ABMGG Clinical Genetics & Genomics common mistakes
ABMGG Clinical Genetics & Genomics · Understand the traps and how to avoid them.
Assigns a pathogenicity class on the strength of a single computational predictor, a single ClinVar submitter, or wording like 'predicted damaging,' instead of applying ACMG/AMP criteria code by code.
Pattern matching feels fast and familiar under time pressure, and supporting-level evidence (PP3/BP4) is easy to overvalue; ClinVar aggregates conflicting, often non-curated submissions, so a single entry reads like an answer when it is only one piece of evidence.
What to do instead
Force an explicit evidence table: list each criterion (PVS1, PS1-4, PM1-6, PP1-5) with the data supporting or refuting it, apply the combining rules, and record the rationale. Treat ClinVar assertions as one line of evidence and state how much weight you gave them and why.
Quotes a recurrence risk from memory (for example '1 in 4') without drawing the pedigree and setting up a Bayesian table, so residual risk after a negative carrier test or in an unaffected sibling is reported as unchanged.
Textbook risks are anchored to the simplest scenario; conditional probability arithmetic requires deliberate setup and is easy to skip when clinic time is short, and the resulting number sounds authoritative enough that nobody rechecks it.
What to do instead
Draw the pedigree, then write prior, conditional probability for each piece of evidence, and posterior as a three-column table before stating any number. Sanity-check that the posterior moved in the expected direction and that unaffected relatives and negative tests appropriately reduce risk.
Computes carrier frequency as 2 times the square root of incidence, or applies p squared plus 2pq plus q squared to a population with consanguinity, founder effects, or assortative mating, without flagging that the assumptions fail.
The formula is memorized as a plug-in; its assumptions (random mating, no selection or mutation, large population, negligible migration) are not recalled at the moment of application, so the calculation looks rigorous while the model is wrong.
What to do instead
Before using Hardy-Weinberg, state the required assumptions and check each against the actual population. Where consanguinity or founder effects apply, use the inbreeding coefficient or empirical carrier rates, and explain the adjustment rather than quoting an unadjusted number.
Orders the most familiar test rather than the discriminating one, such as a panel or exome that cannot detect copy-number changes when a deletion syndrome fits, or a molecular test before the biochemical assay that would establish the diagnosis.
Test menus are learned by specialty habit, and assay characteristics such as CNV detection, mosaicism sensitivity, tissue requirement, and turnaround are not systematically compared against the clinical question being asked.
What to do instead
Write the clinical question first, then choose the assay whose detection scope, sensitivity, and tissue requirements match it. Confirm that panels include exon-level deletion/duplication analysis, and know in advance which conditions require biochemical confirmation or a specific tissue.
Performs a targeted dysmorphology evaluation keyed to the referral question without recording occipitofrontal circumference and growth plotted on appropriate curves, or without examining the parents for the same discriminating features.
The referral letter frames the question, clinic time is short, and the physical exam drifts into hypothesis-confirming rather than hypothesis-generating mode, so findings that would redirect testing are never collected.
What to do instead
Use a fixed minimum dataset for every genetics visit: occipitofrontal circumference, height and weight plotted on curves, and a hands-on exam of both parents for the relevant features. Document pertinent negatives as explicitly as positives and let the data drive the differential.
Stops at the first plausible diagnosis and attributes every later finding to it, never asking whether a second independent condition explains a discordant feature or a later-onset problem.
Premature closure feels efficient and is reinforced in the short term; two diagnoses in one patient are uncommon but not rare, and age-dependent conditions present sequentially so the second looks like a complication of the first.
What to do instead
At each follow-up, list the features that are atypical for the established diagnosis. If any are present, explicitly consider a second diagnosis or a dual molecular finding, and use the known natural history of the named condition to judge which findings are expected and which are not.
Treats consent as a signature rather than a conversation: never asks about opting out of secondary findings, never discusses implications for at-risk relatives, or offers predictive testing to a minor without weighing the child's future autonomy.
Consent forms compress complex choices into checkboxes, and the ethical reasoning behind secondary findings policy, cascade testing, and deferred autonomy for minors is distinct from the administrative act of obtaining a signature.
What to do instead
Use a structured consent script covering test purpose, secondary findings policy, implications for relatives, data sharing, and minors' autonomy; document what was discussed and what the patient chose, not merely that a form was signed.
Gives surveillance recommendations by gene name only, without specifying onset age, interval, modality, or guideline source, for example 'annual MRI' for a condition whose risk begins in childhood and follows a different schedule.
Management is recalled as a headline such as 'screen the heart' or 'watch the kidneys,' while the operational details that make a recommendation actionable are not consolidated alongside it.
What to do instead
Tie every recommendation to a published disease- and age-specific guideline: state start age, interval, modality, end point, and the source. Note explicitly when evidence is extrapolated, weak, or absent rather than presenting all recommendations as equally established.
Reads a variant allele fraction of 20 to 30 percent in blood as proof of germline heterozygosity, or as proof of somatic mosaicism, without considering tissue specificity, age-related clonal hematopoiesis, or the assay's limit of detection.
The variant allele fraction is mentally collapsed directly onto a genotype; the biology of mosaicism, its tissue distribution, and the validated sensitivity of the assay are not tracked alongside the number on the report.
What to do instead
Interpret variant allele fraction jointly with the tissue sampled, the patient's age, and the test's validated detection threshold. When mosaicism would change recurrence risk or phenotype, state the possibility of tissue-limited mosaicism and test a second tissue rather than concluding from blood alone.
Closes the encounter without a documented plan for at-risk relatives: no family letter, no cascade testing offer, no specification of who should be tested and in what order.
The patient in front of you consumes the visit, relatives remain abstract, and responsibility for initiating contact is ambiguously shared between clinician and proband, so nothing gets assigned.
What to do instead
End every visit with a written family action plan: which relatives are at risk, at what magnitude, which test they should be offered, and who will make first contact. Provide a family letter and document the plan in the record.