A defensible comparative effectiveness study in wound care lives or dies on decisions made before a single data pull happens. By the end of this guide you’ll have a framework for designing a wound care comparative effectiveness study (CES) that holds up under payer, regulatory, or peer-review scrutiny. You’ll need access to a real-world data source with wound-level granularity, a defined comparator product or care pathway, and clarity on whether the output is intended for post-market clinical follow-up (PMCF), market access, or publication. The steps below assume you already understand the clinical territory and are focused instead on the methodological choices that determine whether your findings survive contact with a skeptical reviewer.
Step 1: Define the Research Question and Comparator Set
Every unusable CES starts with a vague question. “Does Product X improve wound healing” is not a research question; it’s a marketing aspiration. A defensible question specifies the exact comparison: a named CTP against standard of care in a defined wound type, or a head-to-head CTP-to-CTP comparison within a matched population. Write the question in a single sentence that names the intervention, the comparator, the population, and the outcome. If you can’t do that, you’re not ready to pull data.
Decide your primary use case before anything else, because it dictates endpoint selection and sample-size requirements downstream. A study built for PMCF documentation under EU MDR has different evidentiary expectations than one built to support a CMS local coverage determination (LCD) submission, a GPO or IDN value analysis committee packet, or a peer-reviewed manuscript. Trying to serve all four with one design usually means serving none of them well. Pick the primary audience, design to their standard, and treat the others as secondary beneficiaries if the data happens to support them.
The comparator itself deserves more scrutiny than it typically gets. A common and often fatal error is choosing a comparator that isn’t actually used concurrently in the same care settings as the intervention. Comparing a product used almost exclusively in hospital-based wound centers against one used predominantly in skilled nursing facilities introduces selection bias that no amount of statistical adjustment can fully resolve, because the underlying patient populations, staffing models, and follow-up intensity differ systematically by setting. Before finalizing a comparator, verify utilization overlap: are both products actually being applied to similar patients in similar settings during the same time period? If not, either narrow the population to where genuine overlap exists or reconsider the comparator entirely.
Step 2: Define the Wound Population and Inclusion Criteria
Wound etiology is not a demographic detail, it’s a confounder. Pooling diabetic foot ulcers, venous leg ulcers, pressure ulcers, and surgical wounds into a single analysis without stratification is one of the most frequent sources of confounded results in wound care research, because healing trajectories, standard of care, and typical time to closure differ substantially across these categories. Specify etiology up front and decide whether your design will restrict to a single etiology or stratify across several with etiology as a covariate.
Baseline inclusion criteria need to reflect what actually drives outcomes, not just what’s easy to pull. At minimum, define:
- Wound size at baseline (area and, where available, volume)
- Wound duration prior to index treatment
- Wound bed status (granulation, slough, necrotic tissue percentage)
- Prior failed therapies, including standard-of-care duration before escalation
These criteria matter both clinically and statistically. Wounds that are older, larger, or have already failed conservative care behave differently than fresh, smaller wounds, and if these characteristics are unevenly distributed between your intervention and comparator groups, they will masquerade as a treatment effect.
Care setting scope is the other major population decision. Utilization patterns and healing trajectories differ meaningfully across hospital-based wound centers, skilled nursing facilities, and home health, both because of patient acuity differences and because of variation in visit frequency and debridement practice. This is precisely where a single-site EHR pull falls short: it can tell you what happened at one facility type but can’t tell you whether that pattern holds elsewhere. A multi-setting dataset, such as Intellicure’s Wound Care Industry Dashboard, which tracks product-level utilization across hospital-based wound centers, SNFs, and home visits for DFUs, VLUs, and pressure ulcers, lets you either restrict analysis to a single setting for cleaner comparison or explicitly model setting as a covariate when a broader population is required.
