Clicky

Medical & Clinical Research Journal

Medical & Clinical Research Journal — Peer-reviewed health analysis, supplement safety reviews, and clinical research summaries.

Info

Primary Outcomes and Secondary Outcomes: What a Trial Was Built to Measure

A primary outcome is the prespecified result a trial treats as most important. Secondary outcomes are other planned measurements that add context. Results that were never planned in advance are exploratory, so this page suggests reading them as questions for further study, not as what the trial set out to answer.

The Short Answer

  • Primary outcome: the prespecified outcome considered of greatest importance. According to the CONSORT abstracts guidance, it is usually the one used in the sample size calculation.
  • Secondary outcomes: other planned measurements, such as symptoms, safety, or function, that add context.
  • Exploratory findings: results that were not prespecified as the main question.

The practical habit: find what the trial said it would measure before it started, then compare that with what the paper emphasizes.

Why the Distinction Matters

A standard textbook chapter on field trials describes Phase III trials as aiming to give a definitive assessment of an intervention's efficacy against the primary outcome(s) of interest (Smith, Morrow and Ross, Field Trials of Health Interventions, Chapter 2). Secondary outcomes help describe the wider picture.

The CONSORT abstracts guidance notes evidence that outcomes with significant or favourable results are more likely to be published than others, and says authors should state the primary outcome and when it was assessed (CONSORT for abstracts: Explanation and Elaboration). That is why readers benefit from checking the plan themselves.

The CONSORT 2025 Statement is described by its authors as a minimum set of recommendations for reporting the results of randomised trials, with a 30-item checklist and a flow diagram. SPIRIT 2025 covers the trial protocol in a 34-item checklist (SPIRIT-CONSORT website). One guides how the plan is written down; the other guides how results are reported.

Where to Find the Plan

  1. Trial registry entry. ClinicalTrials.gov instructs data providers to report the specific measure and time frame for each primary and secondary outcome at registration (Zarin et al., NEJM). Other registries may differ, so check what each one lists.
  2. Methods section of the paper. Look for wording such as “the primary outcome was.”
  3. Abstract. CONSORT guidance says that when an abstract reports something other than the primary outcome, the secondary nature of the outcomes should be indicated and enough detail on the primary outcome given for context (CONSORT for abstracts).
  4. Protocol or statistical analysis plan, when published or linked.

Reader Worksheet: Match the Question to the Outcome

Use one copy per trial. Fill in each line from the registry entry first, then from the paper.

  1. Trial question. Write it in one sentence: in which people, comparing what, to see what change?
  2. Primary outcome. Registry wording: ________ Paper wording: ________ Do they match? If the paper names several primary outcomes, or none clearly, note that here.
  3. Time point. When was the primary outcome assessed? Registry: ________ Paper: ________
  4. Secondary outcomes. List each planned one. Mark each “in registry” or “not in registry.”
  5. Result location. Note where the primary result appears (abstract, results text, figure, or supplement) and where each secondary result appears.
  6. Emphasis check. Does the abstract lead with the primary result? If it leads with a secondary or unplanned result, note that.
  7. Changes. Was any outcome or time point changed after the trial began? Is a reason given?
  8. My conclusion. In one sentence, what did the trial answer, and what remains a question for later study?

Limits of This Approach

  • A registry entry can be updated over time. Check its history where the registry provides it.
  • CONSORT and SPIRIT, cited here, are for randomised trials. This worksheet may not fit other study designs.
  • Even a clearly labeled primary outcome does not by itself show that a result matters for any individual patient.

Who Should Seek Professional Advice

This article explains how to read research. It is general education and not medical advice. If you are deciding about a treatment, a medication, or joining a trial, discuss the specific study with a physician, pharmacist, or the trial's research team. See our Medical Disclaimer.

What to Do Next

  • Pick one trial you have read about and complete the worksheet.
  • If the registry and paper disagree, write down the difference and look for an explanation in the paper.
  • Bring questions about your own care to a qualified professional.

About This Page

This page was prepared from source-based desk research. It has not been peer reviewed or clinically reviewed. See Editorial Team and Review Process and About Med Clin Res.

