Selecting an outcome score is not the same as designing an endpoint strategy. A defensible arthroplasty study starts from the intended purpose and clinical claims, then defines the question each endpoint must answer. Instrument, timepoint, responder definition and statistical method all follow from that question.
CMO · Orthopedics & Spine
NHS orthopedic surgeon
in orthopedic surgery & Class III implant evidence
Chief Operating Officer · former TÜV SÜD Senior Reviewer · Senior Consultant Surgeon
in device evaluation & Notified Body review
Chief Data Officer · Biometrics & Data Systems
in clinical data management, EDC & biometrics
This page is relevant when you are writing a first-in-human or pre-market arthroplasty protocol, or when an existing design needs strengthening. The situations below are the ones we see most often.
The protocol contains several outcomes but none is identified as the main result.
The selected instrument does not directly support the intended clinical benefit.
An authority, ethics committee or Notified Body has challenged the endpoint structure.
Radiographic and patient-reported results are not aligned or not reconciled.
A threshold was selected after the data were collected rather than prospectively.
Foreseeable events were not planned for in the protocol or SAP.
If existing clinical data now need to be carried into a CER or a PMCF strategy, start at the arthroplasty evidence hub and come back here for the endpoint detail.
Much of the confusion in arthroplasty protocols comes from treating these as one idea. Separating them is the first step toward an endpoint that survives review.
What is being assessed: pain, function, fixation, stability, revision or quality of life.
How the domain is captured: HOOS, OKS, KOOS, ASES, radiographic assessment or revision records.
The prespecified result used to answer the clinical question, for example change from baseline in OKS at twelve months.
How that result is interpreted, for example the proportion of patients achieving a prespecified meaningful improvement without revision.
Key message. A questionnaire becomes a study endpoint only when the population, variable, timepoint, scoring method and analysis are defined.
Much of the confusion in arthroplasty protocols comes from treating a questionnaire as an endpoint. Domain, instrument, endpoint and responder definition are four separate decisions.
A patient-reported outcome measure is completed directly by the patient. Several scores in routine arthroplasty use are mixed or clinician-assessed, and labeling them as PROMs creates problems later.
Where a hip PROM is required, the Oxford Hip Score is worth including alongside the Oxford Knee Score: it is the hip counterpart to OKS and was developed specifically to assess patient-reported pain and function after hip replacement.
An endpoint should be derived step by step from the claim the device is intended to support.
The hierarchy is what protects the primary analysis from multiplicity and keeps the conclusion readable.
The main result answering the principal clinical question. Clinically meaningful, relevant to the device claim, measurable within the planned follow-up, supported by the sample-size rationale, defined before enrollment, and interpretable if patients undergo revision, withdraw or miss follow-up.
Additional safety, performance or benefit questions: a joint-specific PROM, pain, imaging, revision, range of movement, quality of life.
Hypothesis-generating rather than confirmatory: wearable-device data, gait analysis, novel imaging metrics, activity tracking, early biomarker or sensor measures.
The primary endpoint carries the study. Secondary endpoints are prespecified and ordered. Exploratory endpoints are reported as exploratory and never used to rescue a conclusion.
Device-related, procedure-related and serious adverse events, infection, dislocation or instability, periprosthetic fracture, neurovascular injury, thromboembolic events, wound complications, implant breakage or malfunction, reoperation, revision and death.
Successful implantation, completion using the intended system, conversion to conventional instrumentation, device or instrument deficiency, procedure duration, blood loss, intraoperative fracture, unplanned instrumentation, planned versus achieved position, registration or tracking failure, aborted robotic workflow, surgeon-reported technical difficulty.
Pain reduction, functional improvement, stability, range of movement, ability to complete rehabilitation, implant position and alignment, fixation, migration, radiographic progression, return to activity, satisfaction and revision-free survival.
Important distinction. Revision, reoperation and additional procedure are not interchangeable terms. Each has a different clinical meaning and should be defined and counted separately. Equally, procedural success must not be allowed to stand in for patient benefit or longer-term implant performance.
Every safety endpoint needs an event definition, device and procedure relatedness, severity, seriousness, adjudication process, reporting window, denominator, handling of repeated events and a decision on whether the analysis is component-specific or device-wide, all prespecified with our biostatistics team before enrollment.
