Research methodology
Last reviewed 2026-11-08 · Sources cited for every value
Source hierarchy
- 1
Official prescribing information
Current FDA prescribing information (section 12.3) from DailyMed, or another regulator’s product information when no US label exists.
- 2
Peer-reviewed human pharmacokinetic study
Human studies that measured the compound in plasma and report a half-life with their design and population.
- 3
Registered clinical trial results
Results posted to a trial registry, or regulator reviews of trial data, with enough detail to know what was measured.
- 4
Other primary research
Other primary research, such as FDA reviewer documents summarizing an application’s pharmacology.
- 5
Nonclinical study
Animal or laboratory studies, always labelled with the species and never shown as a human value.
Secondary sources (reviews, databases, vendor pages, encyclopedias) are never used as a value's source, even when they repeat a primary figure. If the primary source can't be read, the value isn't published.
Conflicting estimates and uncertainty
Never averaged
Different labels, populations and studies are separate records. Where they disagree, a note says so (for example, Mounjaro vs Zepbound, or two tesamorelin formulations).
Ranges and bounds as reported
A reported range is drawn as a band around its midpoint. A value given only as an upper bound (“less than 3 hours”) is shown as text and never drawn.
Routes and formulations matter
The same ingredient can behave differently by product: oral and injected semaglutide share a half-life but not absorption. Values are never carried across routes, formulations or species.
Missing data
When no qualified human value exists, the page says “Half-life not reliably established” and explains the gap. Numbers are never filled in by estimation or AI.
Evidence quality
Each record carries a quality grade based on how much the value can be leaned on:
- High — regulatory label values from a full clinical pharmacology program.
- Moderate — label values with limited detail, controlled human studies of reasonable size, or regulator reviews.
- Low — small studies, values read from an abstract only, or animal data.
- Very low — reserved for single, poorly reported measurements; not currently used.
Editorial review checks
- The value, unit and range match the source exactly, and the source’s own term is recorded.
- Species, population, route, formulation and single- or repeated-administration context are recorded from the source.
- The source has a stable identifier (DailyMed SetID, PMID, DOI or NCT number) on a regulatory or scholarly host.
- Abstract-only readings are flagged in the record notes.
- No text quotes an amount or a frequency of administration; automated tests block it.
- The record is marked reviewed with a verification date; drafts never reach a page.
Values are re-checked when a label is revised and at least every six months; regulatory status is re-checked at every review. Each record shows its last verification date, and every value is listed on the sources page.
The mathematical model
Charts use R(t) = 100 × 2^(−t / t½), the idealized percentage of a starting reference amount remaining after time t under single-compartment, first-order elimination. When a source gives a range, the curve uses its midpoint and a band shows the range.
The curve is an idealized first-order elimination model. It is a mathematical illustration of what a half-life means, not a medical prediction. It does not model:
- Patient-specific pharmacokinetics
- Drug absorption from the site of administration
- Multi-compartment distribution
- Active metabolites
- Receptor occupancy
- How long a clinical effect lasts
- Individual blood concentrations
- Drug interactions
- Changes in kidney or liver function