2.4 Empiric Risk, Multifactorial Recurrence & Risk-Assessment Models

2.4 Empiric Risk, Multifactorial Recurrence & Risk-Assessment Models

Quick Answer: When no single-gene mechanism is established, use empiric risk — observed recurrence in studied families. Multifactorial recurrence rises with more affected relatives, closer relationship, greater severity, consanguinity, and affection of the less-susceptible sex (Carter effect). Named models split into carrier-probability tools (BRCAPRO, PREMM5, MMRpro) and absolute-risk tools (Gail, Claus, Tyrer-Cuzick, BOADICEA/CanRisk); choosing the wrong class is the classic error.

Domain 2A gives 17 scored items to risk assessment, and only some of them are Mendelian or Bayesian. ABGC also lists a task statement about evaluating the applicability of reference resources, risk assessment tools, position statements, and practice guidelines — which is a direct instruction to know what each model can and cannot do.

Empiric risk: what it is and when to use it

An empiric risk is derived from observed recurrence in families ascertained through an affected proband, not from calculating gamete segregation. Use it when:

• The condition is multifactorial (neural tube defects, cleft lip/palate, most isolated congenital heart disease, autism, schizophrenia).
• The mechanism is heterogeneous and no molecular diagnosis was found.
• A chromosomal event is sporadic (recurrence after a de novo trisomy).
• Germline mosaicism makes an apparently de novo dominant condition carry a small nonzero recurrence.

Empiric figures are population- and study-dependent, so counsel them as ranges tied to a source rather than as exact constants. Commonly cited sibling recurrence ranges for an otherwise unremarkable family with one affected child:

Do not memorize these as fixed exam constants. Board stems that require a number usually supply it; what is scored is whether you recognize that an empiric figure — not 1/4 or 1/2 — is the right kind of answer.

The multifactorial threshold model

Liability to a multifactorial trait is normally distributed in the population; individuals whose liability exceeds a threshold are affected. Relatives of an affected person have a liability distribution shifted toward the threshold, which is why recurrence is elevated but far below Mendelian values.

Five modifiers raise recurrence, and every one of them shows up in exam vignettes:

1. Number of affected relatives. Two affected siblings shifts the family further right than one; recurrence roughly doubles or more.
2. Degree of relationship. Risk falls sharply from first-degree to second- to third-degree relatives — a much steeper fall-off than the halving seen in dominant inheritance.
3. Severity of the defect. Bilateral cleft lip and palate implies higher family liability than unilateral cleft lip alone.
4. Consanguinity. Shared background liability alleles raise recurrence for multifactorial as well as recessive conditions.
5. Sex of the affected proband — the Carter effect. When a condition affects one sex more often, an affected member of the less commonly affected sex implies a higher liability load in the family and therefore higher recurrence for relatives. The classic teaching example is infantile pyloric stenosis, which is more common in males: relatives of an affected female proband carry the higher recurrence risk.

Trap: reversing the Carter effect. The intuitive but wrong answer is that the more commonly affected sex signals more risk. It is the opposite — affection in the rarer sex means the family needed more liability to cross a higher threshold.

Prevention as part of the risk conversation

Empiric risk counseling is incomplete without the modifiable lever. For neural tube defects, the standard preconception recommendation is 0.4 mg (400 µg) of folic acid daily for average-risk pregnancies, escalating to 4 mg (4,000 µg) daily beginning at least one month before conception and through the first trimester for a person with a prior NTD-affected pregnancy. Naming both figures, and the timing, is the counseling half of the item.

Named risk-assessment models

The most-missed distinction in this subdomain is what the model outputs.

Applying the right model, correctly

• Guideline thresholds specify the model class. NCCN-style discussion of annual breast MRI uses a lifetime risk of roughly ≥20% calculated with models that are largely dependent on family history — Tyrer-Cuzick, Claus, or BOADICEA. Using a Gail score to clear or justify MRI is a documented misapplication, because Gail was never designed to capture a strong pedigree.
• Verify the pedigree before you run anything. Every model inherits the errors of the family history fed into it. A "maternal aunt with ovarian cancer" that turns out to be cervical cancer can move a Tyrer-Cuzick output across a management threshold.
• Do not apply a model outside its validated population. Most of these tools were developed and validated primarily in populations of European ancestry; their calibration in other ancestral groups is weaker, and that limitation belongs in the counseling.
• A model output is not a test result. A 24% lifetime risk is a probability statement supporting a screening discussion, not a diagnosis, and a 6% BRCAPRO carrier probability does not exclude a pathogenic variant.
• Model output can conflict with testing criteria. A patient can meet testing criteria while sitting below an MRI threshold, or vice versa. Answer the question that was asked.

Common traps

• Applying a 1/4 or 1/2 Mendelian figure to an isolated multifactorial malformation.
• Reversing the Carter effect.
• Using Gail for a woman with two affected first-degree relatives and an affected paternal aunt.
• Quoting a carrier-probability model output as the patient's cancer risk, or the reverse.
• Presenting an empiric range as a precise constant without naming the source or population.
• Forgetting the 4 mg folic acid escalation for a prior NTD-affected pregnancy and counseling the routine 0.4 mg dose instead.