A Study to Evaluate Efficacy and Safety of SAGE-547 in Participants With Severe Postpartum Depression (547-PPD-202B): Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Evaluate Efficacy and Safety of SAGE-547 in Participants With Severe Postpartum Depression (547-PPD-202B) (NCT02942004), a randomized phase 3 study. The primary outcome is Change From Baseline at 60 Hours in the 17-Item Hamilton Rating Scale for Depression (HAM-D) Total Score.

2. Clinical Question

The clinical question is about a measured symptom score over a defined period, not about every aspect of postpartum depression or longer-term outcomes. The registry describes the study purpose as determining whether SAGE-547 reduces depressive symptoms compared with placebo. The primary endpoint uses the 17-item HAM-D total score, assessed at baseline and Hour 60. It is a continuous outcome, so the main comparison concerns group mean changes rather than the proportion of participants meeting a response threshold.

The record also names secondary outcomes covering HAM-D scores at Day 30 and other time points, HAM-D response and remission, HAM-D subscale and individual-item scores, MADRS scores, and CGI-I response. Those measures broaden the set of questions, but the primary comparison remains the Hour 60 HAM-D change. The registry record does not report a clinical threshold for interpreting a difference in HAM-D points, so the estimates describe score differences rather than establishing how large a difference would be important to an individual participant.

3. Trial Design

The registry describes a phase 3, randomized, quadruple-masked, parallel-group, placebo-controlled study. It lists the study as completed, with 138 participants enrolled. The study is described as multicenter in its official title. The start date is 2016-08-01, and the primary completion date is 2017-09-24. The registry record does not report the trial's original sample-size justification or the detailed procedures used to generate the randomization sequence.

Three groups are listed: placebo, SAGE-547 60 μg/kg/h, and SAGE-547 90 μg/kg/h. Infusion lasted 60 hours. The 60 μg/kg/h regimen began with 30 μg/kg/h for 4 hours, continued at 60 μg/kg/h from 4 to 56 hours, and tapered to 30 μg/kg/h from 56 to 60 hours. The 90 μg/kg/h regimen began at 30 μg/kg/h from 0 to 4 hours, continued at 60 μg/kg/h from 4 to 24 hours, then reached 90 μg/kg/h from 24 to 52 hours. It tapered to 60 μg/kg/h from 52 to 56 hours and 30 μg/kg/h from 56 to 60 hours. Placebo participants received infusion rates equivalent to either active regimen.

The primary outcome analysis lists 43 participants for placebo, 37 for SAGE-547 60 μg/kg/h, and 39 for SAGE-547 90 μg/kg/h. These are the numbers with data available for analysis at the relevant time point, under the registry's description of the analysis population; they are not a statement that every randomized participant contributed to every analysis. The registry record does not report the exact allocation ratio or which specific roles were included in quadruple masking. The primary question is focused on severe postpartum depression, while the official title also names adult female subjects with moderate postpartum depression; the registry record does not provide further detail in the registry record describing that distinction.

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 138 participants were randomized into the study to receive either placebo or SAGE-547. Forty-three received placebo, and 79 received SAGE-547.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All randomized participants who started study drug infusion, had a valid baseline HAM-D assessment and at least one post-baseline HAM-D assessment. Overall number of participants analyzed is the number of participants with data available for analysis at the given time-point.

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: All randomized participants who started study drug infusion, had a valid baseline HAM-D assessment and at least one post-baseline HAM-D assessment.

6. Primary Endpoint

The registry lists the primary outcome as Change From Baseline at 60 Hours in the 17-Item Hamilton Rating Scale for Depression (HAM-D) Total Score, measured over Baseline, Hour 60.

The primary endpoint is change from baseline at Hour 60 in the 17-item HAM-D total score. The scale sums scores from 17 items covering symptoms such as mood, sleep, anxiety, and activity. The total can range from 0 to 52, with higher scores indicating a greater degree of depression. A negative change from baseline indicates less depression, and a positive change indicates more depression.

Change scores are used to describe how the measured symptom burden shifts during the study period, while comparing groups helps distinguish the observed change under each regimen from change under placebo. The endpoint is continuous: the registry reports least-squares mean changes and their standard errors, rather than only classifying participants as responders or nonresponders. The primary analysis population is described as randomized participants who started study drug infusion, had a valid baseline HAM-D assessment, and had at least one post-baseline HAM-D assessment. The number analyzed at a time point is the number with data available for that analysis.

