A Study to Compare SB16 (Proposed Denosumab Biosimilar) to Prolia® in Postmenopausal Women With Osteoporosis: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Compare SB16 (Proposed Denosumab Biosimilar) to Prolia® in Postmenopausal Women With Osteoporosis (NCT04664959), a randomized phase 3 study. The primary outcome is Percent Change From Baseline in Lumbar Spine BMD at Month 12.

2. Clinical Question

The primary endpoint is a continuous measure: percentage change in lumbar spine bone mineral density (BMD) from baseline to Month 12. The posted results report least-squares mean changes of 5.63 for SB16 and 5.3 for Prolia, with a reported mean difference of -0.33 for Prolia compared with SB16. The reported confidence interval spans values on both sides of zero, and the unadjusted p-value is 0.354. These statistics describe the observed comparison and its uncertainty; they do not, by themselves, establish equivalence or a clinically acceptable degree of similarity. The registry record does not report an equivalence or non-inferiority margin.

3. Trial Design

This was a Phase III, randomized, double-blind, multicentre, parallel-group study. The registry describes masking as quadruple. The study compared SB16, a proposed denosumab biosimilar, with Prolia, denosumab, in postmenopausal women with osteoporosis. The registry lists Samsung Bioepis Co., Ltd. as the sponsor. The study is marked completed, with enrollment of 457.

Participants assigned to SB16 were scheduled to receive SB16 at a dose of 60 mg in 1 mL by subcutaneous injection every 6 months. Participants assigned to Prolia were scheduled to receive Prolia at the same stated dose and route on the same interval. The registry also describes a change in treatment assignment at Month 12 for the Prolia group: participants were to be re-randomized in a 1:1 ratio either to continue Prolia or to transition to SB16. Details of subsequent treatment and follow-up are not reported in the registry record.

The primary endpoint was assessed at Month 12, so it concerns the comparison of the originally assigned groups at that stated time point. The post-Month-12 re-randomization is relevant when interpreting later follow-up, but details of later outcomes are not reported in the registry record. The registry lists 225.0 participants in the SB16 group and 231.0 in the Prolia group for the primary outcome. These outcome group counts are not the same description as total enrollment; the record does not explain the difference between them.

The study start date is 2020-11-26, and the primary completion date is 2022-06-20. The randomized and masked design is intended to support a structured comparison between assigned groups, but the registry record does not provide enough information to assess implementation details such as the randomization procedure, concealment, or the people included in each masking category. It also does not report the protocol’s original sample-size justification or the full statistical analysis plan.

4. Participant Flow and Accounting

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

Main Period

Transition Period

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 Percent Change From Baseline in Lumbar Spine BMD at Month 12, measured over Baseline and Month 12.

The primary endpoint is percent change from baseline in lumbar spine BMD at Month 12. BMD is a continuous measurement, and expressing its change as a percentage relates the follow-up measurement to the participant’s baseline value. The registry identifies the outcome as a least-squares mean, a model-based group average. The exact model, its covariates, and the handling of repeated measurements are not reported in the registry record.

Lumbar spine BMD provides a defined measurement at a specific anatomical site and time point. A continuous endpoint can retain information about the size and direction of measured change, rather than reducing each participant to a binary category. In general, a group mean describes the average outcome, while the spread around that mean and the uncertainty in the group comparison help show how precisely the data estimate the underlying difference. Those summaries do not describe every participant’s experience.

The reported least-squares mean changes are 5.63 for SB16 and 5.3 for Prolia. The reported mean difference for Prolia compared with SB16 is -0.33. The primary endpoint is a measurement outcome; it is not itself a direct measure of fractures, symptoms, function, or quality of life. Whether the observed difference is important for patients cannot be determined from the BMD estimates alone. A threshold for clinical importance or an equivalence margin is not reported in the registry record.

7. Results Posted in the Registry

• Prolia: mean 5.30 versus 5.63 on control; difference -0.33 (95% CI -1.03 to 0.37), standardized effect d = -0.09; z-test p = 0.354.

