Skip to main content
Paper Buddy

Reporting guidelines · manuscript writing

How to Write a Medical Abstract: Structure, Examples and Word Limits

Write an English abstract for a medical journal by choosing the reporting framework for your study, checking the journal’s headings and word limit, and comparing the finished text with your manuscript. This guide brings together four study-design templates, original English examples, linked journal requirements and a practical consistency check.

By the Paper Buddy team · Sources checked:

The Paper Buddy team wrote these explanations, templates and fictional examples. They are teaching materials, not official journal forms or clinical evidence. The English templates and example data are identical in all three language versions.

In shortChoose your study design and target journal first. Shorten repetition while preserving the study population, outcomes, uncertainty and important harms. Check every number and conclusion against the final manuscript.

Structured and unstructured abstracts

A structured abstract uses labeled sections to help readers find the study’s purpose, methods, results and conclusion. Background, Methods, Results and Conclusions form a useful starting layout, but journals use different headings. An unstructured abstract conveys the relevant information in continuous prose.[2]

ICMJE calls for structured abstracts for original research, systematic reviews and meta-analyses. It also asks authors to keep the abstract consistent with the article and avoid overstating the findings. A reporting checklist helps select content; it does not override the target journal’s format.[1]

Once the main text and tables are stable, draft the results first, add the methods needed to interpret them, and finish the objective and conclusion. This writing order is our practical suggestion. Recheck the abstract whenever the analysis or manuscript changes.

If the body is not ready yet, start with the whole-manuscript drafting sequence →

Journal versus conference abstracts →

Background / Objective
Name the gap and the specific study question. A short opening usually leaves more room for methods and results.
Methods
Identify the design, setting, eligibility, intervention or exposure, comparator, primary outcome and time point, and essential analysis. Preserve the details needed to interpret the result.
Results
Report the analyzed population, group-specific denominators, primary outcome, effect size and uncertainty. Distinguish randomized, followed and analyzed participants when they differ.
Conclusions
Answer the stated question within the study’s design and uncertainty. Retain a limitation that materially changes interpretation; do not replace a null primary result with a favorable secondary finding.

Adapt these content blocks to the journal’s exact headings. For example, JAMA separates Importance and Objective and asks for more detailed methods headings. Trial registration and funding also need the location required by the journal; ICMJE places direct funding information separately after the abstract. [1][8]

Journal word limits and headings

Abstract limits depend on both the journal and article type. These examples were checked on October 27, 2026; they are a reference table rather than a complete journal directory.

Abstract limits depend on both the journal and article type. These examples were checked on October 27, 2026; they are a reference table rather than a complete journal directory.
Journal · article typeMaximumRequired headings · notesOfficial source
NEJMOriginal research250 wordsBackground / Methods / Results / Conclusions

Original Article. Add trial registration information when applicable; abbreviations are strongly discouraged.

Official source ↗
JAMAOriginal research350 wordsImportance / Objective / Design / Setting / Participants / Intervention(s) or Exposure(s) / Main Outcome(s) and Measure(s) / Results / Conclusions and Relevance

Reports of Original Data. Trial Registration applies to clinical trials. Avoid abbreviations in the abstract; do not report P values without the comparison data.

Official source ↗
Internal MedicineOriginal research250 wordsObjective / Methods / Patients or Materials / Results / Conclusion

The original-article and case-report limits differ. The case-report instruction specifies 100 words without prescribing the original-article headings.

Official source ↗
Internal MedicineCase report100 wordsNo specific headings stated in the case-report instruction.

The original-article and case-report limits differ. The case-report instruction specifies 100 words without prescribing the original-article headings.

Official source ↗
Japanese Journal of OphthalmologyClinical or laboratory investigation250 wordsPurpose / Study Design / Methods / Results / Conclusion

Define abbreviations at first use in the abstract and again in the text. Do not include references in the abstract.

Official source ↗
Toho Journal of MedicineOriginal research250 wordsIntroduction / Methods / Results / Conclusions

Original research article. Define abbreviations at first use and provide trial registration details when applicable.

Official source ↗
  • Use the journal’s current article-type instructions, not another journal’s policy or the limit for a review or letter. JAMA’s 350-word entry was checked on JAMA’s own author page.
  • Check the submission form’s counting rules, including headings, separately. These source checks do not establish one shared counting convention.
  • For brevity, this table covers selected original articles and one case-report example. Other article types can have different limits or no abstract.

Templates and examples by study design

Choose a design below, then adapt the headings and length to the journal. Brackets mark facts you must supply. Every worked example is fictional, including the trial and review registration statements. Never copy example results into a real manuscript. These are original teaching templates, not official checklist forms.