General statistical context: a mean change gives a group-level summary and does not describe every participant's experience. It also does not by itself establish whether a difference is clinically important. The registry record does not report a threshold for a clinically meaningful between-group difference on this endpoint. Thus, the results quantify the contrast in HAM-D score changes; interpreting the importance of that contrast requires context beyond the point estimate alone.

7. Results Posted in the Registry

• SAGE-547 60 μg/kg/h: mean -19.54 versus -14.04 on control; difference -5.50 (95% CI -8.80 to -2.20), standardized effect d = -0.73; z-test p = 0.001.
• SAGE-547 90 μg/kg/h: mean -17.72 versus -14.04 on control; difference -3.68 (95% CI -6.93 to -0.43), standardized effect d = -0.49; z-test p = 0.026.

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 HAM-D Total Score at Day 30

• 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-547 60 μg/kg/h: mean -19.45 versus -13.82 on control; difference -5.63 (95% CI -9.46 to -1.80), standardized effect d = -0.66; z-test p = 0.004.
  • SAGE-547 90 μg/kg/h: mean -17.62 versus -13.82 on control; difference -3.80 (95% CI -7.58 to -0.02), standardized effect d = -0.45; z-test p = 0.049.
• Interpretation:
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for SAGE-547 60 μg/kg/h than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.
  • The 95% confidence interval excludes no difference, so the mean is statistically lower for SAGE-547 90 μg/kg/h than for the control. This is an unadjusted, secondary comparison, so it is suggestive rather than confirmatory.

The registry names secondary outcomes that cover later HAM-D assessment, response and remission classifications, HAM-D subscales and individual items, MADRS scores, and CGI-I response. The posted numerical secondary summaries include the Day 30 HAM-D total score comparison. Its group means show a direction similar to the primary endpoint: both active groups have more negative mean changes than placebo. The sponsor's Day 30 MMRM comparisons are also reported as superiority analyses.

This broad directional alignment between the posted Day 30 result and the Hour 60 primary result offers context about the measured HAM-D trajectory. It does not establish a consistent effect across every secondary outcome. Numerical results for the other named secondary measures are not reported in the registry record discussed in these results, so the overall pattern across all secondary endpoints cannot be determined from the available summaries.

Secondary results generally carry less weight than the prespecified primary question when many outcomes and time points are examined. General statistical context: testing multiple outcomes increases the chance that at least one result will appear notable by chance, unless the analysis plan accounts for multiplicity. The registry record does not report whether or how multiplicity was handled. The recalculated p-values are unadjusted; they do not account for the family of secondary outcomes. The sponsor's adjustment approach for secondary analyses is not reported in the registry record.

For that reason, the secondary results are best treated as supportive or exploratory context rather than as a series of independent confirmations. The Day 30 finding can be read alongside the primary result, but it should not be used to claim confirmation across the full set of outcomes. The registry record does not report a hierarchy of secondary hypotheses or identify which secondary results, if any, were intended to support formal confirmatory claims.

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

• Change From Baseline in HAM-D Total Score at Day 30
• Change From Baseline in HAM-D Total Score
• Percentage of Participants With HAM-D Response
• Percentage of Participants With HAM-D Remission
• Change From Baseline in HAM-D Bech 6 Subscale
• Change From Baseline in HAM-D Individual Item Scores
• Change From Baseline at Key Time Points in Montgomery-Åsberg Depression Rating Scale (MADRS) Total Score
• Percentage of Participants With Clinical Global Impression - Improvement (CGI-I) Response

9. Statistical Interpretation

10. Statistical Methods Explained

The primary outcome is continuous, and its posted group summaries are least-squares means with standard errors. The primary comparison is expressed as an LS mean difference from placebo. A negative difference means that the active group has a more negative average change in HAM-D score than placebo. Since a negative HAM-D change indicates less depression, the direction of the estimate favors a larger decrease in the measured score in the active group. This is a statement about the score contrast, not a clinical recommendation.

What does the mean difference summarize?

General statistical context: a mean difference subtracts one group's mean outcome from another group's mean outcome, with the sign depending on the comparison order. Here the posted comparisons are active treatment versus placebo. The mean difference is in HAM-D score units, not a percentage or a ratio. It summarizes the estimated difference between group averages and does not imply that every participant experienced a change equal to that estimate.