8. Statistical Interpretation

9. Statistical Methods Explained

The main outcome is continuous: percent change in lumbar spine BMD from baseline to Month 12. The registry reports group least-squares means, standard errors, and a between-group mean difference with a confidence interval. These statistics answer different questions. The group means summarize estimated average change, the standard errors describe uncertainty in those estimates, and the confidence interval describes a range of differences compatible with the data and the analysis assumptions.

What is a least-squares mean?

In general, a least-squares mean is a model-based estimate of a group average. It can account for variables included in the fitted model, rather than being only the arithmetic average of the observed measurements. The model specification matters: included covariates, assumptions, and analysis population can affect the estimate. The registry record names the reported parameter as a least-squares mean but does not report the model or its covariates. It is therefore not possible to say which adjustments, if any, distinguish these estimates from unadjusted group averages.

How should the mean difference be read?

The registry reports a mean difference of -0.33 for Prolia compared with SB16. Read in that stated direction, the negative sign indicates that the posted estimate for Prolia is lower than the SB16 estimate. The reported group means are 5.3 for Prolia and 5.63 for SB16. This is a contrast of averages, not a description of the difference for every participant. Individual outcomes can vary around group estimates, and group averages do not show the full distribution of responses.

For continuous outcomes, a mean difference keeps the result in the outcome’s original scale. Here, the endpoint is expressed as percent change in BMD. A standardized effect size can express a difference relative to a measure of variability, but interpretation then depends on the variability estimate and the assumptions used to obtain it. The posted standardized estimate is -0.087; it is a supporting summary, not a replacement for the outcome-scale difference or its confidence interval.

What does the standard error describe?

In general, a standard error describes the estimated sampling variability of a statistic, such as a group mean. It is not the standard deviation of individual participants’ outcomes. The posted standard errors are 0.25 for SB16 and 0.254 for Prolia. A standard error is useful for understanding the precision of an estimated average, but it does not reveal how widely individual measurements vary. The registry reports these dispersion values as standard errors, not standard deviations.

The posted comparison also includes a pooled standard deviation of 3.806, with the record marking the approximation used for this value. A pooled standard deviation combines information about variability across groups under assumptions. Since the record indicates an approximation, this value should not be treated as a directly reported participant-level standard deviation. Details of how the approximation was made are not reported in the registry record.

How should the p-value be interpreted?

The unadjusted p-value is 0.354. In general, a p-value describes how unusual data at least as incompatible with a specified null hypothesis would be under that hypothesis and the test assumptions. It is not the probability that the null hypothesis is true, nor does it measure the size or clinical importance of a difference. The word “unadjusted” means the posted value is not described as corrected for multiplicity. The registry record does not report a full testing strategy or enough test details to reconstruct the analysis.

Interpreting a p-value alongside an effect estimate and confidence interval is more informative than interpreting it alone. The estimate gives direction and scale; the interval shows uncertainty; the p-value reports a test result under specified assumptions. In this study, the interval crosses zero, and the unadjusted p-value is 0.354. These registry entries do not prove that the groups are identical. They indicate that the data, under the reported comparison, do not provide strong evidence against a zero-difference null at conventional thresholds. A clinical equivalence conclusion would require an explicitly defined acceptable range and a design and analysis suited to that question.

Why can over-dispersion matter?

In general, over-dispersion occurs when observed variability is greater than a statistical model expects. If a model understates variability, standard errors can be too small and confidence intervals too narrow, which can make evidence appear more precise than it is. Appropriate modeling and diagnostics depend on the outcome type and analysis method. The primary outcome is continuous, and the registry does not report diagnostic checks or a method addressing over-dispersion. This general issue should not be taken to mean over-dispersion was present in this trial; its presence is not reported in the registry record.

The statistical details available in the registry record support a limited reading of the result: the estimated average changes, the reported comparison, its interval, and its unadjusted test value. The original analysis plan, model specification, handling of missing measurements, analysis populations, and multiplicity procedures are not reported in the registry record. Those omissions limit independent assessment of how the estimates were produced and how robust they are to analytic choices.

10. Confidence Intervals and Precision

The reported mean difference for Prolia compared with SB16 is -0.33. Its confidence interval crosses zero, so it is compatible with a negative difference and with a positive difference under the reported analysis. The lower confidence limit is -1.029. The interval’s upper limit is positive, but the exact value is not restated because the registry lists it with more than three decimal places.