Original research: observational cohort

Use STROBE items 1a–1b to identify the design and summarize the work fairly. This template is specifically for a cohort. In a case-control study, sampled case/control counts do not by themselves estimate population risk; in a cross-sectional study, keep prevalence separate from incidence. Choose effect measures and wording appropriate to the actual design. [3]

Adaptable English template

Background: [Knowledge gap]. We examined the association between [exposure] and [outcome].
Methods: This [prospective/retrospective] cohort included [population, setting, dates]. The primary outcome was [definition and time point]. [Analysis] adjusted for [prespecified confounders].
Results: Among [N] participants, [group-specific n/N and percentages]. The [adjusted effect measure] was [estimate] (95% confidence interval, [limits]).
Conclusions: [Exposure] was associated with [outcome], with [uncertainty and design limitation].

Worked example — fictional data · 119 words

Background: Whether a discharge telephone call is associated with fewer emergency visits is uncertain. We examined this association in adults discharged from medical wards.
Methods: This retrospective cohort included 240 adults discharged from two hospitals during 2025. The exposure was a nurse telephone call within 48 hours. The primary outcome was an emergency visit within 30 days. A modified Poisson model adjusted for age, comorbidity and previous emergency visits.
Results: Emergency visits occurred in 18 of 120 patients (15.0%) receiving a call and 30 of 120 (25.0%) receiving no call. The adjusted risk ratio was 0.62 (95% confidence interval, 0.37–1.04).
Conclusions: The point estimate favored telephone follow-up, but the confidence interval included no association. Residual confounding limits causal interpretation.

Teaching counts use whitespace-separated tokens, with hyphenated expressions counted as one token. Recheck the target submission system.

See the STROBE guide for observational studies →

Randomized controlled trial

Use CONSORT 2025 item 1b and Table 2. Keep the primary outcome prominent, identify the analysis population and report important harms even when words are scarce. The example uses a superiority framework; an inconclusive superiority result does not establish equivalence. [4]

RCT abstract content to check

  • Objective and trial design/framework
  • Eligibility, setting, interventions
  • Allocation method and blinding
  • Primary outcome and analysis
  • Numbers randomized and analyzed per group
  • Recruitment status, results and uncertainty
  • Important harms and interpretation
  • Registration and funding

These short prompts summarize content, not required abstract headings. Check each sub-item in the linked official checklist; the RCT prompts group several Table 2 items. [4]

Adaptable English template

Background: [Question and objective].
Methods: In this [design/framework] randomized trial at [setting], [eligible participants] were assigned [allocation ratio and method] to [intervention] or [comparator]. [Who was blinded]. The primary outcome was [definition/time point]; [analysis population/method].
Results: [Numbers randomized and analyzed per group; recruitment status]. [Primary outcome by group, effect estimate and uncertainty]. [Important harms by group].
Conclusions: [Benefit–harm interpretation within the trial's limits].
Trial registration: [Verified registry and identifier; date if required].
Funding: [Verified direct support, in the journal's required location].

Worked example — fictional data · 169 words

Background: We tested whether a discharge telephone call reduces emergency visits within 30 days.
Methods: In a two-hospital, parallel-group superiority trial, 300 adults discharged from medical wards were assigned 1:1 by computer-generated randomization to a nurse call within 48 hours or usual care. Patients and nurses were unblinded; outcome assessors were blinded. The primary outcome was any emergency visit within 30 days, analyzed by intention to treat as a risk difference.
Results: Recruitment ended at the planned sample size. Each group had 150 participants randomized and analyzed, with complete outcome ascertainment. Emergency visits occurred in 24 of 150 (16.0%) versus 39 of 150 (26.0%); the risk difference was −10.0 percentage points (95% confidence interval, −19.1 to −0.9). Serious adverse events occurred in 3 of 150 versus 4 of 150.
Conclusions: Telephone follow-up reduced 30-day emergency visits in this trial. The small number of serious adverse events limits conclusions about safety.
Trial registration: Fictional teaching study; no registry record exists.
Funding: No actual study or funding exists for this example.

Teaching counts use whitespace-separated tokens, with hyphenated expressions counted as one token. Recheck the target submission system.

