A patient attends a visit, but the laboratory result never reaches the database. Another completes the joint assessment but misses the symptom questionnaire. Both visits happened; neither necessarily provides the complete endpoint.
Missing data in rheumatology trials can weaken evidence while creating avoidable work for CROs. In a 2021 review of rheumatoid arthritis drug trials, nearly one-third of trials with assessable missingness had over 20% missing primary outcome data. The review examined trials published in selected years between 2002 and 2016, not today’s trial population.
The practical response starts before analysis: make expected assessments visible, detect gaps early and understand why information is absent.
Why rheumatology endpoints are vulnerable
What are missing data in clinical trials? They are measurements needed for the planned analysis that were not obtained or are unavailable. They can involve an entire visit, one questionnaire item or a component of a composite endpoint.
Composite scores depend on complete inputs
The four-variable DAS28-ESR calculation combines tender and swollen joint counts, erythrocyte sedimentation rate and a patient assessment. A missing required component can prevent calculation of the protocol-specified score. Validated alternative formulas exist, but switching formulas opportunistically changes the measurement.
Imaging and PROs introduce different gaps
Radiographs may be unavailable, unreadable or unsuitable for the planned comparison. Specify acquisition, quality review and reading procedures early.
Patient-reported outcomes (PROs) add timing and accessibility challenges. Pain, fatigue, difficult interfaces and unclear instructions can hinder completion. In multilingual studies, validated wording and usable local-language support matter alongside translation.
MCAR, MAR and MNAR explain different assumptions
A blank field does not reveal its statistical mechanism. Use reasons, timing and observed patient information to investigate the pattern.
- MCAR, missing completely at random: missingness is independent of observed and unobserved data. A random technical failure could fit this assumption; systematic failures affecting particular devices or sites may not.
- MAR, missing at random: after conditioning on observed information, missingness does not depend on the unobserved value. Age-related completion differences could be compatible with MAR if relevant observed factors are adequately captured and modelled.
- MNAR, missing not at random: missingness still depends on unobserved outcomes after accounting for observed information. Patients avoiding assessments during an unrecorded flare illustrate a plausible concern.
CTA: Withdrawal for inefficacy or adverse events needs investigation, but does not establish MNAR. Observed data generally cannot distinguish MAR from MNAR.
Why missing outcomes threaten trial interpretation
Incomplete endpoints reduce information and can lower statistical power. If missingness relates to treatment response, estimates may also be biased.
Excluding participants with incomplete records can undermine an intention-to-treat approach. Missing data themselves do not automatically determine whether an analysis respects ITT; the analysis population and handling strategy matter.
Last observation carried forward assumes an earlier measurement can stand in for a later one. Non-responder imputation classifies specified missing binary outcomes as treatment failures. Neither is universally appropriate or automatically conservative.
The EMA guideline on missing data emphasises prevention and prespecified handling. ICH E9(R1), adopted by FDA, connects analysis to the intended treatment effect and sensitivity analyses. CONSORT 2025 supports transparent reporting; it is not a universal imputation algorithm.
How to prevent missing data during study conduct
Configure checks around required endpoint components, expected laboratory results and visit schedules. Distinguish data not yet received from assessments not performed. Give sites actionable queries and clear escalation routes.
Allow a documented reason when an assessment cannot be completed. Forcing entries can replace missing values with unreliable data.
Make patient completion practical
Use protocol-defined questionnaire windows, appropriately timed reminders and accessible interfaces. Test the workflow with representative users, including people experiencing hand pain or fatigue.
ACTide’s article on multilingual ePROs with short time windows explores these operational design questions.
Review patterns while intervention is still possible
Track expected versus received assessments by endpoint and visit, then examine site, language and country patterns. Separate outstanding transfers from genuine missing outcomes. Assign owners and review whether corrective actions improve completion.
A useful alert identifies the missing component, its due date and who can resolve it. For example, an absent ESR result needs a laboratory follow-up, while a missed questionnaire window needs a review of patient support and protocol rules. A single overall completion percentage can conceal both problems.
What an integrated eClinical ecosystem can change
ACTide ePRO supports multilingual patient interfaces, on-device notifications and storage directly in the eCRF database. Connecting patient reporting with ACTide eCRF can reduce manual reconciliation.
ACTide Data Manager Tools & Services supports data controls, completion monitoring and data integration. Laboratory feeds still need study-specific mapping and validation.
Bring your endpoint dependencies, questionnaire windows and escalation rules to an ACTide demo to assess the configuration your study needs.
Prevent gaps before database lock
Technology can reduce avoidable missingness, but cannot guarantee complete endpoints or remove statistical uncertainty. Combine patient support, site processes and an agreed analysis strategy.
Request an ACTide demo to discuss how your CRO can detect and manage rheumatology data gaps earlier.