Optimizing HPV Vaccine Introduction in Shanghai, China: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Optimizing HPV Vaccine Introduction in Shanghai, China (NCT03972813), a randomized na study. The primary outcome is To Determine How the Framing of the HPV Vaccination Across Several Dimensions Affects Short-term Willingness to Receive it.

2. Clinical Question

3. Trial Design

The registry describes a randomized, unmasked factorial study of educational messages about HPV vaccination. It lists enrollment of 1021 and identifies the study as terminated. The listed phase is NA. The sponsor is the University of Michigan. The registry record gives a start date of 2019-05-28 and a primary completion date of 2019-08-31.

The experiment varied message content across dimensions that included the cancers associated with HPV, whether HPV was characterized as sexually transmitted or as an infectious disease or was left without communicability information, and whether the message referred to vaccination at 12 years old, 18 years old, or gave standard age information. The posted arms include combinations of these message features. The arm descriptions illustrate that the experimental contrast concerns information and framing, rather than receipt of a vaccine.

The posted comparisons use the cervical-cancer, sexually transmitted, standard message as their reference. Comparisons are shown separately for the other message combinations, rather than as one overall factorial estimate. Although the registry labels the design factorial, estimated main effects or interactions are not reported in the registry record. The record also does not report how randomization was generated, whether allocation was concealed, or how the factorial dimensions were incorporated into the analysis.

The primary outcome was assessed immediately after the experiment. A secondary outcome concerns sustained willingness, but its measurement results are not included among the posted comparisons. The registry status is terminated, but the reason for termination is not reported in the registry record. The record does not report a detailed protocol or the original sample-size justification.

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.

Only the first 6 of 18 groups are shown.

Overall Study

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.

Only the first 6 of 19 groups are shown.

6. Primary Endpoint

The registry lists the primary outcome as To Determine How the Framing of the HPV Vaccination Across Several Dimensions Affects Short-term Willingness to Receive it, measured over same day as intervention.

The primary endpoint was the count of participants who indicated that they were somewhat willing or very willing to give a daughter an HPV vaccine at 12 years of age. Parents answered immediately after the informational experiment, using a 5-point response scale anchored by not at all willing at 1 and very willing at 5. The registry record specifies that responses of 4 or 5 were counted toward the primary outcome.

This endpoint turns a graded response into a binary outcome: counted as willing according to the stated threshold, or not counted as willing. Such a summary can make the proportion meeting a prespecified response threshold straightforward to compare across randomized message groups. It also means the endpoint does not preserve all distinctions among the five response categories. For instance, the primary count does not show how many participants selected each individual scale point.

The outcome is an immediate, self-reported intention in a hypothetical question about a daughter. It is not a measure of receiving a vaccine, booking an appointment, or sustained willingness. The registry lists sustained willingness as a secondary outcome, but results for that outcome are not reported in the registry record. Thus the primary endpoint addresses short-term expressed willingness after exposure to a message, not subsequent behavior or long-term change.

7. Results Posted in the Registry

• Cervical Cancer - STD - 12 Years Old (y.o.): 68.9% (31/45) versus 58.2% (46/79) on control; risk difference 10.7 percentage points (95% CI -6.7 to 28.0), risk ratio 1.18 (0.90–1.55); unadjusted two-proportion z-test p = 0.239.
• Cervical Cancer - STD - 18 y.o.: 55.8% (29/52) versus 58.2% (46/79) on control; risk difference -2.5 percentage points (95% CI -19.8 to 14.9), risk ratio 0.96 (0.71–1.30); unadjusted two-proportion z-test p = 0.781.
• Cervical Cancer - Infectious - Standard: 41.2% (21/51) versus 58.2% (46/79) on control; risk difference -17.1 percentage points (95% CI -34.4 to 0.3), risk ratio 0.71 (0.48–1.03); unadjusted two-proportion z-test p = 0.058.
• Cervical Cancer - Infectious - 12 y.o.: 58.3% (28/48) versus 58.2% (46/79) on control; risk difference 0.1 percentage points (95% CI -17.6 to 17.8), risk ratio 1.00 (0.74–1.36); unadjusted two-proportion z-test p = 0.991.

8. Statistical Interpretation

9. Statistical Methods Explained

The principal effect measure in the posted comparisons is the risk difference, which is the difference between the proportion meeting the willingness threshold in a message arm and the proportion in the reference arm. A positive value indicates a higher observed proportion in that arm; a negative value indicates a lower observed proportion. A value near zero indicates similar observed proportions. This is a general explanation of the measure, not an additional trial result.

The reference proportion is reported as 0.582. For each comparison, the registry results also show the arm-specific proportion and counts. For example, the cervical-cancer, sexually transmitted message with the 12-year-old framing has a reported proportion of 0.689 and a risk difference of 0.107. These are posted summary results; they do not by themselves adjust for participant characteristics or for the factorial structure.