Sources

  • Smith PG, Morrow RH, Ross DA, editors. Field Trials of Health Interventions: A Toolbox, 3rd edition, Chapter 2. NCBI Bookshelf NBK305514.
  • SPIRIT-CONSORT. consort-spirit.org (the consort-statement.org address redirects here).
  • Hopewell S, et al. CONSORT for reporting randomized controlled trials in journal and conference abstracts: explanation and elaboration. PLoS Med 2008. PMC2211558.
  • Zarin DA, et al. The ClinicalTrials.gov results database: update and key issues. N Engl J Med 2011. NEJM.

Adverse Event Tables in Clinical Trials: What They Report and What They Leave Open

By MedClinRes.org Clinical Research Team

A study's adverse event table can look like a verdict on safety. It isn't one. That table is a record of what happened to a specific group of trial participants, under specific monitoring conditions, over a specific period of time — and reading it well means understanding what those numbers do and don't say before drawing any conclusion about a particular medicine. This guide walks through how to read an adverse event (also called a “harms” or “safety”) table in a clinical trial report, using the reporting standards researchers are expected to follow and the FDA's own public adverse-event data as reference points, without translating any of it into a personal safety judgment.

What an Adverse Event Table Actually Shows

Most randomized trial reports include a table listing adverse events by treatment group, usually with a count and percentage of participants who experienced each event. That structure looks simple, but each column carries assumptions that change what a number means:

  • Event definition — what counts as “headache,” “nausea,” or “serious adverse event” is set by the trial's own protocol and coded against a standard terminology (commonly MedDRA). Two trials can use different thresholds for the same word.
  • Group size (the denominator) — a rate only means something next to the number of people who could have had the event. “12 cases” is a different finding in a trial of 40 participants than in a trial of 4,000.
  • Severity — trials generally grade events as mild, moderate, or severe, and separately flag “serious adverse events” — a regulatory term defined by specific outcomes such as death, hospitalization, or a life-threatening event, not by how unpleasant an event felt.
  • Timing — when an event occurred relative to starting treatment, and how long participants were followed, shapes whether an event could plausibly relate to the drug at all.
  • Discontinuation — the number of participants who stopped treatment because of an adverse event is often a more informative figure than the raw event count, since it reflects tolerability in practice.

The CONSORT Statement (CONsolidated Standards Of Reporting Trials) is the reporting framework most medical journals ask randomized-trial authors to follow. It is, in its own description, “an evidence-based, minimum set of recommendations for reporting the results of randomised trials,” built around a checklist and a participant flow diagram so readers “have all the information they need to critically appraise, interpret and use the research.”

Why Harms Reporting Gets Its Own Standard

Benefit outcomes in a trial are usually prespecified precisely. Harms reporting has historically been looser — which is why CONSORT has a dedicated extension for it. The 2023 update, CONSORT Harms 2022, revises 13 checklist items and adds three new ones specifically for how trials should report adverse events. Among its requirements: authors should state whether harms were collected systematically (structured questionnaires or clinical exams) or passively (spontaneous reporting), describe how and when harms were assessed, report any methods used to judge whether an event was related to treatment and who made that judgment, and report zero-event outcomes rather than omitting them — a step aimed at preventing selective reporting. The guideline's authors cite prior research on underreporting of serious harms — including deaths — in published trial results as part of the rationale for these changes, which is the kind of gap a table-reading checklist can't fix but should make a reader cautious about.

In practice, that means a table with cleaner, better-defined categories and a stated collection method deserves more confidence than one with vague event labels and no explanation of how harms were tracked.

What FDA Adverse Event Data Adds — and What It Doesn't

Trial safety tables aren't the only source of adverse event information. Once a drug is on the market, the FDA collects postmarket reports through what was long known as FAERS (the FDA Adverse Event Reporting System). In 2026, FDA consolidated FAERS and several other legacy safety databases into a single public platform, the Adverse Event Monitoring System (AEMS), which the agency says is intended to make postmarket safety data available in closer to real time. The underlying caution that applied to FAERS data applies to AEMS data as well, and FDA states it directly: “The existence of adverse event reports for a drug or biologic…does not mean that the drug or biologic caused the adverse event.” Reports reflect what a patient, caregiver, clinician, or manufacturer chose to submit and describe — not a confirmed medical determination.