The registry reports least-squares means for the primary endpoint and identifies the sponsor's comparison method as MMRM, with the parameter described as an LS mean difference. General context: a mixed model for repeated measures can analyze outcome measurements taken at multiple times while accounting for their within-participant relationship and representing change over time. The exact covariance structure, fixed effects, covariates, missing-data assumptions, and model specification are not reported in the registry record. Those details should not be inferred from the method label alone.

How should the standard errors be read?

General statistical context: a standard error describes the estimated uncertainty in a group-level estimate, such as a mean. It is not the standard deviation of participant scores and does not show the full spread of individual outcomes. The standard errors listed for the primary least-squares means are 1.148 for placebo, 1.231 for SAGE-547 60 μg/kg/h, and 1.193 for SAGE-547 90 μg/kg/h. A confidence interval for a between-group contrast is more directly useful for seeing the uncertainty around that contrast than the separate group standard errors.

The simple recalculations use the posted means and standard errors to estimate an unadjusted mean difference and an approximate pooled standard deviation. Because those standard deviations are approximate and the calculations do not reproduce the sponsor's MMRM, the recalculated intervals and p-values may differ from the sponsor-reported results. The registry flags a discrepancy, so the sponsor's own analysis takes precedence. These recalculations are best read as a transparent, unadjusted check based on summary statistics, not as a replacement for the trial analysis.

What does a confidence interval convey?

General statistical context: a confidence interval represents the uncertainty associated with an estimate under the assumptions of the statistical method. A narrower interval indicates greater precision, while a wider interval leaves a broader range of values compatible with the data and method. The sponsor-reported primary intervals are 95% confidence intervals. They convey uncertainty about the mean difference, not a range expected to contain a specified proportion of individual participant scores. Their interpretation depends on the analysis model and its assumptions.

The sponsor's reported primary estimates are -5.50 for SAGE-547 60 μg/kg/h versus placebo and -3.68 for SAGE-547 90 μg/kg/h versus placebo. Their reported 95% confidence intervals are -8.80 to -2.20 and -6.90 to -0.47, respectively. The recalculated intervals are -8.799 to -2.201 and -6.925 to -0.435. The small differences between these reported and recalculated bounds reflect that these are not identical analysis procedures; the record does not explain the precise source of the discrepancy.

Why can over-dispersion matter?

General statistical context: over-dispersion occurs when variation in observed outcomes is greater than a model assumes. For continuous outcomes, a model that understates variability can produce standard errors that are too small and intervals that are too narrow. For this trial's continuous HAM-D outcome, the registry does not report an over-dispersion assessment or a correction for it. It would therefore be inappropriate to claim that over-dispersion was present or that any specific adjustment was used.

For sample-size planning, the listed illustration assumes a normal approximation and equal allocation between two arms. General context: such an approximation uses an assumed effect and variability to estimate the number of participants needed to attain a chosen power under a specified significance level. The planning illustrations are based on fractions of the observed effect; they are not the trial's original design justification. The original sample-size assumptions are not reported in the registry record.

The registry describes the primary analysis population and says that the number analyzed at a given time point is the number of participants with data available for analysis at that time point. That description does not reveal how missing assessments were handled within the MMRM or the recalculations. General statistical context: conclusions from a repeated-measures model can depend on assumptions about missing outcomes. The missing-data assumptions and sensitivity analyses are not reported in the registry record. The registry record also does not report an interim-analysis plan or the full statistical analysis plan.

11. Confidence Intervals and Precision

The sponsor-reported primary estimates are accompanied by 95% confidence intervals. For SAGE-547 60 μg/kg/h versus placebo, the LS mean difference is -5.50, and the interval extends from -8.80 to -2.20. For SAGE-547 90 μg/kg/h versus placebo, the LS mean difference is -3.68, and the interval extends from -6.90 to -0.47. Both intervals lie below zero, indicating that the sponsor's model-based estimates are compatible with a more negative mean change in the active group across the interval values.

General statistical context: a confidence interval is an estimate of precision under a specified analysis and its assumptions. It is not a guarantee that the true value is within the interval, nor does it describe the spread of individual participant responses. In this setting, the intervals concern the difference in group mean HAM-D changes. They do not indicate whether a difference is important to participants, because a clinically meaningful threshold is not reported in the registry record.