A confidence interval expresses uncertainty around an estimate. In general, it is a range generated by a statistical procedure that, under its assumptions and repeated use, has a specified coverage property. The confidence level is not stated in the registry record. The interval should not be read as containing a fixed probability that the true value lies inside it after the data have been observed. It also does not show the range of individual participants’ outcomes.

The interval’s inclusion of zero means the posted comparison does not distinguish clearly between a difference favoring either direction and no difference. At the same time, the interval’s width means the estimate is not exact. For decisions about similarity, the relevant question would be whether the plausible differences fall within a prespecified acceptable margin. The registry record does not report such a margin, so the interval cannot establish equivalence or non-inferiority.

Precision depends on the amount of information in the data, variability, and the statistical model. The outcome group counts are 225.0 for SB16 and 231.0 for Prolia, and the posted standard errors are 0.25 and 0.254, respectively. These statistics help describe estimation uncertainty, but the record does not report the model details, analysis population, or missing-data handling needed to evaluate all sources of uncertainty. The precision should therefore be interpreted as reported rather than as a complete assessment of every design or analysis limitation.

11. Planning a Similar Trial: Sample Size

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

The planning figures are participants per arm, not total participants and not total events. They are illustrations based on the observed effect, rather than a record of the trial’s original design calculation. The registry record does not report the original sample-size justification or the assumptions used when the study was designed.

The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between two arms, and a normal approximation. For each scenario, the effect to detect is a stated fraction of the observed effect. Under these assumptions, planning for 100% of the observed effect gives 2,089 participants per arm for 80% power and 2,796 participants per arm for 90% power. Planning for 75% of the observed effect gives 3,713 participants per arm for 80% power and 4,971 participants per arm for 90% power. Planning for 50% of the observed effect gives 8,354 participants per arm for 80% power and 11,184 participants per arm for 90% power.

The 80% power column describes the number of participants per arm that the stated calculation estimates would be needed for an 80% chance of detecting the specified effect under its assumptions. The 90% power column gives the corresponding planning number for a 90% chance. These are design calculations, not guarantees about a future study’s result. The displayed figures are not observed power.

Smaller effects require more participants to detect with the same stated power and assumptions, because a smaller difference is harder to distinguish from random variation. The planning scenarios therefore illustrate how the required sample size changes when the target effect is a smaller fraction of the observed estimate. They should not be treated as a recommendation or as evidence that the completed study was designed to meet any particular power target.

12. How the Sample Size Is Derived

Power is the probability that a study of a given size rejects the null hypothesis when a specific true effect exists. It depends on four things: the effect to detect, the variability of the outcome, the significance level (here a two-sided α of 0.05, so z_1−α/2 = 1.960) and the desired power (z_1−β = 0.842 for 80% and 1.282 for 90%).

Formula

n is the number of participants in each arm, Δ is the difference in means to detect and σ is the common standard deviation.

Worked calculation for this trial

1. Observed difference Δ = -0.330; pooled standard deviation σ = 3.806.
2. n = 2 × (1.96 + 0.842)² × 3.806² / -0.330² = 2,088.5, rounded up to 2,089 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 <- -0.33                      # observed difference in means
sigma <- 3.80636352177664                     # 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(4664959)
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 = 2,089 (80% power), 2,796 (90% power)
 75% of observed effect: n per arm = 3,713 (80% power), 4,971 (90% power)
 50% of observed effect: n per arm = 8,354 (80% power), 11,184 (90% power)
power.t.test: 2089.4 
Simulated power at n = 2,089 per arm: 0.801 (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.

13. Multiplicity, Missing Data and Analysis Populations

The registry record lists one primary outcome measure and no posted secondary outcome measures. It does not report a multiplicity strategy, such as an ordering of tests or an adjustment to account for testing more than one outcome or contrast. The reported p-value is described as unadjusted. Without the prespecified testing plan, it is not possible to determine what inferential role that value was intended to play or whether other analyses were planned.

