A Study to Evaluate the Efficacy of SAGE-217 in the Treatment of Adult Participants With Major Depressive Disorder: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate the Efficacy of SAGE-217 in the Treatment of Adult Participants With Major Depressive Disorder (NCT03672175), a randomized phase 3 study. The primary outcome is Change From Baseline in the 17-item HAM-D Total Score at Day 15.

2. Clinical Question

3. Trial Design

This was a phase 3, multicenter, randomized, placebo-controlled study with a parallel-group design. The registry record lists quadruple masking. It does not specify which roles were masked. The study status is completed, and the recorded enrollment is 581.

Participants were assigned to matched placebo, SAGE-217 20 mg, or SAGE-217 30 mg. Each regimen was taken orally once daily in the evening with food for 14 days. The primary endpoint compared change in HAM-D total score from baseline to Day 15. Baseline and Day 15 were also the stated time points for the listed secondary continuous outcomes, although the registry gives different time-frame wording for some secondary measures.

The primary analysis used a modified full analysis set. It included randomized participants with a valid baseline HAM-D score, at least one post-baseline HAM-D evaluation, and a baseline HAM-D total score of at least 22. The record defines the number analyzed as the number of participants with data available for analyses. The analyzed counts for the primary endpoint were 141 for placebo, 152 for SAGE-217 20 mg, and 153 for SAGE-217 30 mg.

The study start date was 2018-11-19, and the primary completion date was 2019-09-24. The registry record does not report the original sample-size justification, the planned allocation ratio, detailed eligibility criteria, or the full protocol for handling missing observations. Those omissions limit how far the reported design can be reconstructed from the registry entry alone.

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: A total of 1327 participants were screened of which 581 were randomized and 570 were dosed. This study consisted of up to a 28-day screening period, a 14-day treatment period, a 28-day follow-up (FU) period, and an extended FU period through Day 182 (6 months following the last dose).

Treatment Period (Days 1 to 14)

Reasons for not completing this period:

FU Period (Days 15 to 42)

Reasons for not completing this period:

Extended FU Period (Days 43 to 182)

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: mFAS included all randomized participants with valid baseline HAM-D and at least 1 post-baseline HAM-D evaluation, with a total HAM-D score of at least 22 at baseline. Here, "Overall number of participants analyzed" is the number of participants with data available for analyses.

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.

Population: Safety Set included all participants who were administered IP.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline in the 17-item HAM-D Total Score at Day 15, measured over Baseline (BL), Day 15.

The primary endpoint was change from baseline in the 17-item Hamilton Depression Rating Scale total score at Day 15. The HAM-D is used to rate depression severity in participants already diagnosed as depressed. Its total is the sum of scores across 17 items: 8 items are scored from 0 to 2, and 9 items are scored from 0 to 4. The total score can range from 0 to 52, with higher scores indicating more severe depression. A negative change from baseline indicates less depression on the scale.

Change scores focus on how participants’ measured symptoms differ from their own baseline scores over the specified period. Comparing those changes between randomized groups can help distinguish change observed with an assigned regimen from change observed with placebo over the same period. The endpoint is a scale-based measure, not a direct count of participants who improved or a measure of long-term outcomes.

The primary results are reported as least-squares means with standard errors for each group, and as between-group differences from a repeated-measures analysis. The registry record defines the primary analysis population using baseline HAM-D severity and availability of follow-up data. The choice of Day 15 makes the primary question specific to the measured short-term change; it does not, by itself, answer questions about longer-term outcomes.

7. Results Posted in the Registry

• SAGE-217 20 mg: mean -11.50 versus -11.10 on control; difference -0.40 (95% CI -2.08 to 1.28), standardized effect d = -0.05; z-test p = 0.640.
• SAGE-217 30 mg: mean -12.50 versus -11.10 on control; difference -1.40 (95% CI -3.16 to 0.36), standardized effect d = -0.18; z-test p = 0.120.

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.

