Testing New Ways to Name Antimicrobial Resistance: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Testing New Ways to Name Antimicrobial Resistance (NCT06356285), a randomized na study. The primary outcome is Percentage of Correct Answers by Participants on Questions Designed to Evaluate Attitudes Towards Antimicrobial Resistance (Sentiment).

2. Clinical Question

3. Trial Design

The registry describes a randomized, parallel-group online study with single masking. Participants were assigned to one of four poster framings: the control wording “Antimicrobial Resistance” or one of the alternatives, “Superbugs”, “Antibiotic Resistance” and “The Antibiotic Crisis”. The intervention is a communication presentation, not a medicine or clinical procedure.

Eligible participants were adults aged over 18, living in the United Kingdom and already signed up to a market panel research company. They were invited to complete an anonymous online survey, described as taking around five minutes. The survey included multiple-choice items and short free-text responses. The record describes the intended population as adults representative of the UK population who responded to the survey question, but does not report the method used to recruit or weight participants to achieve that representation.

The registry lists enrollment as 4296. The trial is marked completed, with a start date of 2024-04-10 and a primary completion date of 2024-04-18. The primary outcome and the posted secondary outcomes have a timeframe of 6 weeks. The record does not explain how that timeframe relates to the short survey described in the study summary.

Single masking is recorded, but the registry record does not identify which people involved in the study were masked or explain how masking was maintained. It also does not report the sequence-generation method, allocation concealment procedures, or whether participants could infer their assignment from the poster wording. Those details matter when assessing the risk of selection or response bias, but cannot be established from the information reported.

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.

Overall Study

Analysis population for the primary outcome, as defined in the registry: All adults (over the age of 18, representative of the UK population) who responded to the survey question.

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: Adults residing in the United Kingdom who accessed the online survey and passed the attention check.

6. Primary Endpoint

The registry lists the primary outcome as Percentage of Correct Answers by Participants on Questions Designed to Evaluate Attitudes Towards Antimicrobial Resistance (Sentiment), measured over 6 weeks.

The primary endpoint is the percentage of correct answers to survey questions designed to evaluate attitudes toward antimicrobial resistance, described in the registry as sentiment. The questions concern agreement that antimicrobial resistance is urgent and needs to be addressed, that it poses a risk to human health, and that it may affect the participant’s own health. The analysis description says the primary outcome was the sum of binary ratings from sub-sentiment questions, analyzed using a quasi-binomial regression model.

This outcome is useful for testing whether alternative communication framings are associated with how participants respond to a set of attitude questions. It translates answers to multiple binary items into a measure of correct responses, allowing the study to compare sentiment across the poster groups. It is not a direct measure of antibiotic use, actual requests for antibiotics, or later health outcomes. Nor does the registry record define what makes a response “correct” for each sentiment item, explain how unanswered items were handled, or report the detailed scoring rules. Those omissions limit independent evaluation of exactly what the percentage represents.

The registry reports arm-level percentages of 81.0 for the control, 83.2 for “Superbugs”, 86.3 for “Antibiotic Resistance” and 85.2 for “The Antibiotic Crisis”. These are descriptive values; they do not by themselves establish whether the observed differences are compatible with random variation or support a particular causal interpretation. That assessment depends on the reported model, its comparison definition and uncertainty, details that are incomplete in the registry record.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The trial is described as randomized and parallel group. In a randomized study, assignment is intended to make comparison groups similar, on average, with respect to factors that could influence the outcome. The parallel-group structure means each participant receives one assigned poster framing rather than being described as crossing between framings. Random assignment can support a causal comparison of the assigned framings when allocation, follow-up, outcome measurement and analysis are appropriately implemented. The registry record does not provide enough detail to assess all of those conditions.

What does a percentage of correct answers capture?

