A Study to Investigate the Efficacy and Safety of GSK3196165 in Inflammatory Hand Osteoarthritis: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for A Study to Investigate the Efficacy and Safety of GSK3196165 in Inflammatory Hand Osteoarthritis (NCT02683785), a randomized phase 2 study. The primary outcome is Change From Baseline in 24-hour Average Hand Pain Intensity, Averaged Over the 7 Days Prior to Week 6.

2. Clinical Question

3. Trial Design

This was a randomized, Phase IIa, multicentre, double-blind, placebo-controlled, parallel-group study. The masking designation in the registry record is triple. The record does not state which specific groups or roles were masked. GlaxoSmithKline is listed as the sponsor, and the study status is completed.

The study compared GSK3196165 with placebo in subjects with active inflammatory hand osteoarthritis. The registry summary describes allocation across two treatment arms in a 1:1 ratio. It states that approximately 40 subjects were expected to be enrolled and that at least 40 would be randomized. The recorded enrollment is 44. The reported primary outcome groups contain 21 participants for placebo and 20 for GSK3196165 180mg; the registry record does not explain the difference between those group counts and enrollment.

The planned treatment period was 12 weeks, followed by a follow-up period completing at Week 22. Screening could last up to 4 weeks. The arm descriptions specify a total of 8 doses for each treatment. The study started on 2016-03-17, and primary completion was recorded as 2017-11-29.

The primary objective was to assess the efficacy potential of GSK3196165 on pain. The primary outcome focused on change from baseline in average hand-pain intensity at Week 6. The record lists 20 outcome measures and 67 analyses, while 19 secondary outcomes are posted. The original protocol’s detailed sample-size justification, interim-analysis plan, and full statistical analysis plan are not reported in the registry record.

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 121 participants were screened of which 77 were screen failures and 44 participants were enrolled in the study.

Overall Study

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: Intent To Treat Population. Only non-missing data is included MMRM model.

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 Change From Baseline in 24-hour Average Hand Pain Intensity, Averaged Over the 7 Days Prior to Week 6, measured over Baseline (Day 1 Pre-dose) and Week 6.

The primary endpoint measures change from baseline in 24-hour average hand-pain intensity at Week 6. Participants recorded a daily pain score using a numeric rating scale with anchors of 0 for no pain and 10 for worst imaginable pain. For each assessment, daily scores from the 7 days before the visit were averaged. Baseline was defined from the 7 days before the Day 1 pre-dose visit, and change was calculated as the post-dose visit value minus the baseline value.

A change below zero therefore represents a lower average pain score relative to baseline, while a change above zero represents a higher score. The endpoint is a continuous measure, so the reported arm estimates are least-squares means and their standard errors. Comparing the two arms estimates the difference in average change between treatment and placebo, rather than describing the change in either arm alone.

Week 6 was the primary assessment time in this outcome. The posted outcome population is described as randomized participants who received at least one dose. Only non-missing data were included in the MMRM model. Pain scores are directly interpretable on the stated scale, but the registry record does not give a threshold for judging what size of difference would be important to participants. The endpoint therefore gives an estimate of the measured pain-score contrast, not by itself a complete account of clinical importance.

7. Results Posted in the Registry

• GSK3196165 180mg: mean -1.70 versus -1.34 on control; difference -0.36 (95% CI -1.27 to 0.55), standardized effect d = -0.24; z-test p = 0.440.

8. Statistical Interpretation

9. Statistical Methods Explained

The primary outcome is continuous: it records pain intensity on a numeric rating scale, and the analysis compares average changes in scores. A least-squares mean is a model-based estimate of an arm’s mean outcome. Unlike a simple average calculated directly from all observed scores, it is produced within the specified model. The registry record identifies the primary model as a mixed model repeated measures analysis, or MMRM. Details of all model terms and assumptions are not reported in the registry record.

What does the between-group estimate represent?