Change From Baseline in Clinical Global Impression - Severity (CGI-S) Score at Day 15

• 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:
  • SAGE-217 20 mg: mean -1.60 versus -1.50 on control; difference -0.10 (95% CI -0.39 to 0.19), standardized effect d = -0.08; z-test p = 0.501.
  • SAGE-217 30 mg: mean -1.70 versus -1.50 on control; difference -0.20 (95% CI -0.49 to 0.09), standardized effect d = -0.16; z-test p = 0.179.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between SAGE-217 20 mg 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 SAGE-217 30 mg and the control. The interval gives the range of effects that remain compatible with the data.

Number of Participants Achieving Clinical Global Impression - Improvement (CGI-I) Response at Day 15

• 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:
  • SAGE-217 20 mg: 47.0% (71/151) versus 46.1% (65/141) on control; risk difference 0.9 percentage points (95% CI -10.5 to 12.4), risk ratio 1.02 (0.80–1.30); unadjusted two-proportion z-test p = 0.875.
  • SAGE-217 30 mg: 55.3% (84/152) versus 46.1% (65/141) on control; risk difference 9.2 percentage points (95% CI -2.2 to 20.6), risk ratio 1.20 (0.95–1.51); unadjusted two-proportion z-test p = 0.117.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between SAGE-217 20 mg 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 SAGE-217 30 mg and the control. The interval gives the range of effects that remain compatible with the data.

Change From Baseline in Hamilton Anxiety Rating Scale (HAM-A) Total Score at Day 15

• 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:
  • SAGE-217 20 mg: mean -9.10 versus -8.70 on control; difference -0.40 (95% CI -1.83 to 1.03), standardized effect d = -0.06; z-test p = 0.583.
  • SAGE-217 30 mg: mean -9.40 versus -8.70 on control; difference -0.70 (95% CI -2.11 to 0.71), standardized effect d = -0.11; z-test p = 0.332.
• Interpretation:
  • The 95% confidence interval includes no difference, so these data do not show a statistically significant difference between SAGE-217 20 mg 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 SAGE-217 30 mg and the control. The interval gives the range of effects that remain compatible with the data.

The secondary outcomes cover several kinds of measurement: symptom-severity scales, anxiety, global clinical ratings, and participant response or remission counts. In the detailed results reported, continuous group estimates generally changed in the direction of lower scores, while the CGI-I response counts were numerically higher in the active groups than in placebo. The corresponding between-group estimates remain uncertain: the reported confidence intervals include the no-difference value for the detailed secondary comparisons.

This broad pattern is compatible with the primary endpoint’s direction, where all groups had negative HAM-D changes and active-group point estimates were more negative than the placebo estimate. It does not independently confirm a treatment effect. The secondary comparisons are less conclusive than a prespecified primary result would be, particularly because the registry record does not report whether multiplicity was controlled across the many outcomes and comparisons.

The registry lists 29 secondary outcomes as posted, but detailed numerical results for most are not included in the information summarized in the record. Findings for those outcomes are not reported in the registry record. The listed outcomes include HAM-D response and remission, MADRS change, and HAM-D Core Subscale change; their direction and precision should not be inferred from the other secondary measures.

General explanation: when many outcomes or comparisons are tested, the chance of at least one apparently notable result can increase, even when no true differences exist. Adjustments or a prespecified testing hierarchy can address that concern, but the registry record does not report whether such procedures were used. Accordingly, isolated secondary signals, especially unadjusted ones, carry less evidential weight than the trial’s primary analysis. The secondary results are best viewed as contextual and exploratory unless additional prespecified analysis information is available.

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

• Change From Baseline in Clinical Global Impression - Severity (CGI-S) Score at Day 15
• Change From Baseline in the 17-item HAM-D Total Score
• Number of Participants Achieving HAM-D Response
• Number of Participants Achieving HAM-D Remission
• Number of Participants Achieving Clinical Global Impression - Improvement (CGI-I) Response at Day 15
• Change From Baseline in Hamilton Anxiety Rating Scale (HAM-A) Total Score at Day 15
• Change From Baseline in the Montgomery-Åsberg Depression Rating Scale (MADRS) Total Score at Day 15
• Change From Baseline in HAM-D Core Subscale Score

9. Statistical Interpretation

10. Statistical Methods Explained

For the primary endpoint, the registry reports group-level least-squares means and standard errors, followed by between-group least-squares mean differences from a mixed model for repeated measures (MMRM). It identifies the MMRM as the sponsor’s method. The record does not report the full model specification, such as the included covariates, covariance structure, or precise assumptions used for missing observations; those details are not reported in the registry record.

