Keeping asymptomatic wisdom teeth and watching is the choice that most often fails to close the matter: across seven longitudinal studies, about 64% of retained symptom-free third molars are extracted within 18 years, at a steady rate of roughly 3% per year (Bouloux et al., 2015). A same-era 18-year cohort found 73% of the teeth that were unerupted at baseline came out before follow-up ended (Ventä et al., 2004), though that study’s follow-up window overlaps the period Bouloux pooled, so the two are better read as consistent than as independent. The reasons are ordinary: caries, gum disease, later inflammation, and part of that 64% is care that was simply timed to need rather than regretted. The practical upshot is that watching usually defers the operation rather than avoiding it, and the extraction then falls later in life rather than in early adulthood.
Removing the tooth up front does not escape regret either; it relocates it. Because prophylactic removal is itself an extraction, its regret falls on the mirror-image group: the roughly 36% of removed teeth that would have stayed trouble-free over the same 18 years, operated on for nothing. Every one of those extractions carries about a 4.2% chance of a complication, overwhelmingly dry socket, with infection and jaw-joint symptoms far behind (Chen et al., 2021). The Cochrane review found only two studies and 493 people bearing on whether disease-free impacted wisdom teeth should come out at all, and rated the evidence very low certainty (Ghaeminia et al., 2020). NICE went further in 2000, concluding that prophylactic removal of pathology-free impacted third molars should be discontinued in the NHS — a position US practice frequently sets aside.
The two figures are close on purpose, because the decision is close. Both come from the same fact: about two-thirds of these teeth eventually need removal and about a third do not. Watch, and you carry a 64% chance of the surgery you were trying to avoid; remove, and you carry a 36% chance the surgery was needless. Severity runs opposite to frequency — the deferred operation is common but often appropriate, while the unnecessary one is rarer but incurred up front against a benefit the evidence cannot show. What survives the symmetry is a modest lean, consistent with Gilovich and Medvec’s temporal asymmetry: over a long horizon, more people end up back in the chair than are spared it. Whether that reads as regret or as a reasonable bet that mostly postponed a minor operation is the judgement the evidence declines to settle.