The mean difference (net) compares the modeled change in the GSK3196165 arm with the modeled change in the placebo arm. The reported estimate is -0.36 scores. Since the endpoint is change from baseline, negative values indicate lower pain scores at Week 6 relative to baseline; the negative between-group estimate points toward a larger reduction in the GSK3196165 group. This direction describes the estimate, not a conclusion that the difference is certain or clinically important.

The arm-specific estimates are -1.34 for placebo and -1.7 for GSK3196165 180mg, with standard errors of 0.325 and 0.334. A standard error describes the estimated sampling uncertainty of an estimate: if comparable studies were repeated under the model and its assumptions, estimates would vary. It is not the spread of individual participants’ pain scores. The registry record reports standard errors for the arm means but does not post standard deviations for those means in the primary outcome.

How should the confidence interval be understood?

The 95% confidence interval reported by the sponsor runs from -1.31 to 0.58 scores. In general, a confidence interval gives a range of values that are reasonably compatible with the data and the analysis procedure. It reflects uncertainty in the estimated contrast, not a range containing 95% of participants’ individual changes. Because this interval crosses zero, the posted result does not establish a clear direction of the between-group difference at the conventional two-sided 0.05 significance level.

The p-value of 0.442 is another way of describing how the observed data relate to a null hypothesis of no between-group difference under the analysis assumptions. It is not the probability that the null hypothesis is correct. It also does not establish equivalence: an equivalence conclusion requires a prespecified margin and an analysis framed around that margin, neither of which is reported in the registry record.

Why use repeated-measures modeling?

Generally, a repeated-measures model is designed for data collected at multiple visits from the same participants. Measurements from one participant are related, so treating every visit as independent would fail to represent that within-participant relationship. An MMRM can use repeated observations over time and model differences between treatment groups across visits. The registry record names this method for the primary analysis, but the covariance structure, fixed effects, and any baseline adjustment details are not reported in the registry record.

Generally, repeated-measures models can make use of observed outcomes when some visits are missing, subject to the model and missing-data assumptions. The record says only non-missing data were included. It does not describe the reason for missing observations, the amount of missingness, the assumptions made about missing outcomes, or sensitivity analyses. Those omissions limit the ability to assess how dependent the result may be on incomplete follow-up.

What does the analysis population mean?

The primary outcome population is described as all randomized participants who received at least one dose of study treatment. This definition matters because it is narrower than a definition that includes every randomized participant regardless of treatment received. The registry record does not give further detail about participants who were randomized but did not receive a dose, or how their outcomes were handled.

The outcome groups report 21 participants for placebo and 20 for GSK3196165 180mg. These are the reported group sizes for the primary outcome, while the registry lists enrollment of 44. Safety group counts use 22 participants at risk in each arm. The record does not explain the relationship among these populations and counts, so they should not be treated as interchangeable denominators.

What role does the p-value play?

The p-value is not a measure of effect size or precision. The estimated difference and confidence interval describe the size and uncertainty of the contrast, while the p-value describes how the test statistic compares with what would be expected under the null hypothesis and model assumptions. In general, statistical significance and clinical importance are distinct questions. A result may be statistically distinguishable from zero yet too small to matter, or may be uncertain while still allowing differences that matter. The registry record does not provide a clinical-importance threshold for this endpoint.

The registry also lists many secondary outcomes and analyses. In general, testing multiple outcomes can increase the chance of at least one apparently unusual result if no adjustment or hierarchy is used. Whether and how multiplicity was addressed for this trial’s analyses is not reported in the registry record. The primary result should therefore be read as the posted estimate for its stated endpoint and time point, without assuming an unreported multiplicity procedure.

10. Confidence Intervals and Precision

The sponsor’s 95% confidence interval for the Week 6 mean difference runs from -1.31 to 0.58 scores, around a point estimate of -0.36. The interval includes zero, which corresponds to no between-group difference, and includes values on either side of zero. It therefore conveys substantial uncertainty about both the direction and size of the contrast.

A confidence interval is not a statement that the true effect has a 95% probability of lying inside these particular endpoints. In general, the 95% label refers to the long-run performance of the method: across repeated use under its assumptions, intervals constructed this way would cover the true parameter at the stated rate. For an individual trial, the interval shows which values are compatible with the observed data and the analysis, at the chosen confidence level.