The simple unadjusted recalculations give intervals from -8.799 to -2.201 for the 60 μg/kg/h comparison and from -6.925 to -0.435 for the 90 μg/kg/h comparison. These are close to, but not exactly the same as, the sponsor's reported intervals. The record flags a discrepancy; therefore, the sponsor's own analysis takes precedence. The recalculated intervals are useful as a summary-statistic check, but they do not have the same model basis as the sponsor's MMRM results.

The interval widths also matter: each permits more than one plausible magnitude of mean difference under the analysis. Precision is not the same as certainty about clinical importance. Nor do group-level intervals capture all variation among participants. The registry record does not report whether the sponsor performed additional sensitivity analyses that might address uncertainty from model choice or missing outcome data.

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 figures are planning illustrations, not a report of the trial's original design calculation. For these continuous outcomes, the values are participants per arm, not total events. The planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. Each scenario uses the stated fraction of the observed effect as the effect to detect. The trial's original sample-size assumptions and justification are not reported in the registry record.

The 80% power column gives the estimated participants per arm under the stated assumptions for a design with an 80% chance of detecting the specified effect, if that effect and the other planning assumptions were correct. The 90% power column describes the corresponding planning target for a 90% chance. These are design probabilities under assumptions, not probabilities that the observed result is true. This is not observed power.

Rows that plan for 100% of the observed effect use the full effect estimated in this trial as the hypothetical target. Rows for 75% or 50% use progressively smaller target effects. Smaller effects are harder to distinguish from random variation, so they require more participants per arm under the same planning assumptions. For example, the primary comparison against SAGE-547 60 μg/kg/h lists 30 participants per arm for 80% power when planning for 100% of the observed effect, and 118 participants per arm when planning for 50% of that effect. For the 90 μg/kg/h comparison, the corresponding figures are 66 and 261 participants per arm. These illustrations are sensitive to the effect estimate used, which is itself uncertain, and should not be read as evidence that the original study was planned using these values.

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 = -5.500; pooled standard deviation σ = 7.509.
2. n = 2 × (1.96 + 0.842)² × 7.509² / -5.500² = 29.3, rounded up to 30 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 <- -5.5                      # observed difference in means
sigma <- 7.50947732484144                     # 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(2942004)
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 = 30 (80% power), 40 (90% power)
 75% of observed effect: n per arm = 53 (80% power), 70 (90% power)
 50% of observed effect: n per arm = 118 (80% power), 157 (90% power)
power.t.test: 30.3 
Simulated power at n = 30 per arm: 0.795 (20000 replicates)

The simulation draws many sets of group means and variances at the planned sample size and counts how often the two-sample t-test rejects the null. It should land close to the target power; at very small required sample sizes it can fall below it, because the normal approximation is rough there.

14. Analyses Reported by the Sponsor

The sponsor's own primary analyses use MMRM and report LS mean differences with 95% confidence intervals. For SAGE-547 60 μg/kg/h versus placebo, the sponsor reports -5.50, with an interval from -8.80 to -2.20. For SAGE-547 90 μg/kg/h versus placebo, the reported difference is -3.68, with an interval from -6.90 to -0.47. The sponsor labels both comparisons as superiority analyses.

The unadjusted recalculations use posted group means and standard errors rather than the sponsor's repeated-measures model. They produce differences of -5.5 and -3.68, with intervals from -8.799 to -2.201 and -6.925 to -0.435. The estimates point in the same direction and are similar in magnitude, but the confidence limits and p-values are not identical. For example, the sponsor's 90 μg/kg/h interval ends at -0.47, while the recalculated interval ends at -0.435. The registry record flags a discrepancy, so the sponsor's own analysis takes precedence.

The recalculations should be understood as simple, unadjusted checks on the posted summary statistics. They do not reproduce the sponsor's MMRM, and their p-values should not replace the sponsor's analysis. The registry record does not explain the source of the discrepancy in detail or report enough model specifications to reconstruct the sponsor's analysis independently. Accordingly, the sponsor-reported estimates and intervals are the principal inferential results; the recalculated results are supplementary context.