Each joint has its own failure modes, outcome conventions and ceiling effects. The instrument that answers a hip fixation question is not the one that answers a knee function claim, and anatomic and reverse shoulder arthroplasty need different logic again.
The instrument follows the joint and the claim. A generic health-status measure is collected alongside the joint-specific score, not instead of it.
A widely recognized clinical score covering pain, function, deformity and range of movement, requiring clinician assessment in its original form. Useful clinical context, but it should not be labeled a PROM, and ceiling effects should be considered in high-performing populations.
Patient-reported across pain, symptoms, activities of daily living, sport and recreation, and hip-related quality of life. The official guide describes five separate subscales rather than one interchangeable total, so the protocol must prespecify which subscales or summary version is used.
A twelve-item patient-reported measure focused on pain and function, designed for hip replacement with a lower burden than longer multidomain measures. Licensing and approved translations must be confirmed before use.
Other hip endpoints commonly considered: dislocation, periprosthetic fracture, migration or subsidence, radiolucent lines, revision, leg-length discrepancy, patient satisfaction and return to usual activity.
Other knee endpoints commonly considered: alignment, instability, stiffness, range of movement, manipulation under anesthesia, patellar complications, revision, satisfaction, return to work or activity, and radiographic loosening or migration.
Important distinction. Anatomic and reverse shoulder arthroplasty should not automatically use identical endpoint logic. The failure modes and functional expectations differ, and the endpoints should reflect that.
Other shoulder endpoints commonly considered: instability or dislocation, glenoid loosening, scapular notching, tuberosity healing where relevant, rotator-cuff-related function, range of movement, revision, acromial or scapular spine fracture, and patient satisfaction.
PROM licensing, validated translations and ePRO implementation differ by country. The endpoint plan has to be feasible in every market the study runs in, not only on paper.
Instrument licensing, translation status and ePRO readiness are assessed site by site during feasibility.
The right instrument measures the claimed benefit, in the relevant population, at the relevant study stage, with acceptable burden and responsiveness.
Licensing note. The Oxford Hip and Knee Scores require attention to copyright licensing and approved versions, including electronic implementation. Secure the relevant permission before the database is built.
A mean change is the average improvement from baseline. It does not show how many patients achieved a benefit that matters to them.
The smallest within-patient change considered meaningful in the specified context.
An absolute postoperative state that patients may regard as satisfactory.
A greater degree of improvement than the minimum threshold.
Meaningful PROM improvement, no revision, no serious device-related event, acceptable imaging and no major reoperation, combined into one definition.
Do not copy a published MCID into a protocol without checking the instrument version, joint, primary or revision population, baseline severity, follow-up timepoint, anchor used, study objective and estimation method. Published meaningful-change estimates for OHS and OKS vary by context and method, which is exactly why a study-specific justification is required rather than a universal cut-off.
The schedule should follow the biological and clinical process, not convention.
Key message. Do not select twelve months simply because it is common. Explain why that timepoint is appropriate for the intended claim.
Data go missing for foreseeable reasons: a missed visit, withdrawal, revision, death, an inability to complete a PROM, site closure, lost contact or a technical ePRO failure. Because they are foreseeable, the protocol and SAP should plan for them in advance rather than defaulting to complete-case analysis.
Key message. Complete-case analysis should not be the default simply because it is easy.
Revision, death, loss to follow-up and incomplete questionnaires are foreseeable. Each one has an answer written into the protocol and the statistical analysis plan.
An intercurrent event occurs after treatment and affects either the interpretation or the existence of later measurements. In arthroplasty: revision, reoperation, contralateral joint replacement, major trauma, new neurological disease, death, rescue treatment, delayed rehabilitation, or withdrawal after a device-related event.
Two different questions. What is knee function at twelve months under a treatment policy that includes any revision during follow-up? Compare that with: what would knee function have been without revision? These have different interpretations and may require different methods.
The ICH E9(R1) estimand framework is not specific to medical devices, but its discipline is useful when defining what the study is actually trying to estimate after events such as revision, rescue intervention or death.