Step 3: Choose a Data Source and Study Design
The RCT-versus-real-world-data trade-off is well worn but still worth stating plainly for CTP and CAMP comparisons. Randomized trials offer control over confounding and clean causal inference, but they are slow and expensive relative to the pace of product iteration in this category, and many manufacturers can’t wait two to three years for enrollment and follow-up to support a market access conversation happening now. Real-world data registries offer scale, external validity, and speed, but they require rigorous statistical adjustment to approximate what randomization would have controlled for automatically. Neither is universally superior; the right choice depends on your timeline, your use case from Step 1, and how much residual confounding your reviewers will tolerate.
If you’re going the RWD route, verify that your dataset captures product-level utilization, not just billing codes, across the care settings relevant to your comparator. This is a frequent and costly oversight. Claims data alone often can’t distinguish between similar CTPs billed under the same or overlapping HCPCS codes, which means a claims-only design may silently blend two different products into one comparator arm without anyone realizing it. Before committing to a data source, confirm it resolves utilization down to the specific product, not just the billing category.
Finally, decide between a retrospective cohort, a prospective registry, or a hybrid design based on your timeline and regulatory obligations. Retrospective cohorts are faster to execute but limited to whatever was already documented. Prospective registries allow you to define data capture rules in advance, improving completeness, but take longer to accrue. Hybrid designs, combining a retrospective baseline with prospective follow-up, are increasingly common for PMCF work, since ongoing post-market surveillance is an expectation under EU MDR as of 2026 for CE-marked devices and combination products; confirm current requirements with your regulatory team, since specific timelines and thresholds continue to evolve.
Step 4: Define Outcomes and Control for Confounding
Outcome selection should be driven by what regulators, payers, and peer reviewers actually recognize as defensible, not by what sounds favorable. “Improvement” is not an endpoint. Complete wound closure, time to closure, percent area reduction at fixed intervals (commonly four and twelve weeks), and recurrence within a defined follow-up window are the standard currency. Pick outcomes that match your use case: LCD-oriented work typically leans on closure rates and time to closure, while PMCF documentation often requires safety and performance endpoints tracked over a longer horizon.
Confounding by indication is the structural problem underlying most wound care CES failures, and it deserves direct attention rather than a footnote. In real-world practice, sicker or more complex wounds are frequently assigned to newer, more advanced, or more aggressive therapies, precisely because standard care has already failed. If you compare outcomes without adjusting for this, you’re not measuring product effectiveness, you’re measuring how clinicians triage difficult cases. Propensity score matching, in which patients are matched across treatment arms based on their probability of receiving the intervention given baseline characteristics, is the standard method for approximating balance. Multivariable regression, adjusting directly for baseline covariates within the outcome model, is a reasonable alternative or complement, particularly when sample size is too small to support tight matching.
The single biggest and most avoidable mistake in this space is reporting raw closure rates without adjusting for baseline wound size and duration. These two variables alone frequently explain the majority of the apparent “effectiveness” gap between two products, because a program that happens to enroll smaller, more recent wounds will show better raw outcomes regardless of what’s applied to them. Any comparative claim that doesn’t disclose adjustment for size and duration should be treated as provisional at best, and reviewers increasingly know to ask for it.
Step 5: Determine Sample Size and Statistical Power
Run a formal power calculation before committing to a data pull, not after. Use effect sizes from prior published literature or pilot data as your estimate, and be conservative rather than optimistic, since inflated assumed effect sizes are a common reason CTP comparisons come back underpowered. An underpowered study doesn’t just risk a null result, it risks outright rejection at peer review or dismissal by a payer’s evidence review team, because a non-significant finding in an underpowered study is uninterpretable rather than reassuring.
Real-world cohorts lose patients to follow-up at rates that clinical trial protocols don’t tolerate: transfers between care settings, discharge to hospice, loss of insurance coverage, or simple documentation gaps. Build attrition assumptions into your target sample size from the outset. If your power calculation says you need 200 patients per arm to detect your expected effect size, and your anticipated attrition rate over the follow-up window is 20 to 25 percent, you need to pull closer to 250 to 270 per arm to end the study adequately powered.