That distinction matters because postmarket reporting has structural limits that a trial's controlled reporting doesn't have to the same degree:

  • Reporting is largely voluntary for patients and healthcare professionals, so the database is understood to capture only a fraction of events that actually occur.
  • There's no reliable denominator — the system doesn't know how many people took a drug without incident, so raw report counts can't be turned into a rate.
  • The same case can appear more than once (for example, reported separately by a consumer and by the manufacturer), and a single case can list multiple reactions, so summed counts can look larger than the number of distinct people affected.
  • FDA's own guidance is explicit that this data “by themselves are not an indicator of the safety profile of the drug.”

The practical takeaway: postmarket adverse event data is useful for spotting patterns FDA may want to investigate further, and it's a public window into what's being reported about a drug. It is not, on its own, a safety determination — for the same reason a trial's raw event count isn't one either.

A Short Checklist for Reading Any Safety Table

Before drawing any conclusion from a table, it helps to locate six things:

  1. How is each event defined, and was a standard terminology used?
  2. What is the group size behind each percentage — and is it the same denominator throughout the table?
  3. How is severity graded, and is “serious adverse event” being used in its regulatory sense or its plain-language sense?
  4. When did events occur relative to starting treatment, and how long was follow-up?
  5. How many participants discontinued treatment because of an adverse event, separate from how many merely reported one?
  6. Was harms collection systematic or passive, and does the report say so?

If a table or its surrounding text doesn't answer one of these, that's a gap worth noting rather than filling in with an assumption.

Questions Worth Bringing to a Clinician or Pharmacist

A safety table describes a study population, not any one reader. Translating it into a personal decision is exactly the step this guide doesn't attempt — that's a conversation for a licensed clinician or pharmacist who knows an individual's health history, other medications, and risk factors. Useful questions to bring to that conversation include: How does my own health history compare to the population studied in this trial? Are any of the reported adverse events relevant to conditions or medications I already have? Was the discontinuation rate in this trial higher or lower than what's typical for similar treatments? Is there more recent postmarket safety information, such as an FDA safety communication, that updates what this trial found?

What This Guide Does Not Do

This article explains how to read the structure of a safety table and where FDA's public adverse event data fits alongside it. It does not evaluate any specific drug, supplement, or treatment, and it is not a substitute for reading a trial's full published methods or for guidance from a licensed healthcare provider. Readers with questions about whether a particular medicine is appropriate for them should discuss the full clinical picture with a clinician or pharmacist rather than relying on a single table or database entry.

For related reading on how to evaluate trial evidence, see our guides on how to find a trial's primary outcome and time frame, how to match a study with its registry record, and why the studies included in a systematic review matter. For help interpreting the results a trial reports alongside its harms, see what a confidence interval adds to a result. See also our Editorial Standards and Medical Disclaimer.

Preprints, Peer Review, and Published Papers: How the Evidence Status Changes

Short answer: Before you share or act on a health claim, find out which stage the paper is at: preprint, journal-published, corrected, flagged with an expression of concern, or retracted. Each stage means something different about how far the findings have been checked. You can confirm the stage in a few minutes by looking up the paper in the journal record and in a database such as PubMed.

This article is an educational explainer prepared by the MedClinRes.org Editorial Team. It has not been peer reviewed or clinically reviewed, and it is not medical advice. Source checks were last completed on September 30, 2026.

Why the publication stage matters

A headline that says “a study found” can describe very different documents. Some have been examined by outside experts and edited by a journal. Others were posted online by the authors and have not been examined by anyone. The same paper can also move between stages over time, so a claim that was accurate last year may no longer be accurate now.

The status map: from preprint to corrected or retracted

Stage 1: Preprint

  • A preprint is a manuscript made public before journal peer review. The health-science server medRxiv states that preprints are preliminary reports not certified by peer review, should not be relied on to guide clinical practice or health-related behavior, and should not be reported in news media as established information.
  • medRxiv also notes that it does not edit or typeset manuscripts before posting.
  • Preprints can have several versions. medRxiv supports version-specific links, so record which version you read.

Stage 2: Preprint with public reviews

  • Some journals and independent review services make peer reviews public, and you can access them from the medRxiv dashboard.
  • Read the reviews themselves. Their presence shows that people have examined the work, but on its own it does not tell you the paper was accepted.

Stage 3: Published in a journal

  • In traditional peer review, journal editors take advice from expert referees who assess the paper and may identify weaknesses in its assumptions, methods and conclusions.
  • The published version can differ from the preprint. In one analysis of COVID-19 papers, 124 of 182 cited preprints had later been published in a peer-reviewed journal, and the title, data or conclusion differed for nearly half of them.
  • Peer review is a check, not a guarantee. A 2020 review of preprints in medicine notes that peer review is imperfect and does not always ensure accurate results.