Multiplicity matters when several outcomes, time points, subgroups, or comparisons are tested. In general, conducting multiple tests can increase the chance of at least one apparently notable result even when no corresponding effect exists. Statistical plans may address that issue through prespecified testing sequences or adjusted thresholds. This is a general explanation; the registry record does not say that multiple primary tests were conducted or specify how multiplicity was handled.

Details about missing data are not reported in the registry record. The record does not state how many Month 12 measurements were missing, why any measurements were unavailable, whether imputation was used, or how missingness was considered in the analysis. It also does not describe sensitivity analyses. Missing-data methods can affect a continuous-outcome estimate, especially if the availability of measurements is related to outcomes or treatment assignment. Without these details, the extent and direction of any effect on the reported result cannot be assessed.

The analysis populations are not reported in the registry record. It is not stated whether the primary analysis followed all randomized participants, a treatment-received population, or another defined group. Nor are exclusions or protocol deviations described. These definitions matter because analyses of different populations can estimate different quantities and may have different susceptibility to bias.

The least-squares mean label suggests a model-based estimate, but the model formula, covariates, assumptions, and approach to repeated measurements are not reported. The registry record also does not provide a complete statistical analysis plan. Consequently, the posted estimate and unadjusted p-value can be described, but their reproducibility and sensitivity to analytic choices cannot be fully assessed from the registry record.

14. 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: atrial fibrillation (1.0%), transient ischaemic attack (1.0%), nephrolithiasis (1.0%), urethral caruncle (1.0%), osteoarthritis (1.0%).

For serious adverse events, the posted counts are 12 of 225 participants in the SB16 group and 11 of 231 in the Prolia Overall group. The record also gives post-Month-12 Prolia subgroup counts: 5 of 100 in the Prolia+SB16 group and 3 of 101 in the Prolia+Prolia group. These subgroup summaries concern groups formed after re-randomization and should not be read as additional independent groups to add to the original Prolia Overall count.

The posted counts for other adverse events are 116 of 225 for SB16, 107 of 231 for Prolia Overall, 46 of 100 for Prolia+SB16, and 49 of 101 for Prolia+Prolia. Deaths are listed as 0 for each of the four groups. The registry lists atrial fibrillation, transient ischaemic attack, nephrolithiasis, urethral caruncle, osteoarthritis, and general physical health deterioration among the top serious-event terms, each at 1.0%.

These are descriptive counts, not evidence that one treatment caused or prevented an event. The registry record does not report event-level timing, exposure-adjusted rates, follow-up duration, or a statistical comparison of safety outcomes. The subgroup results also follow a change in treatment assignment at Month 12, so they describe a different treatment period from the original randomized comparison.

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

The primary result describes average percent change in lumbar spine BMD at Month 12 among the posted groups. The reported least-squares mean is 5.63 for SB16 and 5.3 for Prolia. The reported mean difference for Prolia compared with SB16 is -0.33, and the confidence interval crosses zero. Taken together, these figures show that the posted average estimates are close in numerical size, while the comparison remains uncertain in direction and magnitude.

The result does not establish that SB16 and Prolia are equivalent, interchangeable, or clinically identical. Those conclusions require a defined equivalence or non-inferiority margin and an analysis that evaluates the estimate against that margin. No such margin is reported in the registry record. A non-significant or unadjusted test result is not proof of no difference.

The outcome also does not directly report fracture outcomes, symptoms, physical function, or quality of life. A BMD change is a measurement outcome at a specified site and time point. The clinical importance of a difference in that measurement cannot be determined from the reported values alone, and a clinical-importance threshold is not reported in the registry record.

The primary endpoint is at Month 12, after which the Prolia group was to be re-randomized either to continue Prolia or transition to SB16. Findings after that point would need to be interpreted in light of the changed treatment assignments. Later outcome details are not reported in the registry record.

The adverse-event counts provide descriptive summaries for the stated groups. They do not establish that an event was caused by a treatment, and the record does not provide a formal comparative safety analysis. Overall, the posted results support a careful description of the measured Month 12 outcome and its uncertainty, not a broader claim about all benefits, harms, or later outcomes.

