Understanding Research

Plain-language definitions for the evidence, regulatory, guideline, and source-integrity terms used throughout this site — matching exactly what each status label and warning actually means.

How cancer research typically progresses

Research does not always move through every step in order — some steps are skipped, repeated, or run in parallel, and a promising early result does not guarantee it will reach later stages. This is a general orientation, not a prediction about any specific discovery.

Laboratory researchStudies in cells or tissue samples, not yet in a living organism.
Preclinical (animal) researchStudies in animal models, before any human is involved.
Early human researchThe first small studies in people, often focused on safety.
Phase 1 trialA small trial testing safety and dosing in people.
Phase 2 trialA larger trial testing whether a treatment appears to work and is reasonably safe.
Phase 3 trialA large trial comparing a treatment against the current standard, often across many sites.
Regulatory reviewA regulator (such as the FDA) reviews trial results to decide whether to approve a treatment for a specific, stated use.
Clinical / guideline contextClinical guideline bodies decide whether and how an approved treatment fits into standard care.

Evidence

Preclinical: Findings from laboratory or animal research only — no human data yet. The earliest, least certain evidence stage.
Early Human: The first human data exists (for example, a small early-phase trial), but it is not yet enough to draw confident conclusions.
Emerging Clinical: Human clinical evidence is building, but it is not yet as large, replicated, or definitive as later-stage evidence.
Later-Stage Human: Larger, more advanced human studies — evidence is more mature, though not necessarily the highest-certainty tier.
High-Certainty Evidence: The strongest evidence stage this site labels — typically large, well-designed, and/or replicated human evidence. This describes evidence strength only; it is never the same as regulatory approval or a guideline recommendation.

Study concepts

Observational study: Researchers observe outcomes in people without assigning who gets which treatment. This can show an association, but it cannot by itself prove that the treatment caused the outcome. Stories built on this design carry a required non-causation note.
Randomized controlled trial: Participants are randomly assigned to receive the treatment or a comparator, which helps isolate the treatment's actual effect from other differences between groups.
Sample size: The number of participants a study included. A small sample size means a finding is less certain and more likely to be affected by chance.
Comparator: What the treatment was compared against — for example, a placebo, an existing standard treatment, or no treatment. Without a comparator, it is hard to know whether a treatment truly outperforms the alternative.
Primary outcome: The main result a study was designed to measure — the question the study set out to answer.
Surrogate endpoint: A stand-in measurement (such as a lab value or tumor-shrinkage marker) used in place of directly measuring what patients actually care about, like living longer or feeling better. A surrogate endpoint may not translate into real-world benefit, so stories that rely on one carry a required limitation warning.
Absolute effect: The plain difference in outcome rates between groups (for example, "5 in 100 people" versus "3 in 100 people"). Absolute effects give a clearer sense of real-world impact than relative effects alone.
Relative effect: The proportional difference between groups (for example, "a 40% reduction"). Relative effects can sound large even when the absolute, real-world difference is small — this site reports both where available.
Confidence interval / uncertainty: A range showing how precise a result is. A narrow range means a more precise estimate; a wide range means more uncertainty about the true effect.
Replication: Whether independent researchers, using their own data, have reproduced a finding. A result that has not yet been replicated is less certain than one that has.
Follow-up: How long participants were tracked after treatment. Short follow-up can miss delayed benefits or delayed harms.
Harms: Side effects or adverse events observed in a study — reported alongside benefit, never omitted.
Discontinuations: How many participants stopped treatment early, and why (for example, due to side effects) — a signal of real-world tolerability.

Publication status

Peer review: Independent experts examined the study before it was published, checking its methods and conclusions. Peer review does not guarantee a finding is correct, but it is a meaningful quality check.
Preprint (not yet peer-reviewed): A study posted publicly before undergoing peer review. Preprints can share findings faster, but they have not yet had independent expert scrutiny — this site labels them as not yet peer-reviewed and requires an editor to note that limitation.
Conference abstract: A brief early summary presented at a scientific conference, usually without the full peer-reviewed detail of a completed paper. Findings can change by the time (or if) a full paper is published.

Regulatory status

Investigational: The treatment has not received regulatory approval for the stated use and is still being studied. It is not available as standard care.
Clinical-trial access: The treatment is available through enrollment in a clinical trial. Trials have strict eligibility criteria, and eligibility never guarantees enrollment or a spot in the treatment arm.
Expanded access: Also called "compassionate use" — a regulatory pathway that can let some patients outside a clinical trial access an investigational treatment under specific conditions, before it is fully approved.
FDA-approved for the exact stated cancer/use: A regulator (in the United States, the FDA) has approved the treatment for the exact cancer type and use stated in the record — not for every cancer, every patient, or every use. Approval is a regulatory decision; it is shown separately from evidence strength and separately from guideline recommendations, and it never implies either one.
Withdrawn / suspended / recalled: The treatment's approval or authorization for this use was withdrawn, suspended, or recalled by the regulator.

Care & guidelines

Not addressed: Clinical guidelines do not yet address this specific use — this is not the same as a negative recommendation; it simply has not been evaluated yet.
No recommendation: Guidelines considered this use but did not issue a specific recommendation for or against it.
Recommended option: Clinical guidelines list this as one reasonable option among others for the stated use — not necessarily the preferred one.
Preferred / standard care: Clinical guidelines identify this as the preferred or standard approach for the stated use.
Recommended against: Clinical guidelines specifically recommend against this use.

Precision medicine

Biomarker: A measurable biological characteristic — for example, a gene mutation or protein level — used to inform decisions about diagnosis, treatment, or monitoring. When a story states a biomarker requirement, it means the finding applies specifically to patients with that biomarker, not to everyone with the cancer type.
Predictive biomarker: A general research concept: a biomarker that helps predict whether a specific treatment is likely to work for a given patient. This site does not currently label individual biomarkers as predictive versus prognostic — a story's biomarker requirement field states what applies, in the source's own terms.
Prognostic biomarker: A general research concept: a biomarker associated with a patient's likely disease course regardless of treatment choice — distinct from a predictive biomarker, which is specifically about treatment response.
Companion diagnostic: A test used alongside a treatment to determine whether a patient has the biomarker that treatment requires.

Source integrity

Primary source: The actual study, trial registry, regulator record, or clinical guideline a story is based on — the primary evidence supporting the story.
Secondary source: Additional context, such as a press release or news article, that is not itself the underlying evidence. Secondary sources are shown separately from primary evidence and are never treated as proof on their own.
Load-bearing source: A primary source whose integrity a story materially depends on. If a load-bearing source is retracted or receives an expression of concern, the story's evidence label is suppressed pending editorial re-review — the site never keeps asserting a strength that depends on a discredited source.
Correction issued: A published fix to a source's content. Corrections are recorded and shown on the story and in the site-wide Recently Updated / Corrections history.
Expression of concern: A formal notice from a journal flagging unresolved concerns about a source, short of a full retraction. On a load-bearing source, this pauses verification pending review.
Retracted: A source formally withdrawn by its publisher, usually because of a serious error or integrity issue. A retracted load-bearing source blocks a story from being verified and suppresses its evidence label.