Example

 

“The primary endpoint is the difference in mean change of GELP [Genital Erosive Lichen Planus] scores from baseline (week 8) to week 32 between deucravacitinib and methotrexate treatment groups.

 

The secondary endpoints will explore mean changes in:

• Vulvar Quality of Life Index (VQLI) [Reference] at weeks 8, 24 and 32

…

• General Health Questionnaire-28 (GHQ-28) [Reference] at weeks 8, 24 and 32

• Physician Global Assessment (PGA) [Reference] at weeks 8, 24 and 32

• Patient Global Assessment (PtGA) [Reference] at weeks 8, 24 and 32

..." [248]. ​

​

“We will use two primary outcome measures:

 

(1) The patient’s global perceived effect (GPE) at 3 months after start of treatment measured on a 7-point Likert scale. The GPE scale will be dichotomized as “improved” (scores 1–2) or “unchanged/worse” (scores 3–7). GPE is recommended as a core outcome measure in pain studies, as it may cover additional aspects to pain relief and physical function that is important to the individual [Reference].

 

(2) The proportion with a clinically important improvement at 3 months in function measured by the Patient-Specific Function Scale (PSFS; 0–10). An important improvement will be defined as 30% increase on PSFS. The PSFS will also be dichotomized. Percent changes in PSFS scores will be calculated by taking the actual change in score divided by the possible change, to account for baseline values" [249]. 

​

Explanation

 

Trial outcomes are fundamental to study design and interpretation of results. For a given intervention, an outcome can generally reflect benefit (efficacy) or harm (adverse effect). The outcome of main interest is designated as the primary outcome, which usually appears in the objectives (Item 10) and is the basis of the sample size calculation (Item 19). The remaining outcomes constitute secondary or other exploratory outcomes. The inclusion of validated patient-reported outcomes is encouraged to reflect patient perspectives on their health status [23]. 

​

It is recommended that one outcome is designated as primary. Although up to 38% of trials defined multiple primary outcomes, [250-253] this practice can introduce problems with multiplicity of analyses, selective reporting, and interpretation when there are inconsistent results across outcomes [254]. Problems also arise when trial protocols do not designate any primary outcome, as seen in half of protocols for a sample of trials published from 2002–2008, [255] and in 18% and 25% of randomised trial protocols that received ethics approval in Switzerland and Denmark respectively [250, 256]. Furthermore, major discrepancies in the primary outcomes designated in protocols, registries, and regulatory submissions versus final trial publications are common. The discrepancies favour the reporting of statistically significant primary outcomes over non-significant ones, and are often not acknowledged in final publications [64, 257-260] Such bias can only be identified and deterred if trial outcomes are clearly defined a priori in the protocol and if protocol information is made public [78, 261-263]. 

​

The systematic development and adoption of a common set of key trial outcomes for a given health condition can help to deter selective reporting of outcomes and facilitate comparisons and pooling of results across trials in a meta-analysis, [264, 265] helping to reduce research waste. The COMET (Core Outcome Measures in Effectiveness Trials) Initiative aims to facilitate the development and application of such standardised sets of core outcomes for clinical trials of specific conditions [266]. 

​

Reviews of two samples of 108 and 292 trial protocols from 2016 found that the specific measurement variable was reported in 89% to 97% of protocols, while fewer specified the analysis metric (83% to 90%) and time point of main interest (82% to 94%).(9, 10) Similar findings were observed for a sample of protocols of published oncology trials [267]. 

​

For each outcome, the trial protocol should define four components: the specific measurement variable, which corresponds to the data collected directly from trial participants (e.g., Beck Depression Scale, all-cause mortality); the participant-level analysis metric, which corresponds to the format of the outcome data that will be used from each trial participant for analysis (e.g., change from baseline, final value, time-to-event); the method of aggregation, which refers to the summary measure format for each study group (e.g., mean, proportion of participants with score > 2); and the measurement time point of interest for analysis [252, 262]. 

​

It is also important to explain the rationale for the choice of trial outcomes, e.g., use of a surrogate outcome [268]. An ideal outcome is valid, reproducible, relevant to the target population (e.g., patients), and responsive to changes in the health condition being studied [206]. The use of a dichotomous analysis metric reduces statistical power, compared with a continuous one; [269-271] and subjective outcomes are more prone to bias from inadequate blinding (ascertainment bias) and allocation concealment (selection bias) than objective outcomes [272]. 

​

 Individual components of any composite outcome should be clearly defined [22]. Although composite outcomes increase event rates and statistical power, their relevance and interpretation can be unclear if the individual component outcomes vary greatly in event rates, direction of effect, importance to patients, or amount of missing data.

 

Summary of key elements to address

• Specification of which outcomes are primary and secondary

• Rationale for the choice of trial outcomes and whether they are part of a core outcome set

• For each outcome:

  • Specific variable to be measured (e.g., Beck Depression Inventory score, all-cause mortality), with definition where relevant

  • Analysis metric for each participant (e.g., change from baseline, end value, time-to-event)

  • Summary measure for each study group (e.g., mean, proportion with score > 2)

  • Time point of interest for analysis (e.g., 3 months)