PerfO Administration: Record the Conditions Behind the Score

A performance outcome assessment measures a patient completing a standardized task. The FDA PFDD glossary describes tasks administered and evaluated by an appropriately trained person or completed independently, and notes the importance of patient cooperation and motivation. Examples include timed walking and memory recall. Once a trial has selected a PerfO, the next operational decision is how to preserve enough information about each administration to interpret its score. A distance or completion time needs a traceable connection to the task instructions and conditions that produced it.

Consider this synthetic example, not a reported trial: Site 12 has a 20 m corridor although its study manual specifies 30 m. A covering assessor improvises encouragement during a participant’s first-ever six-minute walk test (6MWT). Other assessments immediately preceded the walk, and the preparation record is incomplete. The immediate response is to record the actual corridor length, script departures, prior test exposure, preceding procedures, rest interval, assistance, interruptions, and unaltered result. Compare those entries with the selected manual before deciding whether a departure occurred.

Course length and encouragement deserve prompt comparability review because the field-walking literature shows that administration changes can affect distance. First exposure raises a learning-effect question; preceding procedures raise a preparation and fatigue question. None supplies a known numerical penalty for this participant. Do not add published mean effects together, adjust the score by a guessed multiplier, or automatically exclude it. Clinical operations and biostatistics need the factual record, the intended comparison, and the trial’s predefined handling rules to assess its use.

The final FDA PFDD Guidance 3, October 2025 addresses how to justify a clinical outcome assessment for its proposed context of use. Appendix D advises conducting a PerfO according to the selected published administrator’s manual. It also identifies practical barriers: required equipment or settings, such as stairs or air-conditioned rooms, may be unavailable or inconsistent across sites. Standardization therefore needs particular attention in multisite and multinational trials if results are to be interpretable and pooled.

Section IV.C identifies possible supporting evidence: a clear and detailed user manual, successful completion of standardized training by personnel at all sites, and intermittent observation during the trial. Appendix D also recommends standardized case report forms, recording assistive-device use and type, and collecting the reason when a test is not completed. These are recommendations within a fit-for-purpose rationale. They do not impose a universal ten-field form, certification scheme, or legal requirement that every PerfO use the same preparation rules.

The 2023 ISPOR PerfO selection report emphasizes standardization, feasibility, safety, and evidence for validity, reliability, usability, and interpretability. Its 2026 implementation companion was available online on 28 September 2026 and remains an in-press report. The public abstract covers documentation, administrator training, data collection, digital technologies, and monitoring, and identifies learning and encouragement as validity risks. It emphasizes prevention of inconsistent implementation. The discussion here uses that public abstract; it does not infer detailed procedures from unreviewed full text.

The 2014 ERS/ATS technical standard provides a concrete example for adults with chronic respiratory disease. Its Table 2 summarizes methodological comparisons, while its learning-effect section reviews repeated tests. The standard estimates a minimal important difference (MID) of about 30 m for six-minute walk distance (6MWD). The 25–33 m range describes the synthesis across studies and methods; it is not a formal confidence interval for one pooled MID. Most estimates were derived from group changes, often in COPD rehabilitation settings.

The instruction result comes from Weir and colleagues’ 2013 study of 24 participants with pulmonary arterial hypertension, idiopathic pulmonary fibrosis, or other interstitial lung disease. Each completed four walks with different instruction wording, presented in random order. This supports treating the script as a measurement condition; it does not establish a universal 52.7 m adjustment for protocol departures. Likewise, the standard distinguishes oxygen provision from oxygen transport. Its transport comparisons support documenting who carried the equipment, without assigning the oxygen-versus-no-oxygen range to that different question.

These findings establish the importance of consistency, not a calculator for correcting nonstandard tests. Different comparisons involve different participants, methods, and contexts. Their mean effects cannot be summed to predict a combined deviation. A difference below the MID is not automatically negligible, and one above it does not automatically invalidate a trial. The 30 m benchmark and these effect sizes should not be transferred to cognitive, sensory, pediatric, or neuromuscular PerfO endpoints without appropriate evidence.

Use the selected instrument manual to decide which conditions must be controlled and which entries the visit form needs. The following field list is an operational synthesis of FDA’s documentation recommendations and the ERS/ATS example. It is not a validated universal instrument or regulator-prescribed template. Link it to the participant, visit, and test attempt so the measured score can be traced to its circumstances. Controlled references can point to the applicable manual and training record rather than duplicating those documents at every visit.

Design the form so it can distinguish “performed as specified,” “departed from specification,” and “not known.” A checkbox that merely says “test completed” cannot tell the reviewer whether the correct task was completed. For a nonstandard walk, retaining lap count and the actual course length permits reconstruction of the observed distance. It does not reconstruct what the participant would have walked on another course. Keep the observed result separate from any derived analysis variable and its documented handling decision.

The field list is deliberately task-specific in its application. A memory task does not inherit an oxygen field or a walking-course requirement. It does need an identifiable task version and whatever assistance, practice, environment, and scoring details its own administration procedure makes relevant. This distinction keeps the form usable: capture the conditions needed to evaluate the chosen measure instead of collecting a long generic checklist whose scientific purpose is unclear.

