Design: secondary analysis of a prospective multicentre cohort (MFMU APEX), multivariable logistic regression with tests for interaction
Population: 5,277 US women having operative vaginal delivery of a term cephalic singleton — 3,453 vacuum, 1,824 forceps; episiotomy in 32.3%
Primary outcome: obstetric anal sphincter injury, defined as a third- or fourth-degree perineal laceration
Effect: sphincter injury in 20.1% (95% CI 19.0–21.2); any episiotomy vs none adjusted OR 1.45 (95% CI 1.25–1.67), with midline associated with higher and mediolateral with lower odds

Obstetric anal sphincter injury is the complication that follows a woman for years after an operative vaginal delivery, and the role of episiotomy in preventing or causing it has never settled. A secondary analysis of the MFMU APEX cohort — 5,277 women delivering a term cephalic singleton by vacuum or forceps in the United States — separates the question by the cut that was actually made.

Sphincter injury occurred in 20.1% of these deliveries (95% confidence interval 19.0–21.2), which is itself worth absorbing. Taken as a single exposure, episiotomy was associated with higher odds of injury: adjusted odds ratio 1.45 (95% CI 1.25–1.67), adjusting for length of second stage, maternal age, body mass index and birth weight. But the interaction terms were significant for both delivery type (P<.001) and previous vaginal delivery (P=.02), and the direction reversed with the type of cut. Midline episiotomy at vacuum delivery was associated with more sphincter injury, with or without a prior vaginal birth. Mediolateral episiotomy was associated with fewer injuries in women who had not delivered vaginally before, at both vacuum and forceps delivery.

So the aggregate figure conceals two opposite effects, and the harm sits with the midline cut. For a primiparous woman heading for an instrumental delivery, mediolateral is the defensible choice; for a woman who has delivered vaginally before, the case for cutting at all is thin. Indian units largely cut mediolateral already, which is on the right side of this finding — the value here is in resisting drift towards midline and in narrowing who is cut at all.

Decide episiotomy type before the instrument goes on, not in the moment.
If you cut at operative vaginal delivery, cut mediolateral.
In a woman with a previous vaginal delivery, reconsider whether to cut at all.
Examine every woman rectally after operative vaginal delivery — one in five in this cohort had a third- or fourth-degree tear.
Record the episiotomy type explicitly in the delivery note; 'episiotomy performed' is not enough to audit against.

Why it matters

'Episiotomy at operative delivery' is treated as one decision in most delivery notes, and the two versions of it pull in opposite directions.

The statistics, in plain English

The adjusted odds ratio of 1.45 for episiotomy as a single exposure is the wrong number to act on, because it averages a harmful midline effect with a protective mediolateral one — that is what the significant interaction terms are telling you. This is a cohort, not a trial: episiotomy was chosen by the clinician, and the reasons for choosing it are likely to be linked to the risk of tearing. Adjustment for second-stage length, age, body mass index and birth weight helps but cannot remove that. The 20.1% injury rate is a measured incidence in this cohort and is not adjusted for anything.