A percentage summarizes the share of responses meeting the study’s scoring definition. Here, the primary endpoint combines binary ratings from sentiment questions. Binary outcomes record whether an answer falls into one of two categories, such as meeting or not meeting a defined criterion. Combining such items can be a practical way to represent responses across related questions, but interpretation depends on the scoring key, the number and content of items, and the handling of missing answers. The registry record does not report the full scoring rules or item-level results. The percentage should therefore be understood as the study’s operational measure of sentiment, not as a general test of knowledge or a direct measure of conduct.

Why use a quasi-binomial regression model?

A binomial model is commonly used for binary responses or counts of successes among a defined set of trials. It makes assumptions about the relationship between the expected outcome and its variability. When observed variability is greater than the ordinary binomial model allows, over-dispersion can make conventional uncertainty estimates too optimistic. A quasi-binomial approach allows the variance to be scaled to account for that extra variability. This is a general statistical explanation; the registry does not report a dispersion estimate, model diagnostics, covariates, or other details needed to assess the specific fit.

The primary outcome description says that the sum of binary ratings from sub-sentiment questions was modeled. The reported analysis is labeled “Quasibinominal regression” in the registry and gives an odds ratio on a log scale. The record does not fully explain the model’s parameterization, its reference category, or how the three intervention arms were represented in the reported result. Since an odds ratio is a relative measure, its interpretation depends on the defined outcome and contrast. A log-scale estimate also requires clear specification of the model before it can be translated into a meaningful comparison. The registry’s short result does not provide that complete specification.

What does a confidence interval communicate?

A confidence interval expresses uncertainty around an estimated effect under the assumptions of the statistical method. A narrower interval indicates greater precision than a wider interval, while an interval that includes the no-difference value is compatible with a range of effects that includes no difference. It does not show the probability that the true effect lies within that particular interval. The sponsor’s primary result has no confidence interval reported in the registry record, which prevents readers from judging the estimate’s precision from the posted analysis.

For secondary outcomes, the recalculated comparisons include risk differences with confidence intervals. A risk difference expresses the difference between two proportions on the proportion scale; its sign indicates which group has the higher proportion, given the stated group order. These recalculations are unadjusted comparisons derived from posted summary statistics. They should not be confused with a multivariable model or a sponsor-reported analysis. The registry record lists no sponsor analysis for these secondary outcomes.

How should p-values be read?

A p-value describes how incompatible the observed data are with a specified null model, assuming the model and its assumptions are appropriate. It does not measure the probability that the null hypothesis is true, the probability that a finding is due to chance, or the size or importance of an effect. The primary p-value is reported alongside a model estimate but without a confidence interval or a fully described contrast. For the recalculated secondary comparisons, p-values are unadjusted, so they do not account for the fact that multiple outcomes and comparisons are considered.

The sample-size illustrations use the observed effect as a planning reference and are not observed power. The stated assumptions are a two-sided significance level of 0.05, equal allocation between two arms, a normal approximation, and detection of a stated fraction of the observed effect. Such illustrations show how a planning calculation changes when the target effect is smaller than the observed estimate. They do not establish what sample size the original trial required, what its design assumptions were, or the probability that its observed results are true. The registry record does not report the trial’s own sample-size justification.

10. Analyses Reported by the Sponsor

The sponsor’s primary analysis is reported as quasi-binomial regression, with an odds ratio on a log scale of 0.05 and p = 0.017. The registry labels this an equivalence analysis. It does not report a confidence interval, the equivalence margin, or a sufficiently detailed contrast definition. Those missing elements restrict interpretation of the estimate and make it difficult to assess whether the analysis supports the stated equivalence question.

No recalculated primary comparison is available in the posted data. The arm-level percentages alone are not enough to reconstruct the sponsor’s model, especially because the endpoint is described as a sum of binary sentiment ratings and the reported model’s contrast is not fully specified. Therefore, a simple comparison based on those percentages should not be substituted for the sponsor’s model result.