Systematic review and meta-analysis

Use PRISMA 2020 for Abstracts. Report the last search date for each source, study and participant counts, synthesis methods and uncertainty. If no meta-analysis was performed, describe the actual synthesis and its findings instead of inventing a pooled estimate. [5]

PRISMA for Abstracts: 12 content checks

  1. Title: identify a systematic review
  2. Objective
  3. Eligibility
  4. Sources and last search dates
  5. Bias assessment
  6. Synthesis methods
  7. Included studies and participants
  8. Main findings and uncertainty
  9. Evidence limitations
  10. Interpretation
  11. Funding
  12. Registration

These short prompts summarize content, not required abstract headings. Check each sub-item in the linked official checklist; the RCT prompts group several Table 2 items. [5]

Adaptable English template

Title: [Topic]: a systematic review [and meta-analysis, if performed]
Background: We assessed [review question].
Methods: Eligible studies were [design/population/intervention/comparator/outcomes]. We searched [sources and each last search date]. Risk of bias was assessed using [tool]; results were synthesized using [method].
Results: [Number of studies/participants and characteristics]. [Studies/participants for main outcome; pooled estimate and uncertainty, if applicable, with direction].
Conclusions: [Interpretation and limitations of the evidence].
Funding: [Primary funding source].
Registration: [Verified register and identifier, or actual registration status].

Worked example — fictional data · 148 words

Title: Discharge telephone follow-up and emergency visits: a systematic review and meta-analysis
Background: We assessed whether discharge telephone follow-up reduces emergency visits among adults.
Methods: We included randomized trials comparing a nurse call within 48 hours of medical-ward discharge with usual care. MEDLINE, Embase and CENTRAL were each searched through September 30, 2026. We assessed bias using RoB 2 and pooled 30-day emergency-visit risk ratios using a random-effects model.
Results: Six trials included 1800 adults. All six contributed to the primary outcome. The pooled risk ratio was 0.81 (95% confidence interval, 0.65–1.01; values below 1 favor telephone follow-up). Four trials had some concerns and two had high risk of bias.
Conclusions: The estimate favored telephone follow-up, but the interval included no effect. Risk of bias and imprecision limit the evidence.
Funding: No actual review or funding exists for this example.
Registration: Fictional teaching review; no registry record exists.

Teaching counts use whitespace-separated tokens, with hyphenated expressions counted as one token. Recheck the target submission system.

Case report

CARE item 3 covers why the case matters, key presentation, diagnosis/intervention/outcome and the lesson. Use structured or continuous prose as the journal requires. If headings are needed, Introduction / Case presentation / Conclusions is one possible layout, not a universal requirement. A single case does not establish treatment effectiveness. [6]

Adaptable English template

[What makes this case informative]. A [age]-year-old [patient] presented with [main concern and findings]. [Evaluation] identified [diagnosis]. After [documented intervention], [outcome] was observed at [follow-up]. This case illustrates [bounded lesson and relevant limitation].

Worked example — fictional data · 74 words

Medication reconciliation may clarify recurrent dizziness after discharge. A 72-year-old woman presented with postural dizziness one week after hospitalization. Review of her medication containers identified duplicate prescriptions for the same antihypertensive drug. After the duplicate was discontinued, dizziness resolved by the two-week follow-up, with no further falls reported. This case illustrates the value of checking actual medication use when evaluating symptoms after discharge. Symptom improvement after a medication change does not alone establish causation.

Teaching counts use whitespace-separated tokens, with hyphenated expressions counted as one token. Recheck the target submission system.

Write the full case report with the CARE guide →

Shorten without losing meaning

Remove repetition and wordy phrasing first. The aim is to meet the limit while keeping the information a reader needs to judge the study.

Shorten a complete abstract: 294 → 132 words →

  1. Reduce background to the gap and objective; remove a general textbook introduction.
  2. Keep the design, relevant population and setting, comparator, outcome/time point and essential analysis. Compress wording rather than deleting a method that changes interpretation.
  3. Prioritize the primary outcome. Keep denominators, effect estimates, uncertainty and important harms; select secondary results by relevance to the stated question.
  4. Remove repeated numbers from the conclusion when the result is already clear. Preserve a limitation or cautious interpretation that affects the main claim.
  5. Count again, then compare the shortened text with the source paragraph and manuscript. If required facts cannot fit, revise the allocation of space rather than dropping facts silently.

Small wording changes that can help

  • in order to → to
  • a total of 240 patients → 240 patients
  • It was found that visits decreased → Visits decreased

Abbreviations are not a universal way to save words. JAMA instructs authors to avoid them in abstracts; NEJM strongly discourages them. JJO permits defined abbreviations and requires definitions again in the main text. Follow the target journal’s rule. [7][8][10]

Before · 58 words

A total of 240 patients were included in this retrospective cohort study. Emergency visits within 30 days occurred in 18 of 120 patients (15.0%) who received a telephone call and in 30 of 120 patients (25.0%) who did not receive a telephone call. The adjusted risk ratio was 0.62 (95% confidence interval, 0.37–1.04). Residual confounding limits causal interpretation.