The width of the interval is relevant alongside the point estimate. The estimate alone may appear to favor GSK3196165 because it is negative, but the confidence limits allow for a positive difference as well as a more negative one. The data thus do not identify a precise single effect. The reported p-value of 0.442 is consistent with this interval including zero, but it should not replace direct consideration of the interval and endpoint scale.

The group standard errors are 0.325 for placebo and 0.334 for GSK3196165 180mg. They express uncertainty in the model-based arm estimates, not individual variation in pain scores. The registry record does not report a clinical-importance threshold for the 0-to-10 pain measure, so statistical precision cannot by itself establish whether values within the interval represent meaningful differences. The uncertainty is also conditional on the MMRM and its assumptions, some of which are not detailed in the registry record.

11. Planning a Similar Trial: Sample Size

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

The sample-size figures are planning illustrations based on the observed effect, rather than the trial’s original design justification. All figures are participants per arm, not total events. The stated planning assumptions are a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. The effect used in each scenario is the stated fraction of the observed effect.

For planning at 100% of the observed effect, the illustration gives 270 participants per arm for 80% power and 361 participants per arm for 90% power. At 75% of the observed effect, it gives 479 participants per arm for 80% power and 642 participants per arm for 90% power. At 50% of the observed effect, it gives 1,078 participants per arm for 80% power and 1,443 participants per arm for 90% power.

In plain terms, the 80% power column represents the planning count per arm associated with an 80% chance of detecting the stated effect under the listed assumptions, if that effect were the true effect. The 90% column gives the corresponding planning count for a higher target probability of detection. These are prospective planning quantities, not a description of what happened in the trial.

The pattern illustrates that smaller effects require more participants to detect under the same assumptions. The registry record’s original sample-size assumptions and justification are not reported. These figures are not observed power: they are hypothetical planning calculations based on the observed effect and stated assumptions.

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 = -0.360; pooled standard deviation σ = 1.491.
2. n = 2 × (1.96 + 0.842)² × 1.491² / -0.360² = 269.4, rounded up to 270 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 <- -0.36                      # observed difference in means
sigma <- 1.49146099427236                     # 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(2683785)
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 = 270 (80% power), 361 (90% power)
 75% of observed effect: n per arm = 479 (80% power), 642 (90% power)
 50% of observed effect: n per arm = 1,078 (80% power), 1,443 (90% power)
power.t.test: 270.4 
Simulated power at n = 270 per arm: 0.797 (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. Analyses Reported by the Sponsor

The sponsor’s own primary analysis used a Mixed Model Repeated Measures Analysis and reported a mean difference (net) of -0.36 scores, a 95% confidence interval from -1.31 to 0.58, and p = 0.442. The unadjusted recalculation from posted summary statistics gives a mean difference of -0.36 and p = 0.44. These estimates and tests are close, and the registry information does not flag a discrepancy.

The recalculation is a simple unadjusted comparison based on posted summary statistics; it is not a replacement for the sponsor’s repeated-measures model. The sponsor’s analysis takes precedence because it is the trial’s reported model-based result and uses the stated MMRM method. The two approaches should not be treated as interchangeable: the recalculation does not reproduce the repeated-measures structure or other model details that are not reported.

The sponsor’s confidence interval is the appropriate interval to use when describing the posted primary analysis. It includes zero and extends in both directions, so it indicates uncertainty in the estimated contrast. The close agreement between the point estimates does not remove that uncertainty, nor does it establish that either result represents a clinically important difference.

14. Multiplicity, Missing Data and Analysis Populations

The registry record lists 20 outcome measures, 19 posted secondary outcomes, and 67 analyses. The primary endpoint is one part of a broader set of measures that includes hand-pain outcomes, AUSCAN scores, and counts of soft tissue swollen hand joints. The secondary outcome descriptions also include responder outcomes based on 30 Percentage and 50 Percentage reductions in pain.

