Why Science? | FDA Patient Preference Information, Benefit-Risk and Device Decisions

FDA defines patient preference information as assessments of the relative desirability or acceptability to patients of specified alternatives or choices among outcomes or other attributes that differ among health interventions. The definition is about choices and trade-offs. A general satisfaction score or an isolated opinion may be patient input without answering a preference question.

The cheerful power of ppi studies relative desirability or acceptability is that it makes a human judgement visible enough to examine. Here, FDA defines patient preference information as assessments of the relative desirability or acceptability to patients of specified alternatives or choices among outcomes or other attributes that differ among health interventions. The definition is about choices and trade-offs. A general satisfaction score or an isolated opinion may be patient input without answering a preference question. The honest boundary is: PPI is one type of patient input and must be matched to a defined decision. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Rewrite a broad question such as “Do patients like this device?” as a comparison with explicit outcomes, burdens or risks. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Next chapter

FDA issued the final guidance “Incorporating Voluntary Patient Preference Information over the Total Product Life Cycle” in March 2026. It provides recommendations on when and what methods to use and supersedes the 2016 device PPI guidance. Recording the guidance date and docket prevents an analysis from silently relying on older recommendations.

For the march 2026 guidance is the current final document, methods matter because wording, context and who answers can change the apparent preference. FDA issued the final guidance “Incorporating Voluntary Patient Preference Information over the Total Product Life Cycle” in March 2026. It provides recommendations on when and what methods to use and supersedes the 2016 device PPI guidance. Recording the guidance date and docket prevents an analysis from silently relying on older recommendations. Keep this guardrail close: Guidance states FDA’s current thinking and does not create legally enforceable requirements.

A strong science note separates observation from interpretation. Create a source receipt with title, month, final status, docket FDA-2015-D-1580 and superseded-document note. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A PPI project should begin with the decision it could inform: selecting design features, choosing outcomes, planning a clinical study, understanding benefit-risk trade-offs or communicating information. A narrow decision makes it easier to select participants, attributes and a suitable method. Collecting preferences first and inventing a use later invites selective interpretation.

Imagine explaining start with the regulatory or development decision to a family deciding between two options. The evidence should clarify the trade-off without pretending everyone chooses alike. In this case, A PPI project should begin with the decision it could inform: selecting design features, choosing outcomes, planning a clinical study, understanding benefit-risk trade-offs or communicating information. A narrow decision makes it easier to select participants, attributes and a suitable method. Collecting preferences first and inventing a use later invites selective interpretation. Remember: A study useful for one life-cycle decision may be uninformative for another.

A strong science note separates observation from interpretation. State the decision, decision-maker, timing and action that could change because of the PPI. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The relevant respondent depends on who experiences the condition, device, benefit, risk and burden. Adults may report their own preferences; caregivers may sometimes provide distinct information about care demands or choices. The protocol should explain whose perspective is sought and avoid treating different roles as interchangeable.

The cheerful power of patients and caregivers can illuminate different questions is that it makes a human judgement visible enough to examine. Here, The relevant respondent depends on who experiences the condition, device, benefit, risk and burden. Adults may report their own preferences; caregivers may sometimes provide distinct information about care demands or choices. The protocol should explain whose perspective is sought and avoid treating different roles as interchangeable. The honest boundary is: Caregiver preference is not automatically a proxy for a patient’s preference. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Label each proposed respondent group and the exact perspective it contributes. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Interviews, focus groups and other qualitative methods can uncover outcomes, burdens, language and decision processes that designers did not anticipate. This work can shape attributes and levels for later quantitative study. Rigour comes from a clear sampling plan, interview guide, analysis method, coding decisions and evidence of thematic saturation or adequacy.

For qualitative work can discover what matters, methods matter because wording, context and who answers can change the apparent preference. Interviews, focus groups and other qualitative methods can uncover outcomes, burdens, language and decision processes that designers did not anticipate. This work can shape attributes and levels for later quantitative study. Rigour comes from a clear sampling plan, interview guide, analysis method, coding decisions and evidence of thematic saturation or adequacy. Keep this guardrail close: Qualitative insight describes perspectives; it does not by itself estimate population trade-offs.

A strong science note separates observation from interpretation. Trace one patient statement into a candidate attribute while preserving context and an alternative interpretation. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Quantitative PPI methods ask respondents to evaluate alternatives, rate importance or make choices so that preferences can be estimated. Examples can include discrete-choice experiments, threshold techniques or other elicitation approaches. Method choice depends on the decision, number and complexity of attributes, respondent burden and the quantity the study needs to estimate.

Imagine explaining quantitative methods estimate structured trade-offs to a family deciding between two options. The evidence should clarify the trade-off without pretending everyone chooses alike. In this case, Quantitative PPI methods ask respondents to evaluate alternatives, rate importance or make choices so that preferences can be estimated. Examples can include discrete-choice experiments, threshold techniques or other elicitation approaches. Method choice depends on the decision, number and complexity of attributes, respondent burden and the quantity the study needs to estimate. Remember: A sophisticated model cannot repair a choice task that respondents did not understand.