A confidence interval expresses the uncertainty around an estimated effect under the method used to construct it. In general, a wider interval means the data allow a broader range of plausible effects, while a narrower interval indicates greater precision. If an interval includes zero for a risk difference, the data are compatible with no difference as well as with effects on either side of zero, over the interval’s range. This does not prove that the true difference is zero.

A p-value describes how unusual data at least as inconsistent with a specified null hypothesis would be if that hypothesis and the test assumptions held. It is not the probability that the null hypothesis is true, nor a measure of the size or practical importance of an effect. The posted p-values are unadjusted. With multiple comparisons, chance alone can produce apparently notable results among the collection; the registry record does not report a multiplicity adjustment or a prespecified testing hierarchy.

The registry identifies the outcome as binary and reports counts of participants with the event. A binary analysis compares whether participants crossed the specified response threshold. The posted comparisons also include risk ratios and odds ratios with intervals, but risk differences provide an absolute-scale contrast that can be read as the difference between the observed proportions. The record does not report a sponsor analysis or an analysis model for these comparisons.

Randomization is intended, in general, to make groups comparable with respect to measured and unmeasured characteristics on average. It does not guarantee that each small arm will have identical characteristics, and the registry record does not provide baseline characteristics to assess the realized groups. The record says the design is randomized and factorial, but it does not report the randomization procedure, allocation concealment, or whether the analysis accounted for factorial main effects or interactions.

The registry outcome counts allow the posted arm-level proportions to be read directly. However, the record does not report how missing or incomplete responses were handled, whether every randomized participant was included in the relevant analysis, or the analysis population definition. It also does not report an adjustment for clustering, covariates, or repeated measurements. These details matter because they affect how an estimate and its uncertainty should be interpreted.

General statistical context: when an outcome is binary, uncertainty depends on both the observed proportions and the number of participants contributing to each group. Smaller groups tend to yield less precise estimates, particularly when the observed difference is small. The posted intervals are broad for many comparisons, consistent with substantial uncertainty about the size and direction of the differences. This is not a claim that a particular model was used; the detailed inferential method is not reported in the registry record.

The outcome was measured on the same day as the intervention, while sustained willingness is listed separately as a secondary outcome. A same-day measure can describe an immediate response to the information, but it cannot establish that the response persisted. Nor does a hypothetical willingness response establish behavior. Those distinctions define the scope of the primary endpoint.

10. Confidence Intervals and Precision

The posted confidence intervals are important because the individual message arms contain relatively small groups, and the intervals are broad in many comparisons. For example, the risk difference for the cervical-cancer, sexually transmitted message with the 12-year-old framing is 0.107, with a 95% confidence interval from -0.067 to 0.280. That range includes both a negative difference and a positive difference, so the point estimate alone gives an incomplete picture of the uncertainty.

The cervical-cancer, infectious-disease, standard-age estimate is -0.171, with a 95% confidence interval from -0.344 to 0.003. The many-cancers, blank-communicability, standard-age estimate is -0.142, with a 95% confidence interval from -0.318 to 0.033. These intervals also include zero. The other reported risk-difference intervals likewise include zero, and several estimates lie close to zero while their intervals still span appreciable differences in either direction.

In general, a 95% confidence interval is a range of values compatible with the observed data under the interval method’s assumptions; it is not a guarantee that the true effect lies in that particular calculated range. Wider intervals reflect lower precision. The intervals should not be read as evidence of equivalence: establishing equivalence requires a defined margin and an appropriate analysis, neither of which is reported in the registry record.

All posted pairwise intervals use the same reference arm. They are not a complete account of uncertainty across the multiple comparisons, and the registry record does not report a correction for multiplicity. Overall, the displayed point estimates vary, but the uncertainty is substantial. The results do not precisely identify the differences among message framings.

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 for the binary-outcome comparisons. They are not total event counts. The calculations use a two-sided significance level of 0.05, equal allocation between the two arms, and a normal approximation. For each comparison, the effect to detect is set at the stated fraction of the observed effect. These are planning illustrations based on the observed effect, not the trial’s original design assumptions, which are not reported in the registry record.

The 80% power column gives the number of participants per arm that the planning calculation estimates would be needed to have an 80% chance of detecting the specified effect under the stated assumptions, if that effect were the true effect and the assumptions held. The 90% power column expresses the same kind of planning target at 90% power. These percentages describe a planned study’s operating characteristics; they do not describe the probability that the posted finding is true.

Rows planning for 100% of the observed effect use the full observed estimate as the effect to detect. Rows planning for 75% or 50% use smaller effects. Smaller effects generally require more participants to distinguish them from no difference with the same power and significance level. When the observed estimate is close to zero, the illustrations can become extremely large; this reflects the difficulty of detecting a very small difference, and should not be interpreted as a practical recruitment recommendation.