16. Why This Trial Matters Statistically

This trial raises a central statistical question in biosimilar comparisons: how should evidence of similarity be evaluated? A small estimated difference and a p-value that does not provide strong evidence against a zero-difference null are not, by themselves, evidence that two products meet a defined standard of equivalence. The confidence interval and the prespecified margin are essential to that assessment. The registry record reports an interval that crosses zero but does not report an equivalence or non-inferiority margin.

The study also illustrates the importance of distinguishing a point estimate from its precision. The posted mean difference is -0.33, while the confidence interval allows values in both directions. A reader should consider both the estimate’s scale and the range of uncertainty, rather than relying on a single p-value. Since the endpoint is a least-squares mean, access to the underlying model and analysis population would also help readers understand exactly what the estimate represents.

Trial structure matters for interpretation. The Month 12 primary outcome is linked to the original randomized groups, whereas the Prolia group was to be re-randomized at Month 12 for subsequent treatment. Clear separation of the initial randomized comparison from later treatment periods helps prevent unlike follow-up comparisons from being treated as if they arose from the same assignment.

Finally, registry transparency affects how independently a result can be evaluated. The record contains the primary outcome estimates and adverse-event counts, but does not report the original sample-size justification, full statistical plan, analysis populations, or missing-data methods. Those details can clarify whether the analysis matched the intended question and whether conclusions depend on modeling choices. Their absence limits the statistical claims that can be made from the posted record alone.

17. 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 details needed for a fuller interpretation are not reported in the registry record. These include the full eligibility criteria, baseline characteristics, randomization method, allocation concealment, and the operational details of quadruple masking. Without those descriptions, it is not possible to assess how closely the enrolled participants and trial conduct align with the intended study population and procedures.

The outcome record gives least-squares means and standard errors, but does not describe the model, covariates, analysis population, or handling of missing Month 12 measurements. It also does not report a confidence level, the original statistical analysis plan, or the original sample-size justification. The reported comparison can be described, but its derivation cannot be independently reconstructed from the registry record alone.

No equivalence or non-inferiority margin is reported. This is an important limitation when interpreting a study comparing a proposed biosimilar with a reference product: an estimate close to zero, or a p-value that does not reject a zero difference, is not sufficient to establish equivalence. The registry record also does not report a threshold for a clinically important BMD difference.

Only one outcome measure is listed, with no secondary outcomes posted. The adverse-event summaries do not include reported follow-up duration, event timing, exposure-adjusted rates, or a formal comparison between groups. The Prolia re-randomization at Month 12 also changes the treatment context for later follow-up; details needed to interpret subsequent periods are not reported in the registry record.

Finally, the posted results concern lumbar spine BMD change at Month 12. They do not by themselves establish effects on other clinical outcomes or outcomes at other time points. These limitations do not invalidate the reported estimates, but they constrain how far the registry results can be interpreted.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: A Study to Compare SB16 (Proposed Denosumab Biosimilar) to Prolia® in Postmenopausal Women With Osteoporosis — NCT04664959.
• Registry record last updated 2025-02-04; status last verified 2025-01.
• Results first posted 2025-01-09.
• 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.

21. Record Summary

The registry record describes a completed, Phase III, randomized, double-blind, multicentre, parallel-group study in postmenopausal women with osteoporosis. Enrollment was 457, and Samsung Bioepis Co., Ltd. is listed as sponsor. Participants were assigned to SB16 or Prolia; the Prolia group was to be re-randomized at Month 12 to continue Prolia or transition to SB16.

The primary outcome was percent change from baseline in lumbar spine BMD at Month 12. The posted least-squares means are 5.63 for SB16 and 5.3 for Prolia. The reported mean difference for Prolia compared with SB16 is -0.33, with a confidence interval that crosses zero and an unadjusted p-value of 0.354. Serious adverse-event counts are 12 of 225 for SB16 and 11 of 231 for Prolia Overall; deaths are listed as 0 in each reported group.

The record does not report the equivalence margin, original sample-size justification, full statistical analysis plan, analysis populations, or missing-data methods. The results therefore describe the posted Month 12 comparison but do not establish equivalence or a broader clinical conclusion.