A strong science note separates observation from interpretation. Match a decision question to a quantitative output and explain why that output would change the decision. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Attributes describe features such as benefit, risk, treatment burden, recovery or convenience; levels describe the possible values. They should be understandable, plausible, distinct enough for respondents to evaluate and relevant to the decision. Patient input, clinical evidence and technical feasibility can all inform selection, with documentation of what was excluded.

The cheerful power of attributes need patient meaning and decision relevance is that it makes a human judgement visible enough to examine. Here, Attributes describe features such as benefit, risk, treatment burden, recovery or convenience; levels describe the possible values. They should be understandable, plausible, distinct enough for respondents to evaluate and relevant to the decision. Patient input, clinical evidence and technical feasibility can all inform selection, with documentation of what was excluded. The honest boundary is: Omitted or overlapping attributes can distort apparent importance. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Audit each attribute for plain language, independence, realistic levels and connection to the decision. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The study population should resemble the people whose preferences are intended to inform the decision, including relevant disease severity, experience and treatment context. Eligibility, recruitment channels and non-response can affect who is represented. A convenient online sample may answer a survey well yet differ from the intended device population.

For the target population anchors interpretation, methods matter because wording, context and who answers can change the apparent preference. The study population should resemble the people whose preferences are intended to inform the decision, including relevant disease severity, experience and treatment context. Eligibility, recruitment channels and non-response can affect who is represented. A convenient online sample may answer a survey well yet differ from the intended device population. Keep this guardrail close: Results cannot be assumed to represent groups that were not adequately sampled.

A strong science note separates observation from interpretation. Draw a population funnel from intended users to invited participants, responders and analysed respondents. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Preference tasks may involve probabilities, unfamiliar outcomes or difficult trade-offs. Piloting, cognitive interviews, plain-language explanations, visual aids and comprehension checks can reveal misunderstandings. Excluding everyone who struggles may make the analysed sample less representative, so the protocol should anticipate support and report how comprehension was handled.

Imagine explaining comprehension is a measurement property to a family deciding between two options. The evidence should clarify the trade-off without pretending everyone chooses alike. In this case, Preference tasks may involve probabilities, unfamiliar outcomes or difficult trade-offs. Piloting, cognitive interviews, plain-language explanations, visual aids and comprehension checks can reveal misunderstandings. Excluding everyone who struggles may make the analysed sample less representative, so the protocol should anticipate support and report how comprehension was handled. Remember: Passing a single check does not prove every respondent interpreted every attribute identically.

A strong science note separates observation from interpretation. Design one comprehension question that tests the trade-off rather than simple recall. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The number of respondents should be justified by the intended estimates, design complexity, precision and subgroup plans. Probability sampling may be difficult, but recruitment and weighting choices should still be transparent. A small exploratory study can be useful when described honestly; precision claims should match the information collected.

The cheerful power of sampling and sample size follow the estimand is that it makes a human judgement visible enough to examine. Here, The number of respondents should be justified by the intended estimates, design complexity, precision and subgroup plans. Probability sampling may be difficult, but recruitment and weighting choices should still be transparent. A small exploratory study can be useful when described honestly; precision claims should match the information collected. The honest boundary is: A large sample does not eliminate selection bias or poor measurement. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Write the primary preference quantity to estimate, then list the assumptions used in the sample-size rationale. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The protocol should describe objectives, population, recruitment, attributes, elicitation tasks, endpoints or estimands, quality checks and analysis. Prespecifying primary analyses reduces freedom to choose the most attractive result after seeing the data. Deviations may be reasonable, but they should be dated, explained and separated from exploratory work.

For protocol and analysis should be prespecified, methods matter because wording, context and who answers can change the apparent preference. The protocol should describe objectives, population, recruitment, attributes, elicitation tasks, endpoints or estimands, quality checks and analysis. Prespecifying primary analyses reduces freedom to choose the most attractive result after seeing the data. Deviations may be reasonable, but they should be dated, explained and separated from exploratory work. Keep this guardrail close: Prespecification supports credibility but does not make a flawed design valid.

A strong science note separates observation from interpretation. Create a one-page analysis receipt listing primary quantity, model, exclusions, missing-data rule and sensitivity analyses. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Researchers can examine incomplete surveys, unusually fast completion, invariant responses, inconsistent choices and other signals. Quality rules should be justified and applied consistently. Some unusual patterns may reflect real preferences or accessibility needs rather than carelessness, so exclusions deserve both scientific and ethical scrutiny.

