What “discordant” and “double discordance” mean here

Discordant (in sibling/twin designs)

• Discordant for exposure: within a sibling pair, one sibling was exposed and the other was not (e.g., one fetus was exposed to acetaminophen in utero; the co‑sibling was not). This is a classic “discordant sibling/twin” setup used to control for many shared genetic and familial factors in within-family analysesW. Within-family contrasts are often used precisely because they hold constant much of what siblings share (genes, early-life environment), strengthening causal interpretation when correctly designedP.

• Discordant for outcome: within a sibling pair, one sibling has the outcome (e.g., a neurodevelopmental diagnosis) and the other does not.

In the podcast excerpt, “discordant” is being used in these epidemiologic senses—discordant for exposure and/or discordant for outcome within sibling pairs.

“Double discordance” in this context

• In Attia’s segment, “double discordance” refers to selecting only sibling pairs that are simultaneously discordant for exposure and discordant for outcome (i.e., both conditions must differ within the pair). This is a colloquial label here for “discordant on both dimensions” and is distinct from the cardiology usage of “double discordance” for a congenital heart defect.

Why selecting only “double-discordant” pairs is a problem: collider bias

Collider bias arises when an exposure and an outcome both influence a third variable, and the analysis conditions on (or selects by) that third variable. As the CDC explains, “Collider bias occurs when an exposure and an outcome each influence a common third variable and that variable or collider is controlled for by study design or in the analysis,” which “can make associations appear real when there is not a true causal association in the general population”B. The Catalogue of Bias likewise emphasizes that collider bias can be induced by sampling or analysis choices and can even “inflate a null effect to a positive effect or switch a harmful effect to a protective effect”C.

• How this applies here: Requiring sibling pairs to be discordant for both exposure and outcome means “being included in the analysis” depends on both exposure status and outcome status. That inclusion criterion effectively conditions on a variable that is a downstream function of both exposure and outcome, which is the hallmark of collider conditioning. Even if acetaminophen had no causal effect, such selection can induce a spurious within-family association (or reverse its direction) purely from the way families are chosen for analysisC.

Intuition via a simple thought experiment

• Imagine the true causal effect of prenatal acetaminophen on neurodevelopment is zero.
• Families differ in unmeasured factors (e.g., maternal illness, stress) that both increase acetaminophen use and increase risk of neurodevelopmental diagnoses—these are confounders shared by siblings.
• If you analyze all sibling pairs (not conditioning on outcomes), a within-family comparison can help control those shared confounders.
• But if you only keep pairs where:
  1. one sibling was exposed and the other was not, and
  2. one sibling has the diagnosis and the other does not, you have guaranteed a negative correlation within the selected pairs: exposure and outcome must “mismatch” in each included family. This mechanical selection can create or exaggerate an association even when none exists causally—classic collider bias from selectionB.

That is why Attia notes the senior author’s clarification matters: if a review misstates that the Swedish sibling study kept only pairs discordant for both exposure and outcome, that would imply a design prone to collider bias; the author’s correction is asserting they did not use that biased selection.

How sibling-control designs are meant to work (and common pitfalls)

• Intended approach:
  • Include families with variability in exposure across siblings (exposure-discordant pairs), then compare outcomes within families. This leverages shared genetics/family environment to reduce confoundingP.
• Pitfall to avoid:
  • Conditioning on both exposure discordance and outcome discordance (i.e., retaining only “double‑discordant” pairs). That selection is a form of conditioning on a collider and can distort the exposure–outcome relationship—sometimes even reversing itC.

Bottom line definitions, tailored to the excerpt

• Discordant (for exposure): within a sibling pair, one pregnancy involved acetaminophen use and the other did not.
• Discordant (for outcome): within the same pair, one child has a neurodevelopmental diagnosis and the other does not.
• Double discordance: a selection rule that keeps only sibling pairs that satisfy both kinds of discordance simultaneously—an approach that can induce collider bias because inclusion depends on both exposure and outcomeBC.

As the excerpt notes, whether the Swedish study actually used that “double discordance” restriction is the crux of the dispute; the senior author reportedly states they did not—important because avoiding that selection helps prevent collider bias.