The secondary risk-difference comparisons are recalculations from posted summary statistics, not sponsor analyses. They are explicitly unadjusted and concern separate secondary outcomes. They do not validate or overturn the primary model result. The registry record does not mark a discrepancy between a sponsor result and a recalculation. The appropriate reading is to keep the primary sponsor analysis distinct from the secondary descriptive and recalculated comparisons, while recognizing that the registry does not report enough detail to resolve the primary result’s interpretation fully.

11. Multiplicity, Missing Data and Analysis Populations

The study has a primary endpoint and multiple secondary outcomes, with comparisons between several intervention framings and the control. Considering many outcomes or contrasts raises the chance that at least one result will appear statistically notable even when the underlying effects are absent. A prespecified multiplicity strategy can address this issue by controlling the interpretation of a family of tests, or by clearly distinguishing confirmatory outcomes from exploratory ones.

The registry record reports one primary analysis and lists secondary outcomes, but does not report a multiplicity adjustment, a hierarchy of secondary hypotheses, or which secondary comparisons were prespecified. The recalculated secondary p-values are unadjusted. They should consequently be treated as exploratory evidence rather than as a set of independent confirmatory findings. The primary endpoint’s reported p-value does not resolve multiplicity questions about the broader set of outcomes.

Missing-data procedures are not reported in the registry record. The primary population is described as respondents to the survey question, but the record does not state how many people were invited, how many started or completed the survey, how missing item responses were treated, or whether any participants were excluded from the endpoint analysis. The arm-level endpoint counts are reported, but the registry does not explain whether they reflect all randomized participants or only those with usable outcome data. Without those details, it is not possible to assess the effect of nonresponse or incomplete answers.

The analysis population is also not fully specified. The registry does not report whether analyses followed intention-to-treat principles, whether participants were analyzed according to assigned framing, or whether there were exclusions after randomization. These details matter because analyses that omit participants or reclassify assignment can weaken the protection offered by randomization. The record also does not report a protocol or statistical analysis plan that could clarify planned outcomes, exclusions and decision rules. Accordingly, the posted findings should be interpreted as results reported in the registry, with uncertainty about the handling of missing data and the completeness of the analysis population.

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

The results describe responses to alternative ways of naming antimicrobial resistance in an online survey. They show variation across framings in the reported sentiment, comprehension, recall and stated-intention outcomes. This is evidence about how survey respondents answered after seeing the assigned poster, in the setting described by the registry.

These findings do not establish that a particular wording will change antibiotic use, improve health outcomes, or produce lasting changes in knowledge or behaviour. Stating an intention to request antibiotics is not the same as making a request, and recalling a term is not the same as understanding its implications. The primary sentiment measure is based on survey responses rather than observed actions. The registry does not report follow-up evidence establishing whether responses persisted beyond the survey or generalized to other settings.

The results also do not establish that one framing is universally preferable. The sponsor’s primary model result lacks a confidence interval and a clearly described contrast in the registry record, while the secondary comparisons are unadjusted recalculations. The posted arm-level percentages are useful descriptions, but they are not a substitute for a complete analysis with uncertainty and clear definitions of the comparisons being tested.

Interpretation should also account for the study population and delivery method. Participants were adults in the United Kingdom recruited through a market panel research company and responded to an anonymous online survey. The registry describes the intended population as representative of the UK population, but does not report enough recruitment or weighting detail to assess how well the analyzed respondents represent that population. The study therefore speaks most directly to responses in the surveyed setting, not necessarily to all audiences, communication channels or real-world situations.

In short, the trial examines a relevant communication question and reports differences among survey responses, but its registry record leaves important inferential and implementation questions unresolved. It cannot, on its own, show that a wording causes meaningful behaviour change or that any observed difference would persist beyond the survey.

13. Why This Trial Matters Statistically

This study is statistically interesting because it treats communication framing as an experimental intervention and compares alternative labels using random assignment. Rather than assuming that technical terminology is understood in the same way by everyone, the design tests whether different language is associated with different survey responses. That creates a measurable question about the relationship between wording and attitudes, comprehension, recall and stated intentions.

The outcomes also illustrate why communication effects should not be reduced to a single favorable or unfavorable response. Sentiment, comprehension, recall and intention capture different constructs. A change in one does not necessarily imply a change in another, and survey measures do not automatically translate into action. Reading the outcome domains separately is important for defining what a communication intervention has actually affected.

The trial also highlights the importance of transparent analysis reporting. A model label, estimate and p-value are not enough to fully interpret an equivalence analysis when the contrast, confidence interval and equivalence margin are not reported. Similarly, unadjusted secondary comparisons need to be distinguished from prespecified confirmatory analyses. Clear reporting of outcome definitions, model specifications, missing-data handling and multiplicity decisions would make the evidence easier to evaluate and reproduce.

Finally, the record illustrates a useful distinction between a trial’s observed results and planning calculations derived from those results. Sample-size illustrations can inform future design discussions, but they are not observed power and do not explain the original trial’s design rationale. Careful separation of these concepts supports more credible interpretation of communication research and helps prevent descriptive survey findings from being overstated as evidence of real-world behaviour change.

14. Limitations

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

The registry’s primary outcome description gives a broad account of the sentiment questions but does not report the exact items, scoring key, or the rules used to determine a correct answer. Without that information, readers cannot independently assess how the endpoint was constructed or whether all items contribute comparably to the reported percentage.

The primary regression result is difficult to interpret fully because the record does not give a confidence interval or clearly define the contrast represented by the estimate. It labels the analysis as an equivalence analysis but does not report the equivalence margin. These are important omissions for evaluating the strength and meaning of the primary finding.

The anonymous online survey and market-panel recruitment may limit how closely the sample represents the broader public, despite the registry’s description of a representative UK population. The recruitment, response and weighting details needed to assess representativeness are not reported. Participants saw a poster image in a survey context; the registry does not establish how that exposure compares with encountering communication in everyday settings.

Several important design and analysis details are not reported in the registry record, including the method of random sequence generation, allocation concealment, who was masked, analysis-population rules, treatment of missing answers, and a multiplicity strategy. The record also does not explain the relationship between the 6 weeks outcome timeframe and the short survey described in the study summary. These limitations constrain assessment of bias, precision and generalizability.

Finally, the posted secondary comparisons are unadjusted recalculations, and sponsor analyses are not reported for those outcomes. They may be informative as descriptive comparisons but should not be given the same weight as a clearly specified, prespecified primary analysis. The outcomes are survey responses and do not establish durable change or observed behaviour.

15. Related Tutorials

16. Related Statistical Calculators

17. Sources

• ClinicalTrials.gov: Testing New Ways to Name Antimicrobial Resistance — NCT06356285.
• Registry record last updated 2025-01-14; status last verified 2025-01.
• Results first posted 2025-01-14.
• 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.

18. Record Summary

NCT06356285, titled “Testing New Ways to Name Antimicrobial Resistance”, is a completed randomized, single-masked, parallel online trial sponsored by Imperial College London. The registry lists enrollment as 4296 and dates the study from 2024-04-10, with primary completion on 2024-04-18.

Adults over 18 living in the United Kingdom and signed up to a market panel research company were eligible. Participants viewed one of four poster framings: “Antimicrobial Resistance”, “Superbugs”, “Antibiotic Resistance” or “The Antibiotic Crisis”. The primary endpoint assessed correct responses to survey questions about sentiment toward antimicrobial resistance, with comprehension, stated intention to request antibiotics and terminology recall listed as secondary outcomes.

The registry reports arm-level primary outcome percentages and a quasi-binomial regression result, but does not provide the primary estimate’s confidence interval or a fully specified contrast. Secondary comparison recalculations are unadjusted. The record does not report several details needed to evaluate outcome scoring, missing-data handling, multiplicity or the analysis population.