The ERS/ATS standard’s pooled 26.3 m improvement on a second 6MWT demonstrates why a first attempt can differ from a repeat even without treatment-related change. It is not a penalty to subtract from an individual’s baseline, and improvement on a later visit cannot automatically be attributed to learning. The record needs test number and prior exposure so the intended comparison can be evaluated.

For the 6MWT and incremental shuttle walk test, the standard calls for two tests to establish a stable baseline when detecting change over time. Same-day tests need at least 30 minutes between them, with heart rate and oxygen saturation returned to baseline before the second. For the 6MWT it recommends recording the best distance. Specify the study’s repeat-testing schedule and score derivation in advance, and retain both observed results so the derivation remains traceable. These procedures are tied to the named walking tests, not all physical or cognitive PerfOs.

There are limits to a blanket two-test rule. The same standard says one 6MWT may suffice for one-off staging or risk assessment, and after recent testing when the learning effect may be smaller. It also advises caution near decision thresholds. Thus, test purpose and recent experience matter. For other instruments, consult their own familiarization evidence and manual rather than importing the 6MWT schedule. FDA’s Appendix D similarly asks sponsors to plan for practice effects in the protocol without prescribing one universal number of attempts.

Training records establish preparation; observing a task shows how the procedure is delivered. FDA identifies both as possible support for a fit-for-purpose rationale. ERS/ATS adds a specific quality-assurance pattern: peer review of assessor performance, accurate course measurement, equipment checks, consistent documentation, consideration of regular audit, and programme-director review of procedural changes. The following is a practical way to translate those recommendations into trial operations:

• Before first use, confirm which manual and script each administrator will use and where their training evidence is held. Check replacement personnel against the same task-specific qualification plan.

• During observation, compare the delivered instructions, setup, timing, assistance, and scoring with the selected manual. Record the observed discrepancy rather than inferring poor performance from an unusual score alone.

• When procedures or equipment change, document what changed, when it took effect, who reviewed it, and which assessments used the changed conditions. Evaluate implications for repeat measurements.

• Use recurring discrepancies to target retraining or further observation. Training completion and a complete form do not themselves establish that scores across sites are interchangeable.

This operational focus differs from ClinRO rater training and drift control, where clinical judgment in assigning ratings is central. Here the principal question is whether a standardized task was delivered and quantified consistently. Instrument selection and measurement-property evidence remain separate decisions, addressed in the fit-for-purpose COA guide.

In the synthetic Site 12 example, start with facts that can still be established and identify gaps that cannot. Do not label every discrepancy “major,” presume that a score is unusable, or select a favorable analysis after seeing results. The table below identifies questions for review rather than prescribing statistical exclusions or score corrections.

Shorter tracks and altered encouragement are plausible threats to longitudinal comparability, while exposure history and visit sequence may explain additional differences. A recorded deviation is therefore a review trigger, not an automatic pooling verdict. ERS/ATS notes that differences between tracks longer than 15 m may be small enough for risk stratification, while stressing that more evidence is needed. That limited observation does not establish equivalence for treatment-change endpoints on a 20 m course.

Preserve the observed result, link the deviation record, and apply the trial’s predefined analysis rules with a documented rationale. If those rules do not cover the situation, unresolved comparability should remain explicit; this article supplies no substitute estimator or exclusion rule. A late visit is a related but distinct issue: assessment-window documentation records timing against the protocol, whereas this record explains the task conditions within that visit.

FDA’s equipment and setting examples are a reason to check site capability before the first assessment. Establish whether each location can implement the selected manual, and identify material differences early. A shared score label cannot demonstrate that sites used the same task, preparation, or assistance. Use the same field definitions across sites while keeping the applicable manual version visible.

The ISPOR implementation abstract also covers digital health technologies used to support data collection. As an operational extension of the record, a sensor-assisted standardized task may need device or application version, relevant setup, assistance, and interruptions alongside the task result. These are fields to justify against the particular implementation, not a requirement that software automatically sense every environmental condition. Automation can assist capture but does not establish that the underlying task was comparable.

Keep this discussion tied to actively performed standardized tasks. It does not establish validity for passive wearable measures or prove that an at-home task is equivalent to clinic administration. For that separate evidence question, see digital health technologies and endpoint evidence. The useful deliverable here is narrower: an administration record that makes the observed conditions available for a defensible interpretation of the score.

1. Patient-Focused Drug Development: Selecting, Developing, or Modifying Fit-for-Purpose Clinical Outcome Assessments (final guidance, October 2025)

2. Recommendations on the Implementation of Performance Outcome Assessments in Clinical Trials: A Good Practices Report of an ISPOR Task Force (Bush EN, Edgar CJ, Adams H, et al.; in press)

3. Recommendations on the Selection, Development, and Modification of Performance Outcome (PerfO) Assessments (Edgar CJ, Bush EN, Adams HR, et al.)

4. An official European Respiratory Society/American Thoracic Society technical standard: field walking tests in chronic respiratory disease (Holland AE, Spruit MA, Troosters T, et al.)

5. Patient-Focused Drug Development Glossary (Performance Outcome entry)

6. The Influence of Alternative Instruction on 6-Min Walk Test Distance (Weir NA, Brown AW, Shlobin OA, et al.)