Generally, when many outcomes or analyses are examined, the chance of obtaining at least one result that appears unusual under a null hypothesis can increase. Statistical procedures can address this issue through adjustments or through a prespecified hierarchy of hypotheses. Whether any multiplicity adjustment, testing hierarchy, or other control was used is not reported in the registry record. As a result, the primary p-value should be interpreted as the reported test for this endpoint, without assuming that it was adjusted for the wider set of outcomes and analyses.

For the primary analysis, the record states that only non-missing data were included in the MMRM model. It does not report the number of missing observations at each visit, the reasons for missingness, the extent to which participants discontinued, or sensitivity analyses for missing data. In general, repeated-measures models can use observed measurements without requiring every participant to have a measurement at every visit, but conclusions depend on the model and assumptions about missing outcomes. Those assumptions and their assessment are not described in the registry record.

The primary outcome population is defined as randomized participants who received at least one dose. This definition should be kept in mind when interpreting the estimate; it is not described as all randomized participants without qualification. The primary outcome groups list 21 participants for placebo and 20 for GSK3196165 180mg, while the safety counts use 22 participants at risk per arm. The registry record does not explain the differences among these reported group counts. It also does not report further details on handling participants who were randomized but did not receive treatment.

15. 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 (4.5%), humerus fracture (4.5%), hypertensive crisis (4.5%).

The safety record lists 22 participants at risk in each arm. Serious adverse events affected 1 participant in the placebo group and 2 participants in the GSK3196165 180mg group. Other adverse events affected 4 participants in the placebo group and 10 participants in the GSK3196165 180mg group. No deaths are listed in either group.

These counts describe the events recorded for the study populations and follow-up represented in the registry. They are not, by themselves, a complete comparison of risk: interpretation depends on factors such as event definitions, timing, duration of exposure, and the completeness of reporting. The registry record does not provide those details in the information reported alongside these counts. Counts also do not show how severe or persistent non-serious events were.

The serious adverse event terms listed as top serious terms are ATRIAL FIBRILLATION, HUMERUS FRACTURE, and HYPERTENSIVE CRISIS, each reported as 4.5%. The record does not identify an arm for these terms in that list, so they should not be attributed to a specific treatment group based on the posted information. These safety observations are descriptive and do not establish a comparative safety conclusion.

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

The primary outcome records average hand-pain intensity over the 7 days before the Week 6 assessment, compared with baseline. The posted arm estimates show reductions from baseline in both groups: -1.34 scores for placebo and -1.7 scores for GSK3196165 180mg. The sponsor’s estimated net difference was -0.36 scores, with a 95% confidence interval from -1.31 to 0.58 and p = 0.442.

The negative point estimate indicates a greater reduction in the GSK3196165 arm on this scale, but the confidence interval includes zero and also permits a difference in the opposite direction. The result is not precise enough to isolate one direction or size of between-group effect. It should not be described as proof of benefit, proof of no effect, or proof that the groups are equivalent.

Statistical uncertainty is only one part of interpretation. The registry record does not report a threshold for a meaningful difference on the pain scale, so it is not possible to determine from this result alone whether the point estimate or the values covered by the interval would be important to participants. The observed score contrast also does not summarize every outcome measured during the study or establish what might happen beyond the stated assessment.

The trial was a Phase IIa study with recorded enrollment of 44. Its primary outcome groups list 21 participants for placebo and 20 for GSK3196165 180mg. Those group counts and the uncertainty in the confidence interval matter when considering how much information the result contains. The posted result is evidence about a specific comparison, outcome, and time point under the stated analysis; it is not a general conclusion about all people with hand osteoarthritis or all uses of GSK3196165.

The registry also reports descriptive adverse-event counts, including serious events and other events, with no deaths listed. Such counts should be read alongside their denominators and the limits of the available detail, not as a definitive comparative safety assessment. Taken together, the posted findings describe the measured Week 6 pain contrast and reported safety counts, while leaving important questions about precision, clinical importance, missing data, and broader outcomes unresolved.

17. Why This Trial Matters Statistically

This trial illustrates why a between-group comparison is more informative than looking at whether pain scores fell within either arm. The posted average changes were negative in both groups, but the main statistical question is whether the changes differed between GSK3196165 and placebo. The sponsor’s net difference of -0.36 scores addresses that contrast directly.

