A Non-Monotonic Liver Signal in a 28-Day Study: Real Effect or Artifact?
A contract laboratory runs a 28-day GLP oral gavage study in Sprague-Dawley rats at 0, 10, 50, 250 and 1000 mg/kg-day, 5/sex/group. At 50 mg/kg-day, mean liver weight is 12% above control (p<0.05) and ALT is 2.1x control (p<0.05), with no histologic change. At 250 and 1000 mg/kg-day, liver weight and ALT are indistinguishable from control. Control ALT is 58 U/L against a facility historical mean of 38 U/L (range 22-65), and the high-dose groups cluster near 30 U/L. The 50 mg/kg-day group was the last group processed on the analyzer that day. No toxicokinetic samples were taken. The sponsor proposes to declare a NOAEL of 250 mg/kg-day and move directly to a two-year bioassay.
1. Is the 50 mg/kg-day finding treatment-related? Identify the specific evidence that would move your estimate in each direction.
Read the worked answer
The default assumption of monotonicity makes a mid-dose-only effect suspicious, and several features here point to artifact rather than biology:
Against a true effect
- Non-monotonicity: a 12% liver weight rise and 2x ALT at 50 with complete normalisation at 250 and 1000 is unusual for a directly acting hepatotoxicant. A genuine low-dose-only effect would require a mechanistic explanation (e.g., a saturated detoxification pathway or a metabolite formed only at low dose) - possible, but it must be demonstrated.
- Control within historical range but near the top: 58 U/L against a range of 22-65 means the '2.1x' is largely a comparison with an unusually high control, and the high-dose groups near 30 U/L are unusually low. The ratio is inflated from both ends.
- Batch position: the 50 mg/kg-day group was analysed last; drift, reagent aging, or a calibration event can produce a group-sized shift in an automated chemistry run.
- n = 5: with that many comparisons across endpoints and sexes, one p<0.05 is expected by chance.
- No histopathology: a 2x ALT without any morphological correlate is weak evidence of injury.
For a true effect
- A consistent sex pattern, a dose-related trend in individual (not just group-mean) values, and replication in a second study.
Judgment: obtain individual animal data, the analyzer QC log, and a re-run of blinded samples in randomised order. Until then the finding is 'unexplained,' not 'adverse' - and not dismissible either. The 12% liver weight change is at the threshold usually considered adaptive.
2. What must be done before the two-year study is designed, and what could go wrong at each dose level of the chronic study if you proceed on the current data?
Read the worked answer
Missing information that the chronic study design depends on:
- Toxicokinetics. Without TK there is no evidence that the 1000 mg/kg-day animals were actually exposed. If absorption saturates or the compound induces its own metabolism, the top dose may be functionally lower than the mid dose - which would itself produce an apparent non-monotonic response and would invalidate dose selection. TK at all doses, ideally with a metabolite, is the single highest-value addition.
- Duration. Twenty-eight days is too short to distinguish adaptation from injury; a 90-day study with a recovery arm resolves both the ALT question and whether liver weight changes are reversible.
- Histopathology. Special stains and a peer review of livers, plus a proper blinded re-read.
- Chemistry conduct. Repeat the clinical chemistry with groups randomised across runs and with a low and high control material.
- Statistical approach. Pre-specify trend testing and use historical control data rather than pairwise p-values alone.
Consequences of getting it wrong: if 50 is truly a LOAEL and you set the top dose at 250, the chronic study may never characterise the low-dose region; if 50 is an artifact and you set the NOAEL at 10, you drive an unnecessarily low chronic top dose, risk an inadequate exposure margin, and impose avoidable regulatory burden. Dose selection for a two-year bioassay is the decision with the highest downstream cost, which is why the sponsor's leap from a 28-day study to a two-year study should not be accepted.
3. The sponsor argues that because there was no dose-response, the 50 mg/kg-day result can be ignored. Evaluate that argument.
Read the worked answer
The argument conflates 'no monotonic dose-response' with 'no evidence of effect,' and those are different claims.
Where the sponsor is right: a single mid-dose finding with no trend, no histology, a high control, and an analytic confounder does not meet the usual burden for calling a finding treatment-related. Absence of dose-dependence is a legitimate reason to doubt causality, and regulators do accept that a non-reproducible, non-dose-related, statistically marginal effect is not a valid basis for a LOAEL.
Where the argument fails:
- Non-monotonic dose-response is not the same as no dose-response. The correct response is to ask whether a mechanism exists (saturation, induction, metabolite-specific toxicity, cytotoxicity with compensatory proliferation), not to dismiss the point.
- Ignoring the artifact explanation cuts both ways. The sponsor is willing to accept an uninvestigated mid-dose effect as noise while accepting the 250 and 1000 mg/kg data - obtained in the same uncontrolled run - as clean.
- Regulatory defensibility: a NOAEL justified as 'not dose-related' will be challenged unless the underlying data are re-examined; the record must show the re-analysis was done.
Defensible position: the NOAEL is provisional. Re-run the chemistry blinded and randomised, add TK and a 90-day study with recovery, and then document either that the 50 mg/kg-day finding did not replicate (NOAEL 250 is supportable) or that it did (NOAEL is 10). Either conclusion is acceptable; an unexamined one is not.