What does a least-squares mean represent?

General explanation: a least-squares mean is a model-based group mean. It can be adjusted for terms included in the model, rather than being simply the arithmetic average of all observed values. The precise adjustment depends on the model specification. Because the registry record does not list the full MMRM specification, the particular adjustments behind these least-squares means are not reported in the registry record.

General explanation: a standard error describes the estimated uncertainty of a mean estimate. It is not the same as the spread of individual participants’ scores. The primary group standard errors were 0.59 for placebo, 0.62 for the 20 mg group, and 0.68 for the 30 mg group. These measures describe uncertainty around group estimates and should not be interpreted as the range of individual changes.

Why use repeated-measures analysis?

General explanation: repeated-measures models are designed for outcomes measured at multiple time points and can use the pattern of available measurements in a model-based comparison. They can account for the fact that measurements from the same participant are related. The registry identifies MMRM, but does not provide its full specification or state exactly how each missing observation was handled. It is therefore not possible to verify from the registry record whether the model assumptions are appropriate for the observed missingness pattern.

The primary endpoint is a continuous outcome, so the reported contrast is a difference in mean change between groups. The sponsor’s reported difference for the 20 mg comparison was -0.4, and for the 30 mg comparison it was -1.4. A negative contrast means the estimated change was more negative in the active group than in placebo, given the stated group order and change-score interpretation. The confidence intervals indicate that the estimates are uncertain.

What are the recalculated comparisons?

General explanation: the posted recalculations use unadjusted summary-statistic comparisons. They combine the reported group means and standard errors to estimate a difference and an approximate standard deviation. They do not reproduce the sponsor’s MMRM. The recalculated comparison for 20 mg has a mean difference of -0.4 and an unadjusted p-value of 0.64; for 30 mg, the mean difference is -1.4 and the unadjusted p-value is 0.12. These results are close to the sponsor’s point estimates, but the analysis methods differ. The sponsor’s own analysis takes precedence.

General explanation: an unadjusted p-value describes how incompatible the observed data are with a specified no-difference model, under the assumptions of the analysis. It does not give the probability that the null hypothesis is true, and it does not measure the size or clinical importance of an effect. The registry reports superiority analyses. The record does not report whether or how multiplicity across outcomes and comparisons was controlled, so the p-values should not be read as if a correction were documented.

How should the secondary analyses be understood?

The registry identifies several secondary outcomes, including changes in CGI-S, HAM-D, HAM-A, MADRS, and HAM-D Core Subscale scores, as well as response and remission outcomes. The detailed secondary results described in the record include continuous outcomes and a participant-count outcome. The continuous comparisons use MMRM, while the CGI-I response comparison uses generalized estimating equations (GEE) and reports odds ratios. General explanation: GEE is a regression approach often used for outcomes such as binary responses; an odds ratio compares the odds of an outcome between groups. Those odds ratios are not mean differences and should not be interpreted on the HAM-D score scale.

General explanation: for binary outcomes, a risk difference expresses a contrast between event proportions, while an odds ratio expresses a contrast in odds. The recalculated quantities and the sponsor’s GEE estimates are not interchangeable model outputs. For the CGI-I response endpoint, the sponsor reported odds ratios of 1.05 for 20 mg versus placebo and 1.43 for 30 mg versus placebo, with both 95% confidence intervals including 1. The results do not establish a clear difference on that endpoint either.

The recalculated planning illustrations are separate from the observed trial analyses. They translate observed effects into hypothetical participant counts under stated assumptions; they are not estimates of what the trial’s original design required. The trial’s own original design assumptions are not reported in the registry record.

11. Confidence Intervals and Precision

A confidence interval describes a range of values compatible with the data and the analysis assumptions at a stated confidence level. It helps show the precision of an estimate: a narrower interval indicates less statistical uncertainty than a wider interval, while neither width nor inclusion of a particular value alone establishes clinical importance.