These numbers are not observed power. Observed group sizes and outcomes are reported separately; a planning calculation cannot make an uncertain estimate more conclusive after the fact. The registry record does not report the trial’s own sample-size justification or assumptions, so the illustrations should not be confused with the study’s intended enrollment 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, p₀ and p₁ are the control and treatment proportions, and p̄ is their average.

Worked calculation for this trial

1. Control proportion p₀ = 0.582; treatment proportion p₁ = 0.689; average p̄ = 0.636.
2. First term: z_1−α/2 × √[2p̄(1−p̄)] = 1.96 × 0.6806 = 1.3340.
3. Second term: z_1−β × √[p₀(1−p₀) + p₁(1−p₁)] = 0.842 × 0.6764 = 0.5693.
4. n = (1.3340 + 0.5693)² / (0.1066)² = 318.7, rounded up to 319 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
p0 <- 0.582278481012658                          # control proportion
p1 <- 0.688888888888889                          # treatment proportion
n_per_arm <- function(power, frac = 1) {
  q1 <- p0 + (p1 - p0) * frac               # smaller effect = fraction of the observed difference
  pbar <- (p0 + q1) / 2
  z_b <- qnorm(power)
  ceiling(round((z_a * sqrt(2 * pbar * (1 - pbar)) + z_b * sqrt(p0 * (1 - p0) + q1 * (1 - q1)))^2 / (q1 - p0)^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 (unrounded n per arm)
cat("power.prop.test:", round(power.prop.test(p1 = p0, p2 = p1, power = 0.80)$n, 1), "\n")

# Simulation: how often does the z-test reject at the 80%-power sample size?
set.seed(3972813)
n    <- n_per_arm(0.80)
reps <- 20000
x0 <- rbinom(reps, n, p0); x1 <- rbinom(reps, n, p1)
ph <- (x0 + x1) / (2 * n)
z  <- ((x1 - x0) / n) / sqrt(ph * (1 - ph) * 2 / n)
cat(sprintf("Simulated power at n = %s per arm: %.3f (%d replicates)\n", format(n, big.mark = ","), mean(abs(z) > z_a, na.rm = TRUE), reps))

Output from running the code

100% of observed effect: n per arm = 319 (80% power), 427 (90% power)
 75% of observed effect: n per arm = 576 (80% power), 771 (90% power)
 50% of observed effect: n per arm = 1,315 (80% power), 1,760 (90% power)
power.prop.test: 318.7 
Simulated power at n = 319 per arm: 0.805 (20000 replicates)

The simulation draws many pairs of event counts at the planned sample size and counts how often the two-proportion z-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 posted results compare multiple message combinations with one common reference arm. When many comparisons are examined, the chance of obtaining one or more small p-values through random variation can increase. This is a general statistical concern called multiplicity. The posted p-values are unadjusted, and a multiplicity correction, primary contrast hierarchy, or strategy for interpreting the collection of comparisons is not reported in the registry record. Individual p-values should consequently be interpreted cautiously and in the context of the full set of contrasts.

The study is identified as factorial, but the posted results are arm-by-arm comparisons against the reference. Estimates of factorial main effects, interactions among the message dimensions, or an overall test of the factorial structure are not reported in the registry record. The record also does not specify whether these comparisons were planned in advance or selected after examining the data.

Missing-data procedures and analysis populations are not reported in the registry record. It is not stated whether all randomized participants were included, how participants with no outcome response were handled, or whether any imputation or sensitivity analysis was used. The registry lists 1021 enrolled participants, while outcome counts are posted for the individual arms; the record does not explain the relationship between enrollment and the analysis counts.

The record does not report a statistical analysis plan, covariate adjustment, a protocol-defined approach to interim analyses, or handling of deviations from assigned messages. The number of analyses reported in the registry is 0. The posted unadjusted calculations therefore offer limited information about confirmatory inference, and important details needed to assess the analysis process are not reported in 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 0% reporting threshold.

The registry reports zero serious adverse events, zero other adverse events, and zero deaths in each of the listed study groups. The number at risk is reported separately by group, ranging across the individual message arms as displayed in the safety results. These are counts as recorded in the registry; they are not estimates of comparative safety effects.

This was an educational experiment involving information about vaccination, not a trial in which the listed groups received a vaccine as the study intervention. The safety counts therefore describe events recorded among participants in the study groups, rather than adverse events following study vaccination. The registry record does not provide serious adverse event terms or further information on event collection and follow-up.

Zero reported events means none were recorded in the posted categories for the stated groups and reporting period. It does not establish that no events occurred outside the registry’s reporting or ascertainment, and it cannot precisely characterize rare risks. The adverse-event reporting detail and the duration of safety follow-up are not reported in the registry record.