After · 42 words

This retrospective cohort included 240 patients. Within 30 days, emergency visits occurred in 18 of 120 (15.0%) receiving a telephone call versus 30 of 120 (25.0%) receiving no call (adjusted risk ratio, 0.62; 95% confidence interval, 0.37–1.04). Residual confounding limits causal interpretation.

Both versions retain 240 patients, the retrospective design, 30 days, both group denominators and percentages, the adjusted risk ratio and its confidence interval, and the confounding limitation. The interval still includes 1; shortening does not make this a definitive benefit.

Teaching counts use whitespace-separated tokens, with hyphenated expressions counted as one token. Recheck the target submission system.

Check journal AI-disclosure requirements →

Shorten a complete abstract: 294 → 132 words

This exercise uses a 250-word ceiling. The study and numbers are fictional examples written by our team, not evidence of clinical effectiveness. You need not fill a limit, but a real abstract must first include the reporting details it needs.

Before · 294 words

Background: Follow-up attendance is an important part of outpatient care, and there is interest in understanding whether reminder messages are associated with attendance at planned appointments. The objective of this study was to describe attendance among adult patients who received a reminder and those who received the standard scheduling information at a single outpatient clinic.

Methods: We conducted a retrospective review of appointment records from January through June 2025. A total of 200 adult patients were included in the study, with one index appointment included for each patient. Of these patients, 100 had received a reminder message and 100 had received the standard scheduling information without a reminder message. The primary outcome that was assessed in this study was attendance within 30 days of the planned appointment date. Attendance was summarized using counts and percentages for each of the two groups. The analysis was descriptive, and no adjustment was made for differences in baseline characteristics between the two groups. Available records did not capture the reasons why patients missed their appointments.

Results: Attendance within 30 days was recorded for 82 of the 100 patients in the reminder group, representing 82% of that group. In the standard-information group, attendance within 30 days was recorded for 70 of the 100 patients, representing 70% of that group. The absolute difference in attendance between the two groups was therefore 12 percentage points. These figures describe the observed attendance in the records reviewed for this study.

Conclusions: In this single-clinic retrospective review, attendance was higher among patients recorded as having received a reminder. Because reminder receipt was not randomized and baseline differences were not adjusted for, these descriptive findings do not establish that reminders caused the difference in attendance. Further evaluation would be needed before drawing conclusions about effectiveness.

Shortened example · 132 words

Background: We described follow-up attendance among adults who received a reminder or standard scheduling information at one outpatient clinic.

Methods: We retrospectively reviewed January–June 2025 records for 200 adults, including one index appointment per patient. The reminder and standard-information groups each included 100 patients. The primary outcome was attendance within 30 days of the planned appointment. We summarized counts and percentages without adjusting for baseline differences. Reasons for missed appointments were unavailable.

Results: Attendance was recorded for 82/100 patients (82%) in the reminder group and 70/100 (70%) in the standard-information group, an absolute difference of 12 percentage points.

Conclusions: Attendance was higher in the reminder group in this single-clinic review. Nonrandom reminder receipt and unadjusted baseline differences preclude attributing this difference to reminders. Further evaluation is needed before drawing conclusions about effectiveness.

Counts split text on whitespace and include English headings. Submission systems may handle hyphens differently. The English examples are identical in all three language versions.

Check what was removed alongside what was retained
SectionRemoved or compressedInformation retained
BackgroundGeneral background repeating the importance of attendanceAdults at one clinic; the aim of comparing reminders with standard scheduling information
MethodsPhrases such as A total of and that was assessed in this studyRetrospective review; January–June 2025; 200 adults, 100 per group; one appointment per person; attendance within 30 days of the planned date
ResultsExpanded repetitions of denominators and percentages; a sentence reintroducing the findings82/100 (82%) versus 70/100 (70%); an absolute difference of 12 percentage points
ConclusionsRepetitive wordingOne clinic; nonrandom receipt and unadjusted analysis; no causal attribution; further evaluation needed

Easy details to lose

  • Are unavailable reasons for missed appointments still stated in Methods?
  • Did 12 percentage points remain distinct from a “12% increase”?
  • Did you avoid inventing confidence intervals, P values or adjusted analyses?
  • Have you separately checked uncertainty, harms, registration and other details required for the real study?
Shorten the abstract below to the limit for the journal and article type I specify. Preserve the design, population, unit of analysis, outcome definition and timing, main numbers and limitations. Do not add facts or numbers. Show what you deleted or compressed separately from items the author still needs to verify.