15. Multiplicity, Missing Data and Analysis Populations

The study includes a primary endpoint and a set of secondary outcomes covering multiple measures and time points. When many statistical comparisons are made, the chance of seeing at least one apparently unusual result can increase. General statistical context: multiplicity procedures can control that risk by adjusting significance thresholds or by specifying a testing sequence. Whether such an approach was used for the trial's analyses is not reported in the registry record. The record does not report an interim-analysis plan or the detailed statistical analysis plan.

The recalculated comparisons are explicitly unadjusted. Their p-values therefore describe each comparison in isolation and do not correct for the presence of other outcomes or comparisons. The sponsor's reported analyses identify MMRM and provide estimates, intervals, and p-values, but the multiplicity strategy is not reported in the registry record. The study's labels of superiority do not by themselves establish whether multiplicity was controlled.

The primary analysis population is described as randomized participants who started study drug infusion, had a valid baseline HAM-D assessment, and had at least one post-baseline HAM-D assessment. The number analyzed at a time point is the number with data available for analysis at that time point. This is useful context for understanding why analysis counts may differ across endpoints or time points. The record does not report the number of participants excluded from each analysis or the reasons for missing assessments.

How missing outcomes were handled in the sponsor's MMRM, whether additional imputation was used, and whether sensitivity analyses were conducted are not reported in the registry record. General statistical context: missing-data assumptions can influence estimates when participants with missing measurements differ from those with observed measurements. Without the analysis details, the direction or size of any such influence cannot be assessed. The record also does not report how the analysis population was defined for each secondary endpoint, beyond the primary endpoint description.

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: intentional overdose (2.6%), suicidal ideation (2.6%).

The registry reports serious adverse events for the groups, with 43 participants at risk in placebo, 38 in SAGE-547 60 μg/kg/h, and 41 in SAGE-547 90 μg/kg/h. The corresponding numbers affected by serious adverse events are 0, 1, and 0. Deaths are reported as 0 in each group. The listed top serious-event terms are intentional overdose and suicidal ideation, each reported as 2.6%. The record does not assign those percentages to a particular group in the registry record describing the terms.

For other adverse events, the numbers at risk are 43 for placebo, 38 for SAGE-547 60 μg/kg/h, and 41 for SAGE-547 90 μg/kg/h. The numbers affected are 9, 17, and 13, respectively. These are counts, not rates calculated for comparison, and no additional event definitions or exposure-adjusted summaries are reported in the registry record.

These figures describe events recorded in the study and do not establish that an event was caused by a treatment. They also do not by themselves provide a complete comparison of safety, because the registry record does not report detailed event-level timing, severity, relatedness assessments, or the full safety analysis methods in the registry record listed. The serious adverse event counts and the other adverse event counts should be read as distinct reported categories, without assuming that they measure the same type of event.

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

The posted primary results show that mean HAM-D scores decreased from baseline in all three groups by Hour 60, with more negative group mean changes in each SAGE-547 group than in placebo. The sponsor's model-based estimates indicate a more negative mean change for both active regimens relative to placebo, with 95% confidence intervals below zero. The registry also reports a Day 30 HAM-D result with the same broad direction. These findings describe measured group-level score changes over the reported time frames.

They do not establish that every participant improved, that the size of the score difference was important to participants, or that the result applies to people outside the study population. The registry record does not report a threshold for clinically meaningful change on the HAM-D. Nor do the posted primary results answer questions about outcomes beyond the reported time points, such as longer-term course or functioning. Those outcomes should not be inferred from a change in HAM-D total score alone.

The record's discrepancy flag is important when interpreting the numerical comparisons. The recalculated values are simple unadjusted estimates from posted summaries, while the sponsor reports MMRM results. The sponsor's own analysis takes precedence. The recalculations can offer context about the direction and approximate magnitude, but they are not the definitive model-based analysis.

Likewise, the safety counts are descriptive and do not establish causation or rule out risks. A zero count in a group is a report about observed serious adverse events in that group, not proof that an event cannot occur. The registry record does not provide enough detail to resolve all safety questions or to assess the significance of the serious-event terms by treatment group. Overall, the result is a comparison of measured symptom scores and reported events within this trial, not a complete account of benefit, risk, or individual response.

