2.4 Trial characteristics

Page last updated: September 2016

Note: Where the submission has included a systematic review containing multiple studies, present the trial characteristics for the individual studies, as detailed in the individual publications or, where these are unavailable, as detailed in the systematic review.

2.4.1   Participants

Provide the following details about the trial participants for each trial in an attachment:

• eligibility criteria for participants considered for recruitment into the trial
• baseline demographic and clinical characteristics for each randomised group or study arm
• median duration (and range) of follow-up for each group and for the entire trial (also indicate whether the trial is ongoing).

Where there are differences between treatment arms (or trials) in terms of the extent or timing of patients lost to follow-up, patient withdrawals, or missed or refused assessments, present (in an attachment) the baseline demographic and clinical characteristics for the following groups (or state where these data are unavailable):

• patients who were lost to follow-up compared with those who were not
• patients who withdrew from the trial compared with those who did not
• patients who missed an assessment compared with those who were assessed (this comparison is critical when the assessment is for the purpose of measuring an outcome that is related to the disease or condition, or the medicine).

In the main body of the submission, present differences in the baseline demographic or clinical characteristics across arms in the trials or across trials. Report whether differences are statistically significant, but note that statistical significance may not always correlate with important differences, particularly for subgroups.

For each of any identified differences, discuss the likely impact on the magnitude and direction of the treatment effect. Any differences across arms or across relevant subgroups, whether in prognostic variables or not, may be an indication that randomisation was unsuccessful and should be noted in Subsection 2.3.

It is important that the information requested in this section is provided for the whole trial population as well as any subgroups (and their complements) presented in Section 2.6. The PBAC is concerned when there are imbalances in important prognostic factors across arms, or between a subgroup and its complement. Where baseline characteristics are unavailable for a subgroup(s), state why and provide any relevant details to reduce the uncertainty related to an imbalance of patient characteristics within the subgroup analysis.

2.4.2   Treatment details

For each trial, provide the intended treatment regimen (for both arms) as outlined in the trial protocol. Include details on dose, method of administration, dose timing and frequency, dose titration and criteria for titration, intended treatment duration, continuation criteria or stopping criteria, and prespecified use of subsequent active therapy following treatment completion or failure. State whether the dose or treatment regimen, including the use of concomitant treatments, is supported by high-quality clinical practice guidelines and by the product information for each of the medicines. Justify where the protocol’s specified dose (or the actual dose in the trial) differs from recommended dosing.

Provide details of how the interventions actually occurred in the trial (across each arm). These details should include an average dose that incorporates the frequency (and/or proportion of participants taking particular doses) and average duration of treatment.

If participants received concomitant treatments for the same indication (such as a background therapy), provide details of these treatments as above.

If participants received active treatments following cessation of the proposed medicine or comparator, provide details on dose and duration of these treatments across the trial arms.

Discuss differences of treatment duration across arms and across trials. Explain any differences observed.

If the submission relies on subgroups, present the information requested in this section for the whole trial population, relevant subgroups and their complement.

2.4.3   Outcomes

Present the following outcomes from each trial:

• the primary outcome specified in the trial protocol
• secondary outcomes that are patient-relevant.

Present a surrogate outcome (that is not the primary outcome) only when it is critical to the therapeutic conclusion or economic evaluation. State the target clinical (patient-relevant) outcome for which it is a surrogate and present a transformation of the surrogate outcome to a patient-relevant outcome as described in Subsection 3A.4.2 (or cross-reference to Subsection 3A.4.2 if the transformation is presented there).

For each outcome:

• state whether it was the primary outcome
• state the units of measurement and the method of statistical analysis
• describe and justify the population in which the analysis is performed (ie intention to treat, per protocol)
• describe the timing of the outcome assessment.

Summarise the power calculations for outcomes for which the trial was designed to detect a change, and state how missing data were dealt with.

When describing the method of statistical analysis, include the name of the statistical test and sufficient details to allow the PBAC to ascertain how the analysis was performed. For analyses that are not included in the clinical study report, provide a statistical appendix – including the statistical code and statistical output – with notation explaining the variables used in the analysis. Describe the analysis set (eg total randomised population or described as per protocol subset), and the extent of missing data and how missing data were handled (eg censored, imputed). Comment on the likely effect of missing data on the estimate of the treatment effect. Clearly describe the assumptions for the approach to dealing with missing data. Where missing data has been discussed elsewhere, cross-reference the appropriate subsection.