For the primary comparison of SAGE-217 20 mg with placebo, the sponsor’s 95% confidence interval for the least-squares mean difference was -2.0 to 1.3. For 30 mg versus placebo, it was -3.1 to 0.3. Both intervals include 0, the value representing no between-group difference. The point estimates favor a more negative change in the active groups, but the intervals leave room for no difference and for effects in either direction.

The unadjusted recalculated intervals are -2.077 to 1.277 for 20 mg versus placebo and -3.165 to 0.364 for 30 mg versus placebo. They also include 0. The sponsor’s analysis is the primary reported analysis; the recalculations use a different, simplified approach and should not be treated as a replication of the MMRM.

General explanation: confidence intervals do not assign a probability to the true effect lying in a particular interval after the data have been observed. They indicate the performance of an interval-generating procedure over repeated studies under its assumptions. The intervals should be read alongside the effect estimates, endpoint definition, analysis population, and model uncertainty. The registry record does not report a threshold for a clinically important HAM-D difference, so statistical precision cannot be translated into a judgment about clinical importance from these results alone.

12. Planning a Similar Trial: Sample Size

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

The sample-size illustrations translate the observed effect into hypothetical planning targets. For these continuous and binary outcomes, the figures are participants per arm, not total events. The assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each illustration uses an effect equal to the 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 gives the hypothetical participants per arm for a design with an 80% chance of detecting the specified effect under the stated assumptions, if that effect and the model assumptions were correct. The 90% power column gives the corresponding target for a 90% chance. These are design-planning probabilities under assumptions, not guarantees about a future study’s result.

The rows represent planning for 100%, 75%, or 50% of the observed effect. Smaller effects require more participants to detect under the same assumptions, so targets increase as the assumed effect becomes a smaller fraction of the observed effect. The curve conveys the same general relationship across effect sizes: the smaller the effect a design aims to detect, the larger the required sample. The planning illustrations are not observed power and do not describe the probability that this completed trial detected its actual effect.

Because the observed effect is itself uncertain, especially when its confidence interval includes no difference, targets based on that estimate are unstable illustrations rather than reliable forecasts. They should not be read as evidence that the trial was or was not adequately powered under its original plan.

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 = -0.400; pooled standard deviation σ = 7.344.
2. n = 2 × (1.96 + 0.842)² × 7.344² / -0.400² = 5,291.3, rounded up to 5,292 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 <- -0.4                      # observed difference in means
sigma <- 7.34384816304942                     # 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(3672175)
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 = 5,292 (80% power), 7,084 (90% power)
 75% of observed effect: n per arm = 9,407 (80% power), 12,594 (90% power)
 50% of observed effect: n per arm = 21,166 (80% power), 28,335 (90% power)
power.t.test: 5292.3 
Simulated power at n = 5,292 per arm: 0.799 (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 the sponsor’s primary comparisons using MMRM. For SAGE-217 20 mg versus placebo, the least-squares mean difference was -0.4, the 95% confidence interval was -2.0 to 1.3, and the reported p-value was 0.664. For SAGE-217 30 mg versus placebo, the difference was -1.4, the 95% confidence interval was -3.1 to 0.3, and the reported p-value was 0.116.

The recalculations from the posted group summaries give the same rounded point differences, with confidence intervals of -2.077 to 1.277 and -3.165 to 0.364, respectively. Their unadjusted p-values are 0.64 and 0.12. The close point estimates do not make the analyses equivalent: the recalculations are simple, unadjusted comparisons, whereas the sponsor used a repeated-measures model.

The discrepancy flag is false, but that does not make the recalculations a substitute for the sponsor’s analysis. The sponsor’s own analysis takes precedence because it is the reported trial analysis. The registry record does not provide enough model detail to independently reproduce it. Both sets of intervals include no difference, and both indicate uncertainty around the estimates.

