Academic OB/GYN

This month’s Green brings us not one but two articles on the effect of LEEP on subsequent risk of preterm birth.   The first article, published by Fins and Swedes Jakobsson and Gissler et al, looks at a wide population of patients, while the Danes Noehr and Jensen et al fire back with a look at the effect of LEEP on preterm birth rates in twins.

It’s a Scandinavian throw down!  Who did it better?

Bottom line:  LEEP increases the risk of subsequent preterm birth, and the Danes do a better job at showing it.

Both studies share similarities in their study design, using data extracted from National Databases in their respective countries (Finland and Sweden for Jakobsson et al and The Netherlands for Noehr et al.)   Both are cohort studies created from retrospective data, also known a non-concurrent prospective cohort studies.   In the Jakobsson paper, preterm birth rates in women with history of LEEP were compared to preterm birth rates throughout the whole population, While in Noehr et all preterm birth rates in all twins were looked at, with groups stratified on exposure to prior LEEP prior to birth.  Group sizes were large in both cases, with 624 LEEP patients vs. 550k population controls in the first paper, and 166 LEEP patients vs. 9,702 population controls in the second paper (twins only.)

Jakobsson also reports a subgroup analysis of women who had recorded deliveries before and after LEEP, in an attempt to include a dataset that is less likely to be affected by selection bias.   He also reports indications for LEEP, which is not done in Noehr et al.

When all is said and done, both the Fins/Swedes and the Danes demonstrate that LEEP is a risk factor for future preterm birth.   Jakobson found that across all deliveries incidence of preterm birth increased 160% after LEEP (RR 2.61, 95% CI 2.02 – 3.02, NNT for harm  = 14) and more than 500% after repeat LEEP (RR 5.15, 95% CI 2.45 – 7.84, NNT for harm = 5.)  When looking only at women who delivered before and after LEEP, the risk was still elevated, but was somewhat lower (RR 1.94, 95% CI 1.10 – 3.40, NNT for harm = 18.)    Noehr et al found similar results in his twins data (RR 1.58, 95% CI 1.16 – 2.14), but with much higher underlying prevalence of preterm birth of 32%, as would be expected in twins, as opposed to only 4.6% noted in the entire population in the Jakobsson study.

So should you believe it?  I always ask myself this anytime a read a study.   In this case, I think the answer is yes.  Both studies have a few issues, but overall I think they are done well enough that the outcome is likely the result of a true association between LEEP and preterm birth.

But what were the problems?   There were a few, more so in generalized study than in the twins study.

Both studies are cohort studies, which are inherently difficult to control, and there is one huge bias that needs to be controlled here.   People with cervical dysplasia often smoke, and may, at large, be of different socioeconomic or medical backgrounds than patients that don’t.  Assuming this is true, this creates the potential for LEEP to be a marker for these other risk factors rather than a risk factor itself.

The two studies both dealt with this issue, but each in different ways – subgroup analysis in the first and logistic regression in the second.   In the Swedish/Finnish study, they first compare all patients who had LEEP to the entire population of women who delivered over the study period.   They looked at socioeconomic statuses between these groups, and found a difference between women that had a LEEP and the general population, though after reading it three times I still can’t figure out what that was.  They didn’t report p values in their Table 1, so I can’t quite tell where the significant differences were.  They were not able to report on smoking history across the entire population, though only about 25% of the study group smoked.   Recognizing this problem, the reported a subgroup of women who had deliveries before and after LEEP, thus serving as their own controls.

So what the hell?   To me, this was just all a little bit strange.  They had this massive database with tons of variables for each patient.   It seems to me that they could have created a matched cohort of controls, based on the data on each women who had LEEP.  Using the entire population just seems a bit lazy to me, and ultimately decreases the value of that part of the analysis to me.  I like that they used the subgroup to control this, but to me this is like doing another study, not controlling for the first one.   I would have liked it better if they had just reported the subgroup and forgot the entire population registry.  They have one good nugget of beautiful controlled data, and throw a huge pile of lower quality data on top.  Uncontrolled cohorts tend to over represent effect sizes, and we see this here.  In the large group, the effect of LEEP was RR 2.61, but in the better-controlled subgroup RR was only 1.94.  I suspect the second number is closer to the true effect of LEEP.

I like the way the Danes did it better.  They had a smaller dataset, and they controlled it.   They did a proper logistical regression to remove the effects of smoking, maternal age, calendar time, IVF, and marital status.   This is the way they do it in Copenhagen, and this how you should do it too.

Like the Fins and Swedes, they found an increased risk of preterm birth with LEEP, in their an adjused OR of 1.58.  Interestingly, their better-controlled data lined up almost exactly with the small better-controlled subgroup of the first study.    Hmmm….  Maybe the Jakobsson should have just reported that data and bagged the large registry data?  I say yes.  Still, its difficult to compare the two studies fairly as the first calculated relative risk and the second adjusted odds ratio, and the second study looked at a much higher risk population.

Finally, because of all the machinations the Fins and Swedes had to do to justify their data, they present a six page manuscript in comparison to only five pages for the Danes.  Brevity is good.  The first paper doesn’t have more good stuff in my mind, its just longer.   They also report why all these people had LEEPs, which is distracting to me, and breaks the “least publishable unit” rule.  This data would have made another paper that would have been better than the little bit of data thrown into this one.  Given that in their dataset 40-45% of all LEEPs were performed for CIN I or less, the issue certainly deserved some discussion.

Study 1 – Loop electrosurgical excision procedure and the risk for preterm birth.
Jakobsson M, Gissler M, Paavonen J, Tapper AM. Obstet Gynecol. 2009 Sep;114(3):504-10.

Study 2 – Loop electrosurgical excision of the cervix and risk for spontaneous preterm delivery in twin pregnancies.  Noehr B, Jensen A, Frederiksen K, Tabor A, Kjaer SK.  Obstet Gynecol. 2009 Sep;114(3):511-5.

So what do you think about it?  Leave a comment or email to email@academicobgyn.com