Compare the abstract with the text and tables, then review the whole manuscript. Health Check does not automatically shorten abstracts or certify compliance with submission rules.

Review my manuscript with its abstract →

Compare the abstract with the manuscript

Compare the final abstract with Methods, Results and tables using the same population, analysis and time point. Different denominators can be valid; determine what each number refers to before calling it an error. [1]

Copy this table into your document and fill in exact page/line or table locations. Record checked, needs correction, or needs clarification. Recalculate an explicit n/N percentage at the reported precision. Do not force an adjusted estimate to equal a crude estimate or an analyzed count to equal an enrolled count. Resolve the reason using the source analysis.
ItemWhat to compareAbstract location/valueText/table location/valueResult · action
Design and populationDesign name, eligibility, setting and enrollment period__________________
Sample and denominatorRandomized, followed and analyzed counts; patients versus eyes, lesions or visits__________________
Primary outcomeDefinition, measurement time and analysis population__________________
Numerical resultsn/N, percentages, effect measure, adjustment, confidence interval and units__________________
Harms and limitationsImportant adverse events, missingness and limits on interpretation__________________
Conclusion and recordsClaim strength, registration identifier and direct funding__________________

Continue with the pre-submission manuscript check →

Use Paper Buddy in this workflow

Paper Buddy provides an Abstract section for writing. Paper Chat can be used to request more concise wording from supplied abstract text. Treat this as an editing request: inspect the suggestion, recount words and compare the study facts yourself.

  1. Write the abstract from the completed manuscript and use the design template as a content check.
  2. Give Paper Chat the abstract and explicit preservation instructions when requesting a shorter version. Check the returned wording before using it.
  3. When usable journal requirements are available in submission preparation, check the abstract word count and request suggested length or structure adjustments. Review unresolved requirements yourself.
  4. Use Health Check to help identify numerical or interpretive inconsistencies in the supplied manuscript, then verify the relevant passages and tables.
Shorten the English abstract below toward the specified word limit. Preserve design, population, comparator, outcome/time point, all numbers and uncertainty, important harms and limitations. Do not strengthen the conclusion or add facts. If essential information cannot fit, explain what remains over the limit. Return the proposed text for my review.

These tools support author review. They do not establish that every current journal requirement or reporting item is satisfied, reanalyze raw data or determine acceptance. No dedicated automatic abstract-generation or guaranteed word-limit result is claimed here. AI features have plan-based usage limits.

Prepare the cover letter after the abstract is settled →

Journal versus conference abstracts

This guide covers medical-journal abstracts. A conference submission may have different headings, a character limit rather than a word limit, anonymous-review rules or restrictions on previously presented work. Use the specific meeting’s instructions; do not transfer the journal limits above to a conference form.

Frequently asked questions

Is a medical abstract always limited to 250 words?

No. The checked examples below include 250 words for an NEJM Original Article, 350 for a JAMA report of original data and 100 for an Internal Medicine case report. These are different journal and article-type requirements. Recheck the current instructions for your submission.

Do headings count toward the abstract word limit?

Do not assume that every journal or submission system counts words in the same way. Check the journal’s instructions and the counter in the actual submission form. The simple counts beside our examples split text at spaces and include headings; they do not establish a journal’s counting policy.

Does a case report need a structured abstract?

CARE item 3 allows a structured or unstructured abstract. The journal determines the required format and length. Retain the reason the case is informative, key findings, diagnosis, intervention, outcome and bounded lesson in either format.

Sources and how we checked them

Official sources checked October 27, 2026. Journal limits apply to the article types named in the table. The reporting frameworks describe information to report; the journal determines submission headings and length. These explanations are our own summaries. Recheck the current official requirements before submitting.

  1. ICMJE: Preparing a Manuscript for Submission
  2. NLM: Structured Abstracts
  3. STROBE: Official Checklists
  4. CONSORT 2025: Item 1b and Explanation and Elaboration, Table 2
  5. PRISMA 2020 for Abstracts: Official 12-item Checklist
  6. CARE 2013: Checklist, Item 3
  7. NEJM: New Manuscripts
  8. JAMA: Instructions for Authors
  9. Internal Medicine: Instructions to Authors
  10. Japanese Journal of Ophthalmology: Submission Guidelines
  11. Toho Journal of Medicine: Instructions for Authors