1. Administrative and Scope Changes
- Issuing centers: The 2003 guidance was co-issued by CDER and CBER. The 2026 draft is issued solely by CDER's Office of Clinical Pharmacology, with support from the Division of Hepatology and Nutrition.
- Scope of covered products: The 2026 draft explicitly extends to therapeutic peptides and proteins licensed under the PHS Act, and adds new discussion of altered PK for monoclonal antibodies and antibody-drug conjugates in patients with hepatic impairment — a topic absent from the 2003 version.
- Title change: "Impact on Dosing and Labeling" (2003) becomes simply "Impact on Dosage" (2026), reflecting a lighter emphasis on prescriptive label text (see below).
2. Threshold for Deciding Whether a Study Is Needed
- Hepatic elimination threshold raised: 2003 recommended a dedicated study when hepatic metabolism/excretion accounted for >20% of elimination. 2026 raises this threshold to >30%.
- Narrow therapeutic range language dropped: 2003 separately flagged narrow therapeutic range drugs as needing a study even below the elimination threshold. 2026 replaces this with a broader, less mechanistic trigger: "minimal drug concentration changes may lead to clinically significant differences in safety or effectiveness."
- New triggers added in 2026 that did not exist in 2003:
- Elimination pathways of the drug are not adequately characterized.
- The drug is intended to treat patients with liver disease.
- The drug is likely to be used in patients with hepatic impairment.
- New exemption added in 2026: locally acting drugs with limited or no systemic exposure are now explicitly listed as not warranting a dedicated study (alongside the carryover exemptions for renally-eliminated, single-dose, and gaseous/volatile drugs).
- Orphan drug carve-out (new in 2026): for diseases affecting fewer than 200,000 people, sponsors are directed to consult the relevant FDA review division early regarding timing and study design — a consideration not present in 2003.
3. Hepatic Function Assessment Tools
- NCI Criteria (new in 2026): 2026 introduces the NCI Organ Dysfunction Working Group criteria as an accepted alternative to Child-Pugh specifically for cancer patients (except hepatobiliary cancers, where CP is still preferred). The 2003 guidance relied on Child-Pugh alone, with a passing caution about confounding from cancer cachexia.
- Other scoring systems acknowledged (new in 2026): MELD and ALBI scores are explicitly discussed as useful for further stratification within CP groups B and C, though not as CP replacements. The 2003 guidance did not address these.
- Baseline stability requirement (new in 2026): labs/clinical status must be stable before enrollment; if ascites/encephalopathy scores or labs shift by more than ~30% before dosing, reassessment after about a week is recommended, with classification based on the assessment closest to PK sampling. No equivalent instruction existed in 2003.
- Exclusion of acute disease (more explicit in 2026): 2026 explicitly excludes subjects with acute liver disease or acute-on-chronic liver disease and warns against scoring CP during/shortly after acute events (alcohol recidivism, GI bleed, surgery, infection). 2003 did not spell this out.
- Etiology-specific effects (new in 2026): 2026 explicitly discusses that cholestatic vs. hepatocellular disease may differentially affect drugs undergoing enterohepatic circulation, and asks sponsors to collect/report underlying disease etiology. Not addressed in 2003.
- Pharmacogenomics (sharpened in 2026): 2026 names specific polymorphic enzymes/transporters (CYP2D6, CYP2C9, CYP2C19, OATP1B1, BCRP, OCT1) and directs genotyping at screening. 2003 only generically advised considering metabolic status for CYP2D6/2C19-metabolized drugs.
4. Study Design
- Population studied: 2003's "reduced design" centered on control vs. moderate CP impairment only, with mild impairment's dosing extrapolated from the moderate group's findings and severe impairment generally contraindicated absent further data. 2026 instead recommends studying mild, moderate, and severe groups plus control from the outset, though it allows starting with a severe-only cohort and expanding to other severities only if a significant effect is found.
- Sample size approach changed: 2003 specified fixed minimums — at least 6 subjects/arm for the full design, or at least 8 subjects each in the control and moderate arms for the reduced design. 2026 drops fixed minimums in favor of a statistical target: sizing each group prospectively to achieve a 95% CI within 60–140% of the geometric mean with at least 80% power.
- Unbound drug concentration measurement broadened: both versions require measuring unbound fraction for highly extracted (ER >0.7), highly protein-bound (fu <10%) drugs. 2026 adds a new recommendation to also consider unbound measurement for low-extraction (ER <0.3), highly protein-bound drugs, and requires archiving samples for future binding studies when a drug is not extensively protein-bound — none of which appeared in 2003.
- Population PK (popPK) approach expanded: both versions endorse popPK as an alternative to a dedicated study, but 2026 provides a much more detailed checklist (adequate subjects per HI group, consistent classification method across pooled studies, accurate dosing/sampling records, unbound and active metabolite measurement, exposure-response analysis incorporating HI as a predictor).
5. Data Analysis and Dosage Recommendations
- No-effect boundary framing loosened: 2003 established a default standard of an 80–125% confidence interval for AUC/Cmax as the no-effect boundary, while acknowledging this is hard to meet with small HI study populations. 2026 removes this specific default and instead references the 2024 ICH Drug Interaction Studies guidance for no-effect boundary methodology, alongside a 5th–95th percentile exposure-matching approach.
- Both versions retain the same core direction to report geometric/arithmetic means, variability, point estimates for ratios of impaired-to-control means, and 90% confidence intervals for total and unbound AUC/Cmax.
6. Labeling
- Prescriptive label templates removed: 2003 contained detailed, fill-in-the-blank labeling language for multiple scenarios (no effect found; effect found; no studies conducted, broken into sub-options for renal elimination, limited vs. extensive hepatic elimination, and narrow vs. wide therapeutic range). 2026 removes all of this template language and instead gives a short bullet list of the categories of information the Prescribing Information should cover (PK/PD in HI, hepatic elimination info, clinical effects, risk mitigation recommendations).
- Section structure: 2003 had detailed subsections for Clinical Pharmacology (Pharmacokinetics and Special Populations subsections with model text), Precautions/Warnings, and Dosage and Administration. 2026 consolidates this into one high-level "Labeling Recommendations" section without the granular subsection breakdown.
7. Practical Takeaway for Industry
Sponsors currently designing or already running hepatic impairment studies under the 2003 framework should watch for: the higher 30% elimination threshold (may reclassify some drugs as not needing a study, but new triggers like "likely to be used in HI patients" may pull others back in); the shift toward statistically-justified sample sizes rather than fixed minimums; the expectation to study all three severity groups rather than defaulting to moderate-only; the option to use NCI Criteria for cancer trials; and the loss of ready-made label templates, meaning labeling language will need to be developed drug-specifically in consultation with the review division rather than adapted from a standard menu of options.
The differences are generated by AI models and can contain mistakes. Please double-check the content.