Stage 4: Corrected

Stage 5: Expression of concern

  • An expression of concern is an editor's notice that flags possible problems without retracting or correcting the article.
  • COPE's retraction guideline advises editors to publish one when a determination is delayed.
  • PubMed links the article to its notice with an “Expression of concern in” line. Treat the findings as uncertain until the journal resolves it.

Stage 6: Retracted

How to verify the stage in five steps

  1. Identify the exact record. Write down the full title, journal or server, DOI, and PMID or PMCID if there is one. Do not rely on a news summary.
  2. Open the journal or server page. Look for the article type, the dates, any version history, and any banner or notice.
  3. Check the database record. In PubMed, look at the publication types and for “Erratum in”, “Expression of concern in”, or “Retracted Publication” lines.
  4. Check PMC if the full text is there. PMC is a full-text archive, not a citation database like PubMed, and its open-access subset also includes preprints. Being in PMC does not by itself tell you the stage.
  5. Search the title again for a later version. Preprint servers link to published versions, but a 2023 study of preprint matching notes that such links can miss published works that changed significantly from the preprint.

What this can and cannot tell you

  • The stage tells you how much checking a paper has had. It does not tell you the result is true or that it applies to you.
  • A journal's stated process is worth reading, but its label is not proof that the process is sound. Check whether it is a known, indexed publication.
  • If you cannot find a notice, that is not proof none exists. Databases can lag behind journals.

Who should seek professional advice?

Do not start, stop, or change a medication, supplement, or treatment because of a preprint or a single study. Talk with a licensed clinician or pharmacist who knows your history. If you are a journalist, clinician or researcher relaying a finding, state its stage in plain words: “preprint, not peer reviewed” or “retracted in [year]”.

What to do next

Sources

Confidence Intervals in Health Research: What the Range Adds to a Result

By MedClinRes.org Clinical Research Team. This article has not been peer-reviewed or clinically reviewed, and it is general education, not medical advice.

Short answer: a confidence interval shows how precisely a study pinned down its estimate. It marks a range of effects that are compatible with that study's result. In this article's reading (an editorial interpretation, not a statement from the sources), that range describes the studied group and the uncertainty in the summary number. It is not a forecast of what will happen to you or to any one person.

What the range adds to the number

A study result usually arrives as a single figure, the estimate. The interval beside it addresses a second question: how much uncertainty surrounds that figure. According to the Cochrane Handbook chapter on analyzing data, the interval conveys the precision of a summary estimate, while the P value conveys how strongly the data argue against there being no effect. The two answer different questions, so neither replaces the other.

The Handbook also describes the forest plot, the standard chart for combined results. Each study appears as a block at its estimate, with a horizontal line through it. That line is the confidence interval, most often drawn at the 95% level, and it shows the range of effects compatible with that study's result. The size of the block reflects the weight the study carries. Larger studies have smaller standard errors and receive more weight, and heavily weighted studies usually have narrower intervals.

Most of that chapter concerns meta-analysis, the statistical combination of results from separate studies. Where a point below applies only to combined results, this article says so.

A result card you can fill in

When you read a study, separate the four parts below. If you cannot fill one in from the paper, treat that gap as information too.

  • Estimate: the single summary figure, and which measure it is (for example, a ratio or a difference between groups).
  • Interval: the lower and upper limits and the confidence level, usually 95%. Note whether the interval includes the value that means “no difference” for that measure. The paper should say which value that is.
  • Study population: who was enrolled, in what setting, for how long, and how many people.
  • Unanswered question: what the study did not test, such as other populations, longer follow-up, or outcomes it did not measure.

Four common ways intervals get over-read

1. Treating the range as a personal prediction

The Handbook notes that absolute measures of effect are often easier for clinicians to interpret than relative ones, but are less likely to generalize. Converting a relative result into an absolute one requires assuming a baseline risk for the comparison population. Taken together, these points suggest why an interval calculated across a study group should not be read as a personal prediction. That conclusion is this article's interpretation.

2. Confusing the interval with the spread between studies

For a pooled random-effects result, the Handbook explains that the interval describes uncertainty in the location of the average effect. It does not describe how much the studies disagree, and a tight interval can coexist with large variation between studies. Prediction intervals show that variation, and the Handbook encourages them when there are a reasonable number of studies (it gives about five or more as an example).