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

The results describe how parents in the reported groups answered an immediate hypothetical question after receiving different HPV-vaccine messages. The measured response was whether they selected one of the two higher willingness categories on a 5-point scale. The comparisons show variation in the observed proportions across arms, but their confidence intervals are broad and include no difference. The posted numbers therefore do not precisely establish that any particular message framing changed short-term stated willingness relative to the reference message.

The endpoint is an attitude or intention measure, not actual vaccine uptake. A response about what a parent would do if they had a daughter does not show whether a child was vaccinated, whether an appointment was made, or whether willingness persisted. Although sustained willingness is listed as a secondary outcome, its results are not reported in the registry record.

The estimates concern particular combinations of informational features, each compared with the standard cervical-cancer, sexually transmitted message. They do not, on their own, isolate the effect of cancer framing, transmission language, or age framing across all other dimensions. The record labels the design factorial but does not report factorial effect estimates or interactions. The separate comparisons should not be treated as a definitive ranking of every dimension.

The trial status is terminated. The reason is not reported in the registry record, and that status alone does not explain why the study ended or how it affects the available results. The posted findings are best understood as imprecise, short-term responses in this educational experiment. They do not establish what would happen in routine vaccination settings or whether a particular message would change vaccination behavior.

16. Why This Trial Matters Statistically

Statistically, this trial illustrates the distinction between an observed arm-level difference and a precise estimate of that difference. The posted proportions differ across message groups, but the confidence intervals are broad. Reading the point estimates alone could overstate how clearly the data distinguish one framing from another. Interpreting estimates together with their intervals keeps attention on both direction and uncertainty.

The factorial structure also raises a central design-and-analysis issue: several informational dimensions are combined in the message arms. Separate arm comparisons can describe those combinations, but they do not automatically identify the independent contribution of each dimension. Main-effect and interaction estimates would address different questions, and they are not reported in the registry record.

Multiplicity is relevant because many arms are compared with a common reference. An isolated unadjusted p-value should not be interpreted without considering the full set of comparisons and the absence of a reported adjustment strategy. The record’s lack of detailed analysis-plan and missing-data information further limits assessment of confirmatory inference.

Finally, the endpoint makes the estimand explicit: immediate stated willingness, categorized by a response threshold. That choice is useful for describing a short-term response, but it does not measure sustained intention or actual behavior. The trial therefore offers a setting for learning how endpoint definition, factorial contrasts, multiple comparisons, and precision shape the interpretation of an educational experiment.

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.

The primary outcome is self-reported willingness assessed immediately after the message exposure. It is a hypothetical response about a daughter and does not establish future vaccination behavior. The binary threshold combines responses of somewhat willing and very willing, while the registry does not post the distribution across all five response categories.

Individual message arms have modest posted sample sizes, and most risk-difference intervals are wide. This limits precision for comparing message combinations. Because numerous combinations are compared with a shared reference, the unadjusted results also raise multiplicity concerns; no correction or hierarchy is reported in the registry record.

Although the study is described as factorial, the posted estimates are separate arm comparisons. The registry record does not report estimates that isolate the factorial dimensions or their interactions. It also does not report baseline characteristics, allocation procedures, missing-data handling, analysis populations, a detailed statistical plan, or the original sample-size justification. These omissions limit assessment of comparability, analysis decisions, and the intended precision.

The study is listed as terminated, but its reason for termination is not reported in the registry record. The secondary outcome of sustained willingness is listed, but results are not reported there. The registry also does not explain how enrollment relates to the participants included in each outcome arm. Together, these limitations mean that the posted results provide a partial and uncertain account of immediate stated willingness, rather than a complete evaluation of sustained response or behavior.

18. Related Tutorials

19. Related Statistical Calculators

20. Sources

• ClinicalTrials.gov: Optimizing HPV Vaccine Introduction in Shanghai, China — NCT03972813.
• Registry record last updated 2025-01-15; status last verified 2025-01.
• Results first posted 2020-01-13.
• 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.

21. Record Summary

NCT03972813, Optimizing HPV Vaccine Introduction in Shanghai, China, is listed as a randomized, unmasked factorial educational experiment sponsored by the University of Michigan. The registry reports enrollment of 1021 and status terminated. The primary outcome is the same-day count of parents selecting somewhat willing or very willing on a 5-point question about giving a daughter an HPV vaccine at 12 years of age.

The posted arm results report counts and proportions, with unadjusted risk-difference comparisons against the standard cervical-cancer, sexually transmitted message. The estimates vary, but the posted confidence intervals include zero and are often broad. The registry record does not report factorial effect estimates, multiplicity handling, missing-data procedures, or the reason for termination. The results describe immediate stated willingness and do not measure vaccine uptake.