Case Reports, Case Series, and Retrospective Chart Reviews

A detailed patient history can reveal an unexpected event, unusual presentation, measurement problem, or hypothesis worth testing. A collection of clinical records can show patterns that a single observation cannot. Neither format automatically supplies the comparison needed to estimate what would have happened without the exposure.[5][6]

The most useful first step is to reconstruct the design from case selection, timing, comparator, and analysis. A report’s label may be incomplete or inconsistent with what the investigators actually did.

Separate six commonly blended designs

ClinicalTrials.gov distinguishes a case-only observational model from cohort and case-control models, and separately records whether observations are retrospective, prospective, or cross-sectional.[3] A retrospective chart review could support a cohort, case-control, case series, or descriptive design. The chart is the source; the sampling logic defines the study.

Chronology can suggest a hypothesis, not complete the counterfactual

If an outcome follows an exposure, the timeline establishes temporal order. That can be essential for a causal hypothesis, but it does not show that the exposure caused the outcome. The same sequence may reflect natural recovery, disease fluctuation, concurrent care, behavior change, measurement timing, expectation, or another event.

A striking before-after change is also vulnerable to regression to the mean. People often seek care or receive intensified attention when symptoms are unusually severe. Even without an effective intervention, a later value may move closer to the person’s usual level.

For causal interpretation, ask: what would likely have happened to comparable people at the same point in their course without the exposure? An uncontrolled report does not observe that counterfactual. It can generate a testable question, document feasibility, identify signals, or describe an uncommon course without estimating comparative benefit.

Case selection determines what population is represented

A case series should say whether cases were consecutive, complete, convenience-selected, referred, or chosen because their outcomes were memorable. The eligibility definition and inclusion window should be set before results are selected whenever possible.

Selection can occur at several stages:

• Referral bias: a specialist setting receives unusual, severe, or treatment-resistant cases.
• Survivor bias: inclusion requires remaining in care or alive long enough to be recorded.
• Ascertainment bias: cases with more testing have more opportunities for an outcome to be detected.
• Publication bias: remarkable or favorable cases are more likely to be written up or accepted.
• Availability bias: missing, inaccessible, or sparse charts are omitted in ways related to the result.

The JBI case-series appraisal resource asks reviewers to consider clear inclusion criteria, valid condition identification, consecutive and complete inclusion, reporting of setting and demographics, and appropriate analysis.[2][4] Those questions support appraisal; they do not transform selected cases into a randomized comparison.

The denominator controls which claims are possible

“Three adverse events were described” is a count. A safety rate requires a defensible denominator, a defined observation period, and a process capable of detecting events in everyone at risk. The number of published reports is not the number exposed, and spontaneous documentation is not systematic surveillance.

Likewise, a case series assembled because all included people improved cannot estimate the proportion who would improve among all eligible people. Readers need:

1. the source population;
2. the number screened or potentially eligible;
3. explicit inclusion and exclusion criteria;
4. the inclusion dates and setting;
5. whether cases were consecutive or complete;
6. reasons for missing or excluded records; and
7. the final denominator for each outcome and time point.

When a denominator is absent, the conclusion should stay descriptive.

Chart abstraction can introduce information bias

Clinical records are written for care, not necessarily for research. Symptoms may be recorded only when volunteered, outcome scales may change, time points may be irregular, and clinicians may document differently.

A credible chart review defines the variables before abstraction, trains reviewers, pilots the form, and reports how disagreements and ambiguous entries were handled. Where judgment is required, masking an abstractor to the study hypothesis or exposure can reduce expectation effects. Reliability checks between abstractors can show consistency, although consistent extraction does not validate an inaccurate source record.[8]

Important missingness questions include:

• Was “not documented” treated incorrectly as “no”?
• Were outside-care events unavailable?
• Did follow-up intensity differ by exposure or outcome?
• Were laboratory methods, diagnostic criteria, or record systems stable?
• Did people with poor outcomes leave the system?

Absence of a charted adverse event is not evidence that harms were actively solicited, adjudicated, and absent.

Mixed exposures cannot isolate one contribution

A case may involve several interventions, changes in usual care, withdrawal of another exposure, or a combination. The chronology should record each one. If two interventions began together, the case cannot isolate the contribution of either without additional evidence.