3. Assuming a narrow interval means a trustworthy result

Precision and validity are different things. The Handbook states that statistical synthesis does not guarantee valid results, and that meta-analyses can mislead seriously when study design, within-study bias, or variation between studies is not carefully considered. It also says a fixed-effect analysis run on studies that genuinely differ can produce an interval that is too narrow and even meaningless.

4. Reading “not significant” as “no effect,” or skipping the population

In its discussion of subgroup comparisons, the Handbook cautions that a non-significant result may simply reflect too little information, not a smaller or absent effect. By extension, a wide interval is best read as a study that could not settle the question (this article's interpretation). Population matters too. A field-trials chapter on NCBI Bookshelf notes that results from trials in high-income countries may not apply directly to low- and middle-income countries, for reasons such as differing prevalence of other infections or nutritional deficiencies. This article extends that caution to any mismatch between a study group and a reader.

Limits of this article

  • It draws on one methods handbook chapter, written for people conducting systematic reviews, and one trials-methods textbook chapter. It is not a full statistics course.
  • The Handbook describes intervals mainly for meta-analyses. Applying the same reading habits to single studies is editorial guidance.
  • Different measures, study designs, and analysis methods change how you should read an interval. The paper's own methods section governs.
  • One Handbook caution deserves a special mention: for study-level comparisons, the pattern across trial averages can differ from, or even run opposite to, the pattern within trials. This limits what study-level results can say about individual differences.

Who should seek professional advice?

If a study result bears on your own treatment, screening, or medication decisions, discuss it with a physician, pharmacist, or other qualified clinician who knows your history. They can weigh the result against your circumstances, which no interval can do. For what this publication does and does not provide, see the About page and the Editorial Team and Review Process page.

What you can do next

  1. Find the full text or registry entry, not just the headline number, and locate the interval.
  2. Fill in the four-part result card above.
  3. Ask whether the interval comes from one study or a pooled analysis, and whether the paper reports how much the studies differed.
  4. Compare the enrolled population with your own situation before drawing any conclusion.
  5. Bring your questions, along with the paper, to a clinician.

Sources

  • Deeks JJ, Higgins JPT, Altman DG, McKenzie JE, Veroniki AA, editors. Chapter 10: Analysing data and undertaking meta-analyses (last updated November 2024). In: Cochrane Handbook for Systematic Reviews of Interventions, version 6.5. Cochrane, 2024. cochrane.org
  • Ross DA. Types of intervention and their development. In: Smith PG, Morrow RH, Ross DA, editors. Field Trials of Health Interventions: A Toolbox, 3rd edition. Oxford: OUP; 2015. NCBI Bookshelf (cited only for the point about applying trial results across populations).

Inclusion and Exclusion Criteria: Who a Clinical Study Can and Cannot Describe

By MedClinRes.org Clinical Research Team. Last updated September 30, 2026.

Before trusting a clinical study to answer your own question, check one thing first: did the study actually include people like you? Every clinical study is built around a specific group of people, defined by inclusion criteria (who is allowed to join) and exclusion criteria (who is kept out). Those rules decide who a study's results can honestly describe — and just as important, who they cannot.

This guide explains what inclusion and exclusion criteria are, why researchers set them, and how to check whether a study's population resembles the situation you're trying to understand. It ends with a short checklist you can use on any study you read.

What Inclusion and Exclusion Criteria Actually Are

The NIH's Toolkit for Clinical Research defines eligibility criteria in two parts: inclusion criteria, the medical or social characteristics a person must have to join a study, and exclusion criteria, the characteristics that keep someone out.[1] A person has to satisfy every inclusion criterion and avoid every exclusion criterion to be enrolled.

Those criteria are written into the study protocol before enrollment begins, and they exist to make sure participants are alike on the factors that matter for interpreting the result — commonly age, the disease or stage of disease being studied, general health, and prior treatment.[2] A given study record will spell out its own specific cutoffs and requirements for each of those factors; they are not the same from one study to the next.

Why Studies Narrow Their Population on Purpose

Eligibility criteria aren't red tape. Researchers set them so that the people in a study are similar enough to each other that any difference in outcome can be reasonably attributed to the treatment being tested, rather than to differences among the participants themselves.[2] A tightly defined population also makes a study safer to run: it can exclude people for whom a treatment carries known risk, or who have other conditions that would make results hard to interpret.

