posted 2017-05-18.
• 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

NCT00354432 is a randomized phase III, quadruple-masked factorial trial listed as terminated, with enrollment of 120. It compared placebo, soy protein/isoflavones, venlafaxine, and their combination in men with prostate cancer receiving hormonal manipulation. The primary endpoint was the 12-week hot flash symptom severity score relative to baseline, with lower values representing a better outcome.

The registry reports an overall mixed models analysis with p = .3455. Recalculated pairwise comparisons against placebo are imprecise, and all three intervals include zero. Quality of life is listed as a secondary outcome, with a sponsor-reported ANCOVA p-value of 0.208. The registry does not report the reason for termination, a detailed multiplicity plan, the original sample-size justification, or comprehensive missing-data methods. The posted results do not establish a clear difference between groups on the primary endpoint.