18. Why This Trial Matters Statistically

This trial illustrates why the statistical question should be tied to a clearly defined endpoint, time point, and comparison. The primary result concerns change in a continuous symptom score from baseline to Hour 60. The least-squares mean differences retain the score's direction and units, making it possible to interpret a negative contrast in relation to the scale's definition. The same numeric sign would not have the same meaning for every outcome, so scale direction matters.

It also illustrates the distinction between a model-based sponsor analysis and a simple recalculation from summary statistics. The sponsor used MMRM and reported adjusted least-squares mean contrasts, whereas the recalculations use posted means and standard errors and are unadjusted. They are similar in broad direction, but they are not interchangeable. A discrepancy flag should lead readers to privilege the sponsor's own analysis rather than choosing whichever result appears more favorable.

Precision is another essential part of the interpretation. The 95% intervals indicate a range of uncertainty for each estimated group difference, but they do not establish clinical importance or describe individual variation. The registry does not report a clinical-importance threshold. The sample-size illustrations further show, under their stated planning assumptions, that smaller target effects require more participants per arm. Those illustrations must not be confused with the original trial design or with evidence about the chance that the observed finding is correct.

Finally, this record shows the limits of interpreting a collection of outcomes without the multiplicity plan and missing-data details. A primary comparison, multiple secondary endpoints, and analysis populations require a statistical context that includes how hypotheses were prioritized and how incomplete measurements were handled. Where those details are not reported, interpretation should remain focused on what the posted estimates directly describe.

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 follow from what is and is not reported in the registry record. The analysis counts for the primary endpoint are 43 for placebo, 37 for SAGE-547 60 μg/kg/h, and 39 for SAGE-547 90 μg/kg/h, while the registry's total enrollment is 138. The primary analysis population requires infusion initiation, a valid baseline assessment, and at least one post-baseline HAM-D assessment. The record does not give the reasons participants may be absent from a particular time-point analysis or describe the missing-data assumptions used in the MMRM.

The reported primary sponsor analysis is model-based, but detailed specifications are not reported in the registry record. The discrepancy between the sponsor's analysis and simple recalculations is flagged without a detailed explanation of its source. This limits independent reconstruction from summary values. The sponsor's own analysis takes precedence, but readers cannot assess all modeling choices from the method label alone.

The multiplicity plan, interim-analysis procedures, and original sample-size justification are not reported in the registry record. This matters because the study names multiple secondary endpoints, and unadjusted comparisons across outcomes do not have the same interpretation as a prespecified testing strategy with a stated multiplicity approach. Secondary numerical details beyond the Day 30 HAM-D summaries are not reported in the registry record.

The results focus on HAM-D score changes at Hour 60 and Day 30. The registry record does not report a clinically meaningful threshold, so the clinical importance of the between-group score differences cannot be determined from these values alone. The safety information also consists of selected counts and terms; detailed event definitions, relatedness, and timing are not reported in the registry record. These limits do not invalidate the recorded comparisons, but they narrow the conclusions that can be drawn from them.

20. Related Tutorials

21. Related Statistical Calculators

22. Sources

• ClinicalTrials.gov: A Study to Evaluate Efficacy and Safety of SAGE-547 in Participants With Severe Postpartum Depression (547-PPD-202B) — NCT02942004.
• Registry record last updated 2025-10-14; status last verified 2022-01.
• Results first posted 2019-06-13.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

23. Record Summary

The registry record describes a completed phase 3 randomized, quadruple-masked, parallel-group study of intravenous SAGE-547 versus placebo in severe postpartum depression. The primary endpoint was change from baseline to Hour 60 in 17-item HAM-D total score, where a negative change indicates less depression. The listed analyzed group means were -14.04 for placebo, -19.54 for SAGE-547 60 μg/kg/h, and -17.72 for SAGE-547 90 μg/kg/h.

The sponsor's MMRM analyses report negative LS mean differences for both active regimens versus placebo, with 95% confidence intervals below zero. The record flags a discrepancy with simple unadjusted recalculations, so the sponsor's own analysis takes precedence. A Day 30 HAM-D result shows a similar broad direction. Serious adverse event counts were 0, 1, and 0 across placebo, 60 μg/kg/h, and 90 μg/kg/h, and deaths were 0 in each group.

The registry record does not report the original sample-size justification, multiplicity strategy, interim-analysis plan, detailed missing-data assumptions, or a clinically meaningful HAM-D threshold. These omissions limit interpretation beyond the reported group score comparisons and safety counts.