Where there are multiple trials, clearly present any differences in the definition of outcomes or the method of statistical analysis. An example of how outcomes may be presented is shown in Table 2.4.1.

Describe how each outcome was measured, including:

• the instrument used to measure the outcome (eg questionnaire, criteria such as RECIST,⁷ blood test)
• threshold for categorisation as an outcome (if applicable)
• timing of the measurement of the outcome
• personnel who administered the instrument (eg investigator, study nurse, patient)
• personnel who determined whether the outcome had been achieved (or the magnitude of the outcome).

For each instrument, state whether the instrument is validated in the population and the circumstances in which it is applied in the study, and reference its validation.

Ensure that each outcome is reported as being truly independent, or that the statistical analysis appropriately adjusts for clustering. This issue most often occurs when a single patient can experience multiple events (eg fractures, hypoglycaemic events, hospitalisation episodes) during follow-up.

Where the submission has identified multiple trials, clearly indicate how many trials reported on each relevant outcome. If some trials have not reported on relevant outcomes, indicate this in a footnote when presenting results in Subsection 2.5 or 2.6.

Composite outcomes

A composite outcome is one in which multiple endpoints are combined. It is usually defined as having been experienced when the first of any of the component endpoints is experienced, even though subsequent component endpoints may occur.

If one or more of the reported outcomes is a composite, discuss and compare the clinical importance of each of the components of the composite. Report whether the definition of the composite outcome was explicitly prespecified. Justify the inclusion of the components in the composite outcome, and the exclusion of any components that were considered but subsequently rejected. Disaggregate the composite outcome and present the results (eg comparative rates) of each component as a secondary outcome in Subsection 2.5.

Composite outcomes need to be appropriately handled when disaggregating the component outcomes so that the true estimate for each component outcome is appropriately captured.

Patient-reported outcome measures

Patient-reported outcome measures include generic (‘global’) or condition-specific (eg for respiratory conditions, depression, arthritis) measures of quality of life, symptoms or function.

Patient-reported outcome measures may also include multiattribute utility instruments (MAUIs), in which the scoring method for the instrument is anchored on a quality-adjusted life year scale of 0 (death) to 1 (full health). Several commonly used MAUIs for which a detailed discussion of the validity or reliability is not required are the Health Utilities Index (HUI2 or HUI3), the EQ5D-3L or -5L (‘EuroQol’), the SF-6D (a subset of the Short Form 36, or SF-36), the Assessment of Quality of Life (AQoL) instruments, and the Child Health Utility 9D (CHU9D) index for children and adolescents.

Include any data and references that support the selection of the MAUI in a technical document or an attachment (provide clear cross-references between these data and the main body of the submission).

Where a patient-reported outcome measure is used, or a MAUI that is not listed above, provide, in an attachment, a discussion of (or reference supporting) the:

• domains of quality of life, symptoms or function that are covered by the instrument
• scoring method of the instrument
• validity of the instrument
• reliability of the instrument
• responsiveness of the instrument to differences in health states between individuals and to changes in health states over time experienced by an individual
• clinical importance of any differences detected by the instrument (see Subsection 2.4.4 for guidance on minimal clinically important differences [MCIDs]).

To explain how the patient-reported outcome measure is used within the study, describe:

• the timing of assessments, including how often and at what points in the study the instruments were administered
• who administered the questionnaire and in what setting
• why assessments were missed and how missed assessments were dealt with.

Provide the characteristics of the patients who missed or refused to complete patient-reported outcome measures and compare them with those patients who completed the questionnaires. If this has been presented in Subsection 2.3, cross-reference it. If an investigator assessment of patient wellbeing (or performance score) is captured for all patients at all time points, this may be an appropriate metric to compare patients who completed the questionnaire with those who did not. Describe any methods that were used to adjust for response bias, or describe the effect of missed assessments on the comparison of patient-reported outcome measures across the arms of the study.

2.4.4   Minimal clinically important difference

An MCID is the smallest difference in a particular outcome that patients perceive as beneficial (or detrimental). This is usually determined by patients, although an MCID may be determined by a consensus of experts. An MCID should be specified for the primary outcome and the main patient-relevant outcome (where this is not the primary outcome). For submissions relying on a claim of noninferiority for which a noninferiority margin is specified, present an MCID only if it informs the noninferiority margin.