Imagine explaining data quality includes response behaviour to a family deciding between two options. The evidence should clarify the trade-off without pretending everyone chooses alike. In this case, Researchers can examine incomplete surveys, unusually fast completion, invariant responses, inconsistent choices and other signals. Quality rules should be justified and applied consistently. Some unusual patterns may reflect real preferences or accessibility needs rather than carelessness, so exclusions deserve both scientific and ethical scrutiny. Remember: Mechanical quality filters can remove meaningful minority perspectives.

A strong science note separates observation from interpretation. For each proposed flag, write a benign explanation and a careless-response explanation before deciding how to act. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Average preference estimates can hide groups with different priorities. Disease experience, severity, age, prior treatment, device familiarity or other factors may be associated with variation. Subgroup or latent-class analyses should be planned carefully and interpreted with uncertainty, especially when groups are small.

The cheerful power of heterogeneity is information, not noise is that it makes a human judgement visible enough to examine. Here, Average preference estimates can hide groups with different priorities. Disease experience, severity, age, prior treatment, device familiarity or other factors may be associated with variation. Subgroup or latent-class analyses should be planned carefully and interpreted with uncertainty, especially when groups are small. The honest boundary is: An observed subgroup difference is not automatically causal or stable in a new population. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Report the overall estimate beside subgroup intervals and the number of respondents contributing to each. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

Preference estimates depend on sampling, task design, model form, coding and assumptions. Confidence or credible intervals, alternative specifications, exclusion rules and scenario analyses help show whether the decision-relevant conclusion is robust. A result that changes under modest assumptions should be presented as uncertain rather than hidden.

For uncertainty and sensitivity tests protect the conclusion, methods matter because wording, context and who answers can change the apparent preference. Preference estimates depend on sampling, task design, model form, coding and assumptions. Confidence or credible intervals, alternative specifications, exclusion rules and scenario analyses help show whether the decision-relevant conclusion is robust. A result that changes under modest assumptions should be presented as uncertain rather than hidden. Keep this guardrail close: Statistical precision does not cover every source of bias or misunderstanding.

A strong science note separates observation from interpretation. List the three assumptions most likely to change the preferred option and test each one. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

PPI can help FDA understand how patients value benefits, risks and other attributes when the evidence is fit for the decision. It may illuminate whether some patients accept a risk for a meaningful benefit or which outcomes matter most. It complements clinical, engineering and other evidence rather than substituting for proof of safety or effectiveness.

Imagine explaining ppi can inform benefit-risk assessment to a family deciding between two options. The evidence should clarify the trade-off without pretending everyone chooses alike. In this case, PPI can help FDA understand how patients value benefits, risks and other attributes when the evidence is fit for the decision. It may illuminate whether some patients accept a risk for a meaningful benefit or which outcomes matter most. It complements clinical, engineering and other evidence rather than substituting for proof of safety or effectiveness. Remember: Preference for an option does not establish that the option works or that its risks are controlled.

A strong science note separates observation from interpretation. Build parallel columns for clinical evidence and preference evidence, then show exactly where they meet in the benefit-risk reasoning. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

The 2026 guidance places PPI across the total product life cycle: early development, clinical study design, premarket review, postmarket evaluation and communication. The question, population and method may change at each stage. Reusing an old study requires checking whether the device, alternatives and patient experience remain comparable.

The cheerful power of ppi can contribute across the product life cycle is that it makes a human judgement visible enough to examine. Here, The 2026 guidance places PPI across the total product life cycle: early development, clinical study design, premarket review, postmarket evaluation and communication. The question, population and method may change at each stage. Reusing an old study requires checking whether the device, alternatives and patient experience remain comparable. The honest boundary is: A preference estimate can become stale when options, risks or lived experience change. That combination keeps “patient-centred” from becoming a slogan.

A strong science note separates observation from interpretation. Add a review date and trigger conditions—new device version, new alternative or new safety signal—to the PPI record. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter · Next chapter

A reusable receipt records the decision question, target population, respondent role, recruitment, attributes and levels, method, instrument version, field dates, sample, comprehension work, exclusions, analysis, uncertainty, heterogeneity, findings, limits and official source. It lets another reader distinguish what patients said from what analysts inferred.

For a complete ppi receipt makes patient voice traceable, methods matter because wording, context and who answers can change the apparent preference. A reusable receipt records the decision question, target population, respondent role, recruitment, attributes and levels, method, instrument version, field dates, sample, comprehension work, exclusions, analysis, uncertainty, heterogeneity, findings, limits and official source. It lets another reader distinguish what patients said from what analysts inferred. Keep this guardrail close: The receipt improves transparency; it does not turn one sample into every patient’s voice.

A strong science note separates observation from interpretation. Give the receipt to a reader who has not seen the study and ask what decision the findings can and cannot inform. Record the source, version or date, population or device context, method, result and one uncertainty. If the official record does not show a field, write “not shown” instead of filling the gap from memory.

Contents · Previous chapter