When conclusions don’t tell the whole story, and on paracervical blocks for IUDs
I recently was looking into whether or not you should place a paracervical block before placing an intrauterine device, after a particular patient had a difficult time with a placement. Over my career, I generally haven’t done so, as the