15. Multiplicity, Missing Data and Analysis Populations

The registry identifies a primary endpoint and numerous secondary endpoints, with 30 outcome measures and 150 analyses recorded. It reports superiority comparisons for the primary endpoint and some secondary outcomes. The record does not report a multiplicity adjustment, a testing hierarchy, or whether the secondary analyses were designated confirmatory or exploratory. Without that information, the reported p-values cannot be interpreted as evidence that the overall family of comparisons was controlled at a particular error rate.

General explanation: testing many outcomes or contrasts creates multiple opportunities to observe a small p-value by chance. Procedures such as family-wise adjustments, a hierarchical testing plan, or false-discovery control can address different multiplicity goals. Which approach is suitable depends on the study’s prespecified objectives. The registry record does not report which, if any, approach was planned or used.

The primary analysis population required valid baseline HAM-D data and at least one post-baseline evaluation, in addition to randomization and the baseline severity threshold. This means the primary endpoint analysis did not simply include every randomized participant regardless of data availability. The registry defines the analyzed number as participants with data available for analyses. It does not report the missing-data counts by reason, the missingness pattern, or sensitivity analyses addressing departures from the MMRM assumptions.

General explanation: missing outcome data can affect an estimate when the likelihood of missingness is related to participants’ outcomes or differs across groups. Model-based methods make assumptions about the missing data process; sensitivity analyses can examine whether conclusions depend on those assumptions. The registry record does not report enough detail to assess that sensitivity for this trial.

The record also does not report whether interim analyses were conducted, whether any stopping rules were used, or the original sample-size justification. These are not grounds to assume a particular design decision. They are simply details not reported in the registry record and should remain uncertainties when interpreting the posted analyses.

16. Safety Results

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

Deaths

Other (non-serious) adverse events

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

Most frequent serious adverse event terms

Most frequent non-serious adverse event terms

Serious events with the highest arm-level rate: depression (0.8%), death (0.8%), diabetic foot (0.8%), mania (0.8%), hypertension (0.8%).

The registry reports serious adverse event counts separately for the treatment period, follow-up period, and extended follow-up period. During treatment, serious adverse events affected 0 participants in the placebo group, 0 in the 20 mg group, and 2 in the 30 mg group. During follow-up, the reported count was 1 in each group. During extended follow-up, the counts were 1 for placebo, 3 for 20 mg, and 1 for 30 mg. These are reported counts, not estimates of comparative risk.

Deaths were reported separately: none during the treatment period or follow-up period, and 1 in the extended follow-up period for the 20 mg group. The registry also lists serious-event terms, including Depression, Death, Diabetic foot, Mania, Hypertension, and Agitation, with percentages attached to those terms. The record does not provide enough context in these figures to attribute an event to treatment or to draw a causal conclusion.

General explanation: serious adverse event counts are not a complete account of safety. They do not show all non-serious events, duration, clinical context, or whether an event was considered related to the study drug. The observed counts should be read as registry-reported events during stated periods, not as proof of comparative safety. Additional safety details are not reported in the registry record.

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

The primary endpoint compares short-term change in HAM-D total score from baseline to Day 15 among participants meeting the modified full analysis set definition. The active-group point estimates were more negative than the placebo estimate, but the sponsor’s confidence intervals for both comparisons include no difference. The posted analyses therefore leave substantial uncertainty about the size and direction of the between-group differences.

This result does not establish that the groups were identical. A confidence interval that includes no difference can also include differences favoring either group. Nor does a numerically larger change in one group alone establish a between-group effect; randomized comparisons, their uncertainty, and the analysis plan are needed for that interpretation.

The outcome is a rating-scale score, not a direct measure of every aspect of a participant’s experience. The results concern the specified scale and the Day 15 assessment, not longer-term effects or outcomes that were not measured. The registry record does not report a clinically important difference threshold for interpreting the HAM-D contrasts, so the statistical estimates cannot by themselves determine whether a difference would matter clinically.

The findings also do not establish conclusions for every adult with major depressive disorder. The primary analysis population had a baseline HAM-D total score of at least 22 and required at least one post-baseline evaluation. Results outside that definition are not reported in the registry record. The registry safety counts likewise do not determine whether an event was caused by treatment, and they do not capture all dimensions of safety.

