Improving Communication in Older Cancer Patients and Their Caregivers: Complete Statistical Analysis

1. Trial at a Glance

This page covers the registry results for Improving Communication in Older Cancer Patients and Their Caregivers (NCT02107443), a randomized na study. The primary outcome is Direct Communication About Age-related Concerns: Number of Discussions Related to the Geriatric Assessment That Occur in the Clinic Visit Between the Patient, Oncology Physician, and Caregiver. [Patient-Centered Outcomes Research Institute Specified].

2. Clinical Question

The trial rationale was that age-related health concerns, including function and cognition, may matter to treatment decisions and quality of life for older people with cancer. The study therefore evaluated communication about such concerns, as well as patient-reported quality of life and satisfaction-related outcomes. The central statistical contrast is not simply whether a geriatric assessment was completed: it distinguishes assessment alone from assessment accompanied by a summary and recommendations for patients, caregivers, and the oncology team. The registry describes the study as cluster randomized, so interpretation of the result should account for the design rather than treating every patient as an independently randomized unit.

3. Trial Design

The registry describes a randomized, single-masked, parallel, cluster-randomized clinical trial. It lists enrollment as 546 and status as completed. The listed study dates are 2014-10-27 for the start and 2017-05-12 for primary completion. The phase is listed as NA. The precise unit of randomization and the number of clusters are not reported in the registry record.

In the geriatric-assessment intervention arm, patients and participating caregivers completed the assessment at their first study visit with the oncologist and received a summary with targeted recommendations. The oncology team received the information to discuss and implement if it chose. In usual care, patients and participating caregivers completed the assessment at the first visit, without receiving the summary or recommendations. This distinction is important: the registry does not describe the comparison as assessment versus no assessment.

The primary endpoint is the number of conversations related to the geriatric assessment during the clinic visit among the patient, oncology physician, and caregiver. The listed timeframe is “Baseline.” The record does not clarify how that label relates to the intervention being delivered at the first study visit, so the timing should be interpreted as recorded rather than assumed to represent a later follow-up visit.

The registry describes a linear mixed model for the primary outcome, with arm as a fixed effect, estimation by restricted maximum likelihood, and an F test of the null hypothesis of zero mean difference. It also says that differences among NCORP practice sites would be assessed graphically using best linear unbiased predictors of mean response for each site. Further model details, including the precise covariance structure and the handling of cluster-level allocation in the primary model, are not reported in the registry record. The registry reports five outcome measures and five analyses, but a detailed analysis sequence is not 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

Reasons for not completing this period:

Analysis population for the primary outcome, as defined in the registry: All baseline patients who were evaluable for this primary aim were included in this analysis. * Arm I patients excluded because 2 withdrew, 1 expired, 4 no audio captured and 2 primary aim protocol violations. * Arm II patients excluded (3 withdrew, 1 no audio captured).

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: * Non-Evaluable Patients: Arm I- 3 Patients Study Protocol Violation; Arm II- 1 Patient Study Protocol Violation, 1 Patient Registered then found to be ineligible. * Missing Demographics information from 1 evaluable patient from Arm I. This is denoted as Unknown.

6. Primary Endpoint

The registry lists the primary outcome as Direct Communication About Age-related Concerns: Number of Discussions Related to the Geriatric Assessment That Occur in the Clinic Visit Between the Patient, Oncology Physician, and Caregiver. [Patient-Centered Outcomes Research Institute Specified], measured over Baseline.

The primary endpoint counts conversations about age-related concerns during the clinic visit between the patient, oncology physician, and caregiver. The geriatric assessment defined which age-related topics would count as relevant discussions. Examples of concerns named in the study description include function and cognition. The endpoint is therefore a communication measure: it records discussion of concerns identified through the assessment rather than a clinical event or a direct measure of treatment efficacy.

The registry identifies the outcome as a number of conversations and describes the estimand as a least-squares mean. It specifies the timeframe as “Baseline.” The primary analysis population is described as all baseline patients evaluable for the primary aim. The reported analysis gives a mean difference of 3.59 and p < 0.001, while the numerical limits of the stated 95% confidence interval are not reported in the registry record.

Counting relevant discussions can help assess whether assessment information is being brought into the clinic conversation. It does not, by itself, establish the accuracy, quality, completeness, or usefulness of those discussions. Nor does the endpoint alone establish that a conversation changed a treatment decision or a patient's quality of life. Those are separate questions, reflected in the study's other stated outcomes. Because the outcome is a count, interpretation also depends on how discussions were identified and coded; the registry says that topics were defined using the geriatric assessment but does not report further details about coding reliability or the possible range of the count.