The same boundary applies to a named combination. A response observed after components A and B does not establish that A alone caused it, that B alone caused it, or that any differently sourced mixture is comparable. Biological plausibility may help generate hypotheses, but it does not resolve the missing comparison.

Reporting checklists and appraisal tools have different jobs

The CARE guideline focuses on complete and transparent case reporting. Its checklist includes patient information, clinical findings, a timeline, diagnostic assessment, interventions, follow-up, outcomes, adverse events, discussion of strengths and limitations, patient perspective, and informed consent.[1]

JBI supplies critical-appraisal tools for several designs, including case series and cohort studies.[2] CARE is primarily a reporting guide; JBI tools structure appraisal. Neither is a regulatory approval, a causal-certification system, or a substitute for a comparator.

FDA’s non-interventional-study resource is a March 2024 Draft Level 1 Guidance, marked not for implementation. It discusses prespecification, data fitness, confounding, selection, measurement, missingness, and analytic transparency for potential regulatory use. Its draft recommendations should not be described as binding or final.[7]

A disciplined extraction template

For each uncontrolled report, record:

1. the actual design and data source;
2. setting, source population, and case-selection method;
3. diagnostic criteria and baseline state;
4. exposure identity and chronology, without converting historical facts into use guidance;
5. co-interventions and other major events;
6. outcome definition, assessor, timing, and measurement method;
7. missing records and follow-up losses;
8. the denominator for every count or proportion;
9. adverse-event ascertainment and adjudication;
10. alternative explanations and prespecified versus exploratory analyses.

The conclusion should match what the design observed: a clinical course, a pattern among selected cases, an estimated occurrence in a defined uncontrolled cohort, or an exploratory association. Comparative efficacy language requires a comparative design.

Important limitations

• Temporal order is necessary for many causal claims but is not sufficient.
• A detailed case can support a hypothesis without estimating an average effect.
• A larger case series remains uncontrolled unless it contains a valid comparator.
• Retrospective describes the observation period relative to participant selection and enrollment; it does not uniquely identify the design.
• “No adverse events reported” is not the same as systematic evidence of no adverse events.
• Before-after improvement may reflect natural history, regression to the mean, measurement, or co-interventions.
• CARE and JBI promote reporting or appraisal; checklist completion does not remove bias.
• A report cannot authenticate or establish comparability with a separately sourced material.

Frequently asked questions

What is the difference between a case report and a case series?

A case report describes one person or episode. A case series describes multiple cases sharing defined features. Both are ordinarily uncontrolled, although the cases may be organized and followed systematically.

Is every retrospective chart review a case series?

No. A chart review is a way of obtaining previously recorded data. Depending on how people are sampled and compared, it may be a cohort, case-control study, case series, before-after study, or descriptive audit.

Can one dramatic recovery prove that an intervention worked?

No. It documents a temporal clinical course and may generate a strong hypothesis. Without a suitable counterfactual, natural history, co-interventions, measurement changes, and regression to the mean remain possible explanations.

Can a case series estimate incidence?

Only when it arises from a defined population at risk with complete case ascertainment and a valid time denominator. A collection of published or referred cases ordinarily cannot estimate incidence.

Does no reported adverse event mean the exposure was safe?

No. The report may not have systematically solicited, defined, followed, or adjudicated harms. Safety interpretation requires the ascertainment method, observation window, missingness, and denominator.

What does consecutive case inclusion improve?

It reduces investigator discretion to select only memorable or favorable cases and clarifies the population represented. It does not create randomization or eliminate referral, measurement, or confounding concerns.

Does completing a CARE or JBI checklist validate causality?

No. CARE supports transparent case reporting, and JBI supports structured appraisal. Neither checklist supplies a missing comparator or certifies a causal estimate.

Primary and official sources

1. EQUATOR Network: CARE Case-Report Reporting Guideline
2. JBI Critical Appraisal Tools
3. ClinicalTrials.gov Protocol Registration Data Element Definitions
4. Munn et al.: JBI Critical Appraisal Tool for Case Series
5. Cochrane Handbook Chapter 24: Non-Randomized Studies of Intervention Effects
6. Murad et al.: Methodological Quality and Synthesis of Case Series and Case Reports
7. FDA Draft Guidance: Non-Interventional Studies for Drug and Biological Products
8. Vassar and Holzmann: Retrospective Chart Review Methodology