DABT common mistakes
DABT · Understand the traps and how to avoid them.
Reads an LD50 (or LC50) as a fixed, transferable property of a chemical and extrapolates it directly across routes, species, or dosing durations without adjusting for vehicle, formulation, or exposure regimen.
LD50 is the most memorizable number in toxicology, so candidates latch onto a single value per agent and treat it as a constant rather than a conditional statistic that depends on species, route, vehicle, particle size, and observation window.
What to do instead
For any potency question, first write down the four qualifiers - species, route, vehicle/formulation, and duration - attached to the value, and explicitly state how changing any of them shifts the dose-response curve (e.g., oral vs inhalation, rat vs human, single vs repeated dosing).
Applies the '4–5 half-lives to steady state' and first-order decay equations to compounds that show saturable elimination at relevant doses (ethanol, phenytoin, salicylates at high dose), producing a falsely short predicted accumulation time.
First-order kinetics is taught first and rehearsed most, so candidates default to exponential equations whenever a half-life is given, without checking whether the dose sits near or above Km.
What to do instead
Before using any half-life formula, ask whether the scenario states or implies dose-dependent (Michaelis-Menten) clearance - look for disproportionate increases in plasma concentration with dose, or half-life that lengthens as dose rises - and then reason qualitatively about zero-order behavior (accumulation far exceeds prediction, time to steady state is dose-dependent).
Assumes all Phase I/Phase II metabolism is detoxication and predicts lower toxicity for the metabolite, missing bioactivation pathways such as acetaminophen to NAPQI (CYP2E1/3A4), methanol to formic acid, or benzo[a]pyrene to diol epoxides.
The textbook framing of metabolism as 'detoxification and excretion' is reinforced by clearance-based thinking, and the exceptions are memorized as isolated facts rather than as a default alternative hypothesis.
What to do instead
For every metabolism question, run a two-branch check: does the metabolite have equal or lower reactivity, or is it an electrophile/reactive intermediate? Name the specific enzyme and the specific toxic metabolite, and note the protective role of glutathione conjugation and its saturation.
Labels any serum ALT/AST elevation as 'hepatocellular injury' without computing the R value (ALT/ALP ratio, expressed as multiples of the upper limit of normal) to separate hepatocellular, cholestatic, and mixed patterns.
ALT is the familiar 'liver enzyme', so a rise is taken as generic liver damage; candidates rarely operationalize the pattern classification that drives mechanism and prognosis.
What to do instead
Always compute R = (ALT ÷ ULN) / (ALP ÷ ULN): >5 hepatocellular, <2 cholestatic, 2–5 mixed; then link the pattern to the mechanism (e.g., zone 3 bioactivation injury vs canalicular transport inhibition) rather than naming one generic mechanism.
Explains dose-dependent target-organ toxicity using administered dose only, ignoring whether the effect tracks peak concentration (Cmax) or total exposure (AUC), and ignoring tissue-specific dose metrics.
Dose-response is taught in terms of mg/kg, and candidates carry that framing into situations where the toxicologically relevant metric is a concentration-time profile or a target-tissue burden.
What to do instead
For each toxicity, state the governing metric - peak-driven effects (e.g., aminoglycoside ototoxicity, CNS depression) versus exposure/AUC-driven effects (e.g., aminoglycoside nephrotoxicity, many chronic effects) - and explain how infusion rate, dosing interval, or protein binding change the metric without changing total daily dose.
Treats a class of metals or metalloids as a single toxic entity - 'arsenic', 'mercury', 'chromium' - and assigns one potency and target organ regardless of speciation (arsenite vs arsenate vs MMA/DMA; elemental vs inorganic vs methylmercury; Cr(III) vs Cr(VI)).
Element names are the salient retrieval cue, while speciation, valence state, and biotransformation determine absorption, distribution, and mechanism; the distinction is easily lost under time pressure.
What to do instead
When a metal appears, immediately state the species/form in the stem, the route-dependent absorption it implies, and the mechanism tied to that form (e.g., trivalent arsenite binds vicinal thiols and inhibits pyruvate dehydrogenase; pentavalent arsenate uncouples oxidative phosphorylation; methylmercury is lipid-soluble and CNS-targeted).
Computes a reference dose or acceptable intake by dividing a NOAEL by a single 10-fold factor, omitting interindividual (10H) and database/modifying factors, or mixes up which uncertainty factors apply to which derivation.
The 'factor of 10' is the memorable headline, and the composite product of 10A × 10H × 10S × 10D × 10L is rarely written out under exam time pressure, so components get defaulted away.
What to do instead
Write the full equation explicitly every time: POD ÷ (UF_A × UF_H × UF_S × UF_D × UF_L), state which defaults are being replaced by chemical-specific data, and keep RfD units (mg/kg-day) distinct from cancer slope factor units (mg/kg-day)^-1.
Interprets an odds ratio from a case-control study as if it were a risk ratio or absolute risk, and draws causal conclusions from cross-sectional or ecological designs without checking temporality or confounding.
Effect measures all look alike numerically, and candidates default to 'OR = RR' without invoking the rare-disease assumption; study-design hierarchies are memorized as labels rather than as constraints on inference.
What to do instead
Identify the design first, then the measure it legitimately yields: cohort → RR/AR; case-control → OR, which approximates RR only when the outcome is rare (roughly <10%); cross-sectional → prevalence ratio, no incidence inference. Name at least one plausible confounder and whether the design can control it.
Treats a statistically significant result in a small or multiply-tested study as proof of a toxic effect, without checking confidence interval width, study power, number of comparisons, or biological plausibility.
Significance testing is over-learned relative to estimation, and 'p < 0.05' is read as a binary verdict rather than as one piece of evidence conditional on design quality.
What to do instead
Read the effect estimate with its CI before the p-value: ask whether the CI excludes both the null and biologically trivial magnitudes, whether multiplicity was addressed, whether the study was powered for that endpoint, and whether the finding replicates or has a mechanistic anchor.
Converts an in vitro IC50 or in vivo animal dose directly into a human exposure threshold (or vice versa) without in vitro-to-in vivo extrapolation - free fraction, scaling by body surface area, metabolic clearance differences, and route bioavailability.
The single potency number is treated as a portable constant, and candidates skip the dosimetry bridge because it requires assembling several assumptions at once.
What to do instead
State the bridge explicitly: in vitro nominal concentration → free (unbound) concentration → in vivo plasma concentration after scaling and clearance → administered dose via bioavailability; for animal-to-human, apply body-surface-area scaling and flag species differences in metabolism and receptor sensitivity as residual uncertainty.