For rare wound subtypes or niche products with limited market penetration, adequate power within a reasonable timeframe may simply not be achievable from a single site or even a small multi-site network. In these cases, a comprehensive real-world dataset spanning many care settings and a large patient volume, of the kind aggregated across hospital-based wound centers, SNFs, and home health programs, may be the only realistic path to sufficient sample size without stretching the study timeline past its usefulness.
Step 6: Execute Data Collection and Quality Control
Data quality problems surface late and expensively if they aren’t addressed before analysis begins. Establish validation rules up front for the fields that matter most: wound measurement consistency (same measurement method across sites), application dates for the intervention and comparator, and product identifiers specific enough to distinguish similar CTPs. Inconsistent documentation across sites, different staff recording wound area differently, gaps in application logs, ambiguous product coding, is the most common data quality failure in multi-site wound studies, and it’s far cheaper to catch with automated validation rules than to discover during peer review.
Document data provenance and completeness rates as a standard part of your methods, not as an afterthought. State how many sites contributed data, what percentage of expected follow-up visits were captured, and what proportion of records met inclusion criteria after cleaning. Reviewers and payer evidence teams increasingly ask for this transparency before they’ll accept real-world evidence at face value, and a study that volunteers this information upfront reads as more credible than one that has it extracted under questioning.
Building a bespoke, validated multi-site data pipeline from scratch is a significant undertaking, and many manufacturer teams underestimate the timeline and QA burden involved. Where that’s impractical given internal resourcing, structured support, such as Intellicure’s Comparative Effectiveness Studies or Clinical Trials services, can shorten the path from raw data to analysis-ready dataset considerably compared to standing up an internal pipeline, particularly for teams that don’t have dedicated data engineering staff for this work.
Step 7: Analyze, Report, and Disseminate Findings
Report adjusted effect sizes with confidence intervals, not raw comparisons dressed up as findings. A hazard ratio or odds ratio with a 95 percent confidence interval tells a reviewer far more than a percentage-point difference in closure rates, and it signals that you understand the difference between association and adjusted effect. Disclose every covariate included in your adjustment model, including wound size, duration, care setting, and any comorbidities you controlled for. Omitting this invites the reasonable assumption that you’re hiding an unfavorable sensitivity analysis.
State your limitations explicitly and early in the report, not buried in a final paragraph. Care setting mix, loss to follow-up, and the possibility of unmeasured confounders (documentation practices, referral patterns, formulary constraints) should all be named. This audience, whether a payer medical director, a peer reviewer, or an internal value analysis committee, distrusts unqualified claims more than it distrusts honestly stated uncertainty. A study that says “adjusted for measured confounders; residual confounding from unmeasured comorbidity burden cannot be excluded” reads as more trustworthy than one that claims a clean causal result.
Match your output format to the use case you defined in Step 1. A manuscript intended for peer review needs full methods transparency and should follow relevant reporting guidelines for observational studies. A PMCF report needs to align with your regulatory file structure and safety reporting obligations. A ranking or benchmark summary intended for internal commercial teams or a GPO value analysis submission can be more concise but should still carry the adjusted findings and stated limitations, not just the headline number.
Confirming Fit Before You Finalize
Before finalizing anything, go back to the use case you named in Step 1 and confirm the finished study actually maps to it: a PMCF report needs different documentation than an LCD coverage submission, and a submission built for one rarely repurposes cleanly for the other without revision. Confirm, too, that your confounding adjustments and stated data limitations are documented in enough detail that a skeptical reviewer, whether a payer medical director, a journal referee, or an internal HEOR colleague, can trace exactly what was adjusted for and what wasn’t. A study that survives that level of scrutiny is one that holds up when it matters, in a coverage decision, an audit, or a published record that competitors and payers will read for years.