That same narrowing shows up differently depending on where a study sits in the research pipeline. Early-phase drug studies typically enroll a small number of healthy volunteers to check safety, while later-phase studies enroll larger, more varied groups of patients who actually have the condition being treated.[3] A vaccine study, for example, may start in adults and only move into progressively younger age groups once earlier results support it.[3] The population a study describes is a moving target, not a fixed one.

Why the Same Narrowing Limits Who the Results Describe

The flip side of a well-controlled study population is a well-defined limit on generalizability. A finding that holds in the group actually studied does not automatically hold in a different group. Field trial methodology literature gives a concrete example: results from trials conducted in high-income countries may not transfer directly to low- and middle-income countries, because of differences such as the prevalence of other infections or nutritional deficiencies in the local population.[3] The same logic applies closer to home — a study population defined by a narrow age range, a specific diagnosis definition, or a particular care setting describes that population, not everyone who shares the same general condition.

This is not a flaw in the study. It's a description of its boundaries. The practical task for a reader is to identify those boundaries and compare them to their own situation before deciding how much weight to give the results.

The Population-Comparison Checklist

Use this checklist on any study you're trying to apply to your own situation. It covers the five areas where a study's population most often differs from a reader's:

  • Age range. Find the study's stated age eligibility, as written in that specific study's protocol or registry record, and compare it to your own age or the age of the person you're asking about. A study limited to a specific adult age band has not directly tested people outside that band, even if the condition itself affects a wider age range.[2]
  • Diagnosis definition. Check exactly how the study defined the condition being studied — the diagnostic criteria, disease stage, severity threshold, or test results required to qualify. A study of “moderate” disease does not describe “mild” or “severe” disease unless it says so.
  • Medicines and prior treatment. Look for exclusion criteria or baseline descriptions covering medicines participants were taking, medicines they were required to stop, or treatments they had already tried. If you're on a medicine the study excluded, or haven't tried a treatment the study required first, the population doesn't match yours.
  • Setting. Note where and how the study was conducted — inpatient or outpatient, a particular country or health system, a specialty clinic versus general practice. A result from a specialized referral center does not necessarily describe what happens in routine primary care, or in a different country's health system.[3]
  • Missing details. If a summary or news article doesn't state the age range, diagnosis definition, medicines, or setting, treat the comparison as unverified rather than assuming it's close enough. The original study record or published methods section is the place to look; when that information isn't available, the honest conclusion is that the match can't be confirmed.

If a study's population differs from yours on more than one of these points, treat its results as informative background rather than a direct answer to your specific question, and bring the study to a clinician or researcher who can help interpret it in your context.

Where to Find a Study's Eligibility Criteria

Every study registered with ClinicalTrials.gov has its eligibility criteria drawn from the study protocol and published as part of that study's public record. If you're starting from a news article, abstract, or press release instead of the registry record itself, the registration number (often written as an NCT number) is the fastest way to find it; our companion guide on matching a study to its registry record walks through that process. Once you're at the full study record, the eligibility criteria are stated in the researchers' own words, which is more reliable than a secondhand summary.

What This Guide Is Not

This article is a plain-language explanation of how eligibility criteria work and a checklist for comparing a study's population to your own situation. It is not medical advice, and it is not a substitute for reading a study's full published methods or eligibility criteria yourself. It does not tell you whether any specific treatment is appropriate for you. If you're weighing a treatment decision based on a clinical study, bring the study — and this checklist — to a qualified clinician or researcher who can evaluate it alongside your medical history.

Related Reading on This Site

For more on how this site researches and fact-checks its guides, see our Editorial Standards and About pages.

Sources

  1. National Center for Advancing Translational Sciences (NCATS), National Institutes of Health. Toolkit for Clinical Research glossary: “Inclusion Criteria” and “Exclusion Criteria.”
  2. National Center for Advancing Translational Sciences (NCATS), National Institutes of Health. “Eligibility Criteria,” patient registries glossary.
  3. Smith, P.G., Morrow, R.H., Ross, D.A. (eds.) “Types of Intervention and Their Development,” in Field Trials of Health Interventions: A Toolbox, 3rd edition. National Center for Biotechnology Information, National Library of Medicine.