In short, the record documents group changes and comparative estimates, with uncertainty that includes no difference. It does not support a definitive claim about benefit, lack of benefit, or long-term outcomes. Interpretation beyond the stated endpoint, period, population, and reported analyses would require information not reported in the registry record.

18. Why This Trial Matters Statistically

This trial is statistically informative because it illustrates the distinction between within-group change and a randomized between-group comparison. HAM-D scores fell on average in all three groups. The key question is not whether scores changed from baseline in any one group, but whether the estimated changes differed between active treatment and placebo. The reported contrasts are modest relative to their uncertainty and both sponsor confidence intervals include no difference.

It also shows why analysis method matters. The sponsor used MMRM and reported least-squares mean differences, whereas the simple recalculations use posted summary statistics and are unadjusted. Similar rounded point estimates do not mean the procedures answer the question in precisely the same way. Reproducibility requires model details that the registry record does not report.

The trial’s secondary outcomes and numerous recorded analyses raise a further statistical issue: multiplicity. Without a reported adjustment strategy or testing hierarchy, secondary p-values should not be treated as a collection of independent confirmatory tests. This is especially important when interpreting several measures that assess related aspects of symptoms.

Finally, the planning illustrations show how strongly required sample size depends on the effect a study aims to detect. Targets grow as the assumed effect becomes smaller. Since these illustrations use an observed effect and stated generic assumptions, they cannot reconstruct the trial’s original design or establish that its planned size was adequate. The trial therefore offers a practical example of reading estimates, intervals, model-based results, multiplicity, and planning assumptions together rather than relying on a single p-value.

19. Limitations

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

The registry record does not give the full MMRM specification, including covariates and covariance structure. It also does not report detailed missing-data patterns, reasons for missingness, or sensitivity analyses. These omissions make it difficult to assess the robustness of the primary model-based comparison independently.

The primary analysis population required valid baseline and post-baseline HAM-D data and a baseline HAM-D score of at least 22. The record does not provide results for a population that includes all randomized participants regardless of post-baseline data availability, nor does it describe the impact of participants excluded from the primary analysis.

The endpoint was assessed at Day 15 after a 14-day regimen. The posted primary result therefore addresses a short-term rating-scale change. Longer-term efficacy results are not reported in the registry record. The trial also reports several secondary outcomes and analyses, but multiplicity procedures and a testing hierarchy are not reported. This limits the confirmatory interpretation of secondary findings.

The unadjusted recalculations are approximations based on summary statistics and do not reproduce the sponsor’s MMRM. They rely on approximate standard deviations and should be interpreted as supplemental checks, not as independent versions of the trial’s primary analysis. Planning targets derived from observed effects are likewise illustrative and do not reveal the original sample-size justification.

For safety, the listed serious-event counts do not provide detailed clinical context, relatedness assessments, or a complete account of non-serious events. The registry record does not report enough information to make causal safety comparisons. The record also does not report the details of any interim analyses or stopping rules, so no conclusions about them are warranted.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Study to Evaluate the Efficacy of SAGE-217 in the Treatment of Adult Participants With Major Depressive Disorder — NCT03672175.
• Registry record last updated 2023-11-29; status last verified 2023-11.
• Results first posted 2022-11-10.
• 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

NCT03672175 is a completed phase 3, multicenter, randomized, quadruple-masked, parallel-group, placebo-controlled study of SAGE-217 in adults with major depressive disorder. Participants received matched placebo, SAGE-217 20 mg, or SAGE-217 30 mg once daily in the evening with food for 14 days.

The primary endpoint was change from baseline in 17-item HAM-D total score at Day 15. The sponsor reported MMRM comparisons of each active group with placebo. Both 95% confidence intervals include no difference. Unadjusted recalculations from the posted summary statistics are similar in point estimate, but the sponsor’s own analysis takes precedence.

Secondary outcomes included symptom scales and response or remission measures. The record does not report multiplicity procedures. Serious adverse event counts and deaths are listed by study period; the counts do not establish causation. The original sample-size justification, full model specification, and detailed missing-data analyses are not reported in the registry record.