It also demonstrates the importance of uncertainty intervals. The point estimate is negative, yet the 95% confidence interval from -1.31 to 0.58 includes zero and values on either side. Without the interval, the point estimate could be read too confidently as a clear directional result. The interval makes visible that the data do not precisely determine the size or direction of the contrast.

The repeated-measures model is another important feature. Pain was collected over time, and MMRM is a method designed generally for repeated observations. Knowing the method helps readers understand the type of analysis reported; recognizing that details such as the model terms and missing-data assumptions are not reported helps keep interpretation appropriately bounded.

Finally, the planning illustration shows how the required sample size depends on the effect one aims to detect and the desired power. Smaller assumed effects require larger participant counts under the stated planning assumptions. These calculations are distinct from the trial’s own planning rationale, which is not reported in the registry record. Together, the endpoint, model, interval, and planning assumptions show why statistical interpretation requires more than a single p-value.

18. Limitations

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

The primary outcome result is for Week 6 and does not, by itself, describe pain at every later follow-up visit. The study’s planned follow-up completed at Week 22, but the registry record does not provide a Week 22 result in the primary outcome information. The posted result should therefore be kept tied to its stated endpoint and assessment time.

The endpoint used daily self-reported pain scores averaged over the 7 days before a visit. The scale anchors are stated, but the registry record does not report a threshold for a meaningful difference. That limits interpretation of the estimated difference in terms of practical importance. The posted summary also does not show the distribution of individual participant changes, so the group means do not indicate how responses varied across participants.

The record names MMRM but does not give the full model specification, including model terms and covariance structure. It also does not report the extent or pattern of missing data, reasons for missingness, or sensitivity analyses. Since only non-missing data were included, the assumptions behind the use of observed data are relevant but cannot be fully evaluated from the registry record.

The registry reports 44 enrolled participants, primary outcome group sizes of 21 and 20, and safety group counts of 22 at risk per arm. It does not explain the relationships among these counts. The study’s original sample-size justification and planned power assumptions are also not reported; the available sample-size illustration is based on the observed effect rather than the original design.

Finally, the record lists numerous outcome measures and analyses, but does not report whether multiplicity was addressed. The serious adverse event terms are listed without arm attribution in the reported top-term summary. These limitations mean the primary result and safety counts should be interpreted as the specific posted findings, with uncertainty about additional analytic and reporting details.

19. Related Tutorials

20. Related Statistical Calculators

21. Sources

• ClinicalTrials.gov: A Study to Investigate the Efficacy and Safety of GSK3196165 in Inflammatory Hand Osteoarthritis — NCT02683785.
• Registry record last updated 2021-01-11; status last verified 2020-12.
• Results first posted 2018-12-14.
• Data retrieved from ClinicalTrials.gov on 2026-11-09.
• Group sizes, means, event counts and sponsor-reported p-values are reproduced as posted. Effect estimates, confidence intervals and p-values not attributed to the sponsor are independently calculated from the posted summary statistics.

22. Record Summary

NCT02683785 is a completed randomized Phase IIa, multicentre, double-blind, placebo-controlled parallel-group study sponsored by GlaxoSmithKline. It investigated GSK3196165 in subjects with active inflammatory hand osteoarthritis. Recorded enrollment was 44, with a 12-week treatment period and follow-up completing at Week 22.

The primary endpoint was change from baseline in 24-hour average hand-pain intensity, averaged over the 7 days before Week 6. The sponsor’s MMRM analysis reported a mean difference of -0.36 scores for GSK3196165 180mg versus placebo, with a 95% confidence interval from -1.31 to 0.58 and p = 0.442. The interval includes zero, indicating uncertainty about the direction and size of the between-group difference.

The safety record lists serious adverse events affecting 1 participant in the placebo group and 2 in the GSK3196165 group; no deaths are listed. Original sample-size justification, multiplicity procedures, and detailed missing-data assumptions are not reported in the registry record.