Likely sources for an MCID may be:

• the protocol (often for the purposes of powering the study)
• a previously accepted MCID by the PBAC, as indicated in a public summary document, that is relevant to both the trial population and the proposed indication
• a commonly accepted MCID in the literature, relevant to the trial population and the proposed indication
• an internal study by the sponsor (anchor-based analysis, expert consensus, statistically based analysis)
• a commonly accepted MCID in the literature for a similar indication that can reasonably be expected to be generalisable to the proposed indication.

Present the details for selected MCIDs in Table 2.4.2. Regardless of the method of derivation, describe the influence of consumer or patient input, where possible.

Where data are time to event (eg overall survival) or dichotomous (eg haemorrhage or no haemorrhage), the determination of an MCID is not straightforward. The most common approach for determining a meaningful benefit to patients involves a consensus of clinical experts in the relevant fields, and should account for the values of patients and families. An example of an expert process for determining a meaningful benefit in terms of overall survival to patients with selected cancers was published in 2014.⁸ Where the primary outcome is a surrogate for another endpoint (eg cholesterol for cardiovascular events), the justification of an MCID should be the change in the surrogate required to result in a meaningful change in the target outcome.

Patient-reported outcomes or patient-relevant continuous/ordinal outcomes

For patient-relevant outcomes that are measured on a scale (eg a patient-reported outcome measure, a quality-of-life instrument, the Visual Analogue Scale, the LogMAR vision acuity test, the 6-Minute Walk Distance test), the MCID may be established using an anchor approach.^9,10 Although alternative approaches (statistical or consensus) are available, they are less preferred. For these types of outcomes, the MCID can be used as a threshold, beyond which an individual patient would be regarded as a ‘responder’.¹¹

a Methods for deriving an MCID commonly fall within three categories.^12-15

2.4.5   Noninferiority margin

A claim of noninferiority means that, in terms of safety and effectiveness, the proposed medicine is no worse than the main comparator. However, a lack of a statistically significant difference between the proposed medicine and the comparator does not adequately establish noninferiority. It is common practice to require that the confidence limits of the difference in treatment effect do not include an a priori stated clinically meaningful difference favouring the comparator.

Choice of patient-relevant outcome(s)

Establish a noninferiority margin for the primary outcome and the most important patient-relevant outcome (where this is not the primary outcome). Where the proposed medicine impacts on two distinct indications, or contains two active components (see Product type 1) that affect two distinct indications, establish noninferiority for the primary and most patient-relevant outcomes for both indications.

Justification of the noninferiority margin

Select a noninferiority margin to assure that the proposed medicine is not inferior to the main comparator by an important difference.

Propose a magnitude of difference in outcome that would be regarded as unimportant and can be used as the noninferiority margin. Justify the approach taken to establish the noninferiority margin, noting that a statistical approach by itself is inadequate,¹⁶ and indicate whether there is agreement across multiple sources. It is common to estimate an unimportant difference as less than a minimal clinically important difference (Section 2.4.4).

Prespecified noninferiority margin

Where the included trial has prespecified a noninferiority margin, present and justify the choice of the margin. Explain how the noninferiority margin meets the assurance previously described. Reference the justification presented in the trial protocol. Where the justification provided in the protocol does not adequately address the assurance previously described, provide supporting evidence. Some noninferiority trials are designed to ensure that the proposed medicine retains superiority over placebo. However, a noninferiority margin designed to achieve this may still allow an important reduction in treatment effect compared with the main comparator. In this case, redefine the margin and retest noninferiority.

Non-prespecified noninferiority margin

Where there is no prospectively defined noninferiority margin, justification of such a margin after trial completion (ie post hoc) is difficult¹⁶ and not preferred by the PBAC. Therefore, choose a conservative margin for the submission. This may happen where there are, for example:

• failed superiority trials of the proposed medicine versus comparator¹⁷
• indirect comparisons of the proposed medicine versus comparator via a common reference
• outcomes in noninferiority trials that did not have a prespecified noninferiority margin.

When selecting post hoc noninferiority margins, where possible, present multiple sources of evidence for selecting a margin that represents an unimportant loss of treatment effect, and that converge on a similar estimate. Present and discuss the list of estimates, their sources and the methods used to derive the estimates. Justify the selection of one particular estimate as the proposed noninferiority margin.

2.4.6   Cross-references to source documents

For each trial, specify the source document in the reports or papers accompanying the main body of the submission. For each of the responses, cross-reference to the page, table or figure numbers of the relevant trial report(s) (in a separate technical document or attachment, if necessary).