The imaging side of the strategy (acquisition, projections, core-lab reading and RSA) is covered on imaging, core lab and RSA for arthroplasty studies, and applied study by study alongside our biostatistics team.
At a high level, the analysis plan should address each of the following before the first patient is enrolled.
Sample size, SAP and analysis are produced by our medical device biostatistics team, using the same data management pipeline that carries the study to database lock.
Mapping the flow before enrollment is what keeps every result traceable to the claim it is meant to support.
Key message. The CER should conclude on the same clinical question the protocol was designed to answer. It should not broaden the claim after seeing the results, one of the most common non-conformities in MDR clinical evaluation.
For an existing protocol requiring review. Output: gap map and endpoint recommendations.
For uncertainty over outcome instruments. Output: instrument comparison and recommendation.
For an early-stage program. Output: claim-to-endpoint framework.
For primary endpoint or analysis uncertainty. Output: sample-size and analysis recommendations.
For a new clinical investigation. Output: full synopsis or clinical investigation plan.
For an existing or completed study. Output: traceability from results to CER conclusions.
The arthroplasty program is led by Dr Nikhil Khadabadi, an NHS orthopedic surgeon and former TÜV SÜD Notified Body clinical reviewer.
Endpoints, sample size, CRF structure, SAP and reporting are planned together, not sequentially.
The same evidence logic runs from protocol design through the clinical study report, PMCF and CER.
Commission a bounded endpoint review, or full protocol and study delivery.
Timelines follow protocol maturity, the number of intended claims, the number of joints or implant configurations, the evidence available for the instrument, translation and licensing requirements, statistical complexity, imaging integration and any need for authority or expert feedback. Costs follow whether a full protocol or a focused endpoint review is required, the number of PROMs, electronic implementation, translation and linguistic validation, statistical simulations, responder and sensitivity analyses, CER and PMCF integration, and the number of revision cycles. We quote in defined work packages, using the same model as our European clinical trial cost benchmarking.
Related Eclevar pages across the orthopedic program.
The instrument should follow the claim, population and timepoint. The Oxford Hip Score suits a concise patient-reported endpoint, while HOOS suits a multidomain assessment, provided the subscales are prespecified.
The Harris Hip Score is a clinician-assessed or mixed score, whereas HOOS is patient-reported across five subscales. They answer different questions and should not be treated as interchangeable.
The Oxford Knee Score is concise and low-burden for a focused functional endpoint. KOOS is broader, so if it is used, the relevant subscales should be selected in advance.
No. It combines patient-reported information with clinician-assessed strength and range of movement, so it is a mixed score rather than a pure PROM.
Yes, but anatomic and reverse shoulder arthroplasty differ in failure modes and functional expectations, so the wider endpoint set should reflect the specific procedure rather than being copied across.
Prospectively, using a threshold justified for the specific instrument, joint, population, baseline severity and timepoint, rather than a universal published cut-off.
At timepoints matched to the process being assessed, typically a preoperative baseline followed by assessments of early recovery, function and more stable outcomes. The schedule should be justified by the claim.
By defining it as an intercurrent event in advance, deciding whether it counts as failure and how pre-revision and post-revision data are treated, and stating the estimand the analysis targets.
Plan for it before it happens, with defined windows, reminders, item-level scoring rules, a primary missing-data assumption and sensitivity analyses, rather than defaulting to complete-case analysis.
Each endpoint should be traceable to a claim and flow through the clinical study report and, where applicable, the PMCF evaluation report into the CER. The CER should conclude on the same question the protocol was designed to answer.
Yes. The Protocol and Endpoint Review examines claim alignment, endpoint hierarchy, timepoints, responder logic, missing-data handling and downstream CER use, and can be commissioned before submission or after feedback.
Sometimes, but registry variables are not always sufficient for a prospective claim. The set should be checked against the intended claim, responder definition and CER use before it is adopted.
Send us your device description, intended claims, draft protocol and proposed outcome measures. We review the endpoint hierarchy, PROM selection, assessment timepoints, responder logic, missing-data strategy and downstream CER use.
Your documents are reviewed confidentially. An NDA can be put in place before we receive any technical or clinical information. You can also reach the team through the contact page.