7. Why No Re-analysis Is Shown

8. Statistical Interpretation

9. Statistical Methods Explained

The primary endpoint is a continuous count, and the registry describes its analysis using a linear mixed model. In general, a mixed model estimates an outcome association while allowing for structured variation among related observations or groups. Such models can be useful when measurements are not fully independent, as can occur when allocation is by cluster or when outcomes vary across practice sites. The precise random-effects structure used for this trial is not reported in the registry record, so the model should not be described more specifically than the registry does.

What does the mean difference describe?

A mean difference expresses the contrast between the arms on the outcome's original scale. Here, the scale is the number of conversations. The reported estimate is 3.59, and the primary outcome is described in terms of least-squares means. In general, a least-squares mean is a model-based mean estimate, conditional on the model's specification; it need not be identical to a raw arithmetic mean. The results entry labels the estimate “Mean Difference (Final Values),” while the primary description gives the timeframe as “Baseline.” The record does not explain this difference in labels. It also does not identify which result code corresponds to which named arm, so the sign or direction should not be translated into a claim that one particular arm had more conversations.

Why use a mixed model in a cluster-randomized trial?

In general, cluster randomization assigns groups rather than treating every participant as independently assigned. People in the same cluster may have shared clinicians, clinic procedures, or other common influences, and their outcomes can therefore be correlated. A model that accounts for this dependence can yield more appropriate uncertainty estimates than a calculation that assumes complete independence. The registry identifies the study as cluster randomized and the primary analysis as a mixed model. It also says that specific NCORP practice-site differences would be assessed graphically using best linear unbiased predictors of the site mean response. The number and identity of clusters, the exact level of allocation, the model's full variance structure, and the way site effects entered the primary inferential model are not reported in the registry record.

As a general matter, over-dispersion occurs when variability in count outcomes exceeds what a simple count model would expect. A linear mixed model is the method named in the registry, and no over-dispersion assessment or alternative count-model analysis is reported. The outcome's distributional diagnostics and any sensitivity analyses are also not reported, so they should not be inferred from the selected model alone.

What do restricted maximum likelihood and the F test do?

The registry states that estimation used restricted maximum likelihood. In general, restricted maximum likelihood estimates variance components while accounting for the estimation of fixed effects, and is commonly used in mixed-model settings. The registry says arm was the fixed effect. It then specifies an F test for the null hypothesis of zero mean difference between arms. In general, an F test evaluates whether the observed model contrast is sufficiently inconsistent with that null hypothesis under the model's assumptions. The reported p-value is p < 0.001. The degrees of freedom and the exact reference distribution are not reported in the registry record.

A p-value is not an effect-size measure and does not state the probability that the observed result arose by chance in an everyday sense. It quantifies how unusual data at least as incompatible with the null as the observed data would be, under the null hypothesis and the analysis assumptions. The result's interpretation should therefore consider the estimated difference, its uncertainty, the endpoint's meaning, and the design, rather than relying on the p-value alone.

How should the confidence interval be read?

The primary outcome description identifies a 95% confidence interval as the dispersion measure. In general, a confidence interval gives a range of estimates compatible with the data and the method at the stated confidence level, under the model assumptions. The posted analysis does not include numerical lower or upper limits in the registry record. Consequently, the interval's width, whether it includes zero, and the plausible magnitude of the effect cannot be stated from the posted numerical result. The p-value is reported, but it does not recover the missing interval limits without assumptions and additional information, and no such reconstruction is appropriate.

What does the site assessment add?

The registry says that differences among specific NCORP practice sites would be assessed graphically using best linear unbiased predictors of mean response for each site. In general, these predictors estimate group-specific deviations while borrowing information across groups according to a model. Graphical assessment can help reveal whether outcome patterns vary by site, but it is not a substitute for a reported numerical estimate of the arm contrast or its uncertainty. No site-specific estimates or graphic interpretation are reported in the registry record.

Finally, the trial's cluster-randomized design is why a simple individual-level recalculation is not appropriate. Recalculations that ignore clustering can give misleading precision. The sponsor's mixed-model analysis is therefore the relevant reported inferential result; the available record does not permit a complete independent reconstruction of its assumptions or calculations.

10. Analyses Reported by the Sponsor

The registry reports the sponsor's mixed-model result: a mean difference of 3.59 and p < 0.001. No independent recalculation is shown because the trial was cluster randomized and a simple individual-level reanalysis would be misleading. This is not a case in which an unadjusted estimate should be substituted for the sponsor's analysis.

The primary outcome description identifies a 95% confidence interval, but numerical limits are not reported in the registry record. The result entry also uses coded groups, OG000 and OG001, without mapping those codes to the intervention and usual-care labels. The estimate can therefore be reported as a between-arm difference, but its direction cannot be assigned to a named arm using the posted result alone.

The registry describes arm as a fixed effect in a linear mixed model, with restricted maximum likelihood estimation and an F test of zero mean difference. It also describes site-level assessment using best linear unbiased predictors. The record does not give the complete model specification or the cluster-level details needed to reproduce the analysis. The sponsor's result should be read as the reported cluster-aware analysis, not as proof that every modeling decision can be independently checked from the available record.

11. Multiplicity, Missing Data and Analysis Populations

The registry lists five outcome measures and five analyses. It does not report a multiplicity adjustment, a hierarchy of outcomes, a prespecified testing sequence, or whether the reported p-value was adjusted for examining multiple outcomes. This matters because, in general, testing multiple hypotheses can increase the chance of observing at least one small p-value even when the hypotheses are all null. Without a stated strategy, the reported p-value should be interpreted as the result for the primary comparison under its specified test, not as an adjustment across all outcomes.

The primary analysis population is described as all baseline patients evaluable for the primary aim. The registry lists exclusions in Arm I: 2 withdrew, 1 expired, 4 had no audio captured, and 2 had primary-aim protocol violations. In Arm II, it lists 3 withdrawals and 1 with no audio captured. These are exclusions from the primary analysis population as described; the record does not state that the analysis included every randomized participant.

The registry does not report how missing outcome measurements were handled beyond identifying these exclusions. It does not describe imputation, assumptions about missingness, sensitivity analyses, or whether the exclusions were determined before outcome data were reviewed. The reasons listed suggest that not every exclusion had the same cause, but the record does not quantify how those reasons affected the estimated contrast. No assumptions about missing-at-random or missing-not-at-random mechanisms should be made from the reported information.

It is also not reported in the registry record whether the analysis followed an intention-to-treat principle, how protocol violations were defined, or whether additional analysis populations were used. Since the trial was cluster randomized, the relationship between cluster allocation, individual evaluability, and the analysis population is important; further details are not reported. These gaps limit assessment of potential selection effects and of how robust the result is to alternative missing-data decisions.

12. 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's safety entries report zero serious adverse events among 296 at risk in Arm I (GA Informational Intervention), and zero among 250 at risk in Arm II (Usual Care). It likewise reports zero other adverse events among 296 at risk in Arm I and zero among 250 at risk in Arm II. The registry lists 42 deaths for Arm I and 30 for Arm II.

These entries should be read as counts recorded in the registry, not as evidence that the intervention caused or prevented any event. The deaths are listed separately from serious and other adverse-event counts; the record does not establish their causes or relationship to study participation or treatment. The observation period and the rules used to collect and classify these safety outcomes are not reported in the registry record. The registry also lists no top serious-event terms. These figures alone therefore do not permit a detailed comparison of safety, event timing, severity patterns, or causation between arms.

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

The primary result describes a difference between the trial arms in a communication count: the registry reports a mean difference of 3.59 and p < 0.001. It supports the conclusion that the sponsor's specified mixed-model test found evidence of a difference in the number of age-related conversations under that analysis. Because the result codes are not mapped to the named arms in the registry entry, the direction should not be attributed to the geriatric-assessment intervention or usual care from this result alone.

The result does not establish the size of the difference with precision because numerical confidence limits are not reported in the registry record. Nor does a small p-value establish that the difference is large, important to patients, or sustained over time. Clinical or practical importance would require context about the scale and distribution of the count, the content and quality of conversations, and how the difference relates to patient and caregiver priorities. Those details are not reported in the primary result.

The endpoint is communication about concerns identified by a geriatric assessment. It is not itself a measure of quality of life, caregiver satisfaction, treatment tolerance, or treatment efficacy. The registry describes those as additional areas of interest, but numerical findings for those outcomes are not reported in the registry record for this interpretation. No conclusions about those outcomes should be drawn from the communication result.

The analysis also applies to the evaluable baseline patients described in the registry, after the listed exclusions. The relationship between the “Baseline” timeframe and delivery of the intervention at the first study visit is not clarified. Finally, the cluster-randomized design means the sponsor's mixed-model analysis is central to interpreting uncertainty; a simple individual-level analysis would not respect the design. Taken together, the result is informative about the registered communication endpoint under the sponsor's analysis, while its direction, precision, broader outcome implications, and sensitivity to exclusions remain incompletely described.

14. Why This Trial Matters Statistically

This trial raises an important statistical issue: a patient-level outcome can arise within a cluster-randomized design. When people are grouped by shared clinical settings or other allocation units, their responses may not be independent. General statistical methods must account for that dependence to estimate uncertainty appropriately. The registry's explicit warning against a simple individual-level reanalysis is therefore consequential; a numerically straightforward comparison would not necessarily provide a valid measure of precision.

The endpoint also illustrates the distinction between a process measure and a health outcome. Counting conversations captures whether age-related concerns enter the clinic discussion. It does not directly measure whether the information was accurate, whether decisions changed, or whether patients experienced better quality of life. Such outcomes can be related, but they are not interchangeable. Clear separation of these questions helps prevent a communication finding from being overinterpreted as evidence about downstream clinical effects.

The reporting also highlights why an effect estimate and its uncertainty should be presented together. The registry gives a mean difference and a p-value, but not numerical confidence limits. The estimate identifies the reported contrast, while the missing interval limits the ability to judge its precision. The unassigned group codes additionally restrict interpretation of direction. These are reporting limitations even though the analysis method and test are described.

Finally, the evaluable analysis population, listed exclusions, multiple outcomes, and lack of reported multiplicity or missing-data procedures matter to interpretation. They do not invalidate the result, but they identify aspects that readers need in order to assess how broadly and robustly it applies. The study is statistically notable for combining a communication endpoint with cluster randomization and a mixed-model analysis, while also illustrating how incomplete reporting of model detail and uncertainty can narrow what can be concluded.

15. 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 assessment are not reported in the registry record. The exact unit of cluster randomization, number of clusters, allocation procedures, and complete mixed-model specification are absent. The record mentions site-level assessment but does not provide site-specific estimates or enough detail to determine how site structure contributed to the primary contrast. These omissions make independent reproduction of the sponsor's analysis impractical.

The primary outcome timeframe is recorded as “Baseline,” although the intervention is described as occurring at the first study visit. The registry does not explain how these timing descriptions fit together. The numerical lower and upper limits of the stated 95% confidence interval are also absent, and the results entry's group codes are not linked to the named arms. These gaps constrain interpretation of timing, precision, and direction.

The analysis includes only baseline patients evaluable for the primary aim, with exclusions reported for withdrawal, expiration, missing audio, and protocol violations. The record does not describe the handling of missing data beyond these exclusions, provide sensitivity analyses, or explain whether the analysis was intention-to-treat. Potential effects of exclusion on the estimated contrast therefore cannot be assessed from the registry entry.

The registry lists five outcomes and five analyses but does not report a multiplicity plan or adjustment. It also does not give further information about coding procedures, coding reliability, the outcome's range, or distributional checks. Such details would help assess the measurement and analysis of a conversation count. Finally, the registry entries do not include numerical results for the stated secondary outcomes, so conclusions about quality of life or satisfaction cannot be evaluated alongside the primary result.

16. Related Tutorials

17. Related Statistical Calculators

18. Sources

• ClinicalTrials.gov: Improving Communication in Older Cancer Patients and Their Caregivers — NCT02107443.
• Registry record last updated 2024-04-12; status last verified 2024-04.
• Results first posted 2018-08-28.
• 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.

19. Record Summary

The registry record describes a completed, randomized, single-masked, parallel, cluster-randomized trial with enrollment of 546. It compared usual care, in which patients and participating caregivers completed a geriatric assessment, with an intervention that added a summary and targeted recommendations for discussion by the oncology team.

The primary outcome was the number of conversations about geriatric-assessment-related concerns during the clinic visit. The registry reports a mixed-model mean difference of 3.59 and p < 0.001. Numerical confidence limits and the mapping of result group codes to named arms are not reported in the registry record. The primary analysis included evaluable baseline patients, with exclusions listed in both arms.

The cluster-randomized design makes the sponsor's mixed-model analysis the relevant reported result; an unadjusted individual-level recalculation would be misleading. Details about the complete model, missing-data handling, multiplicity, and numerical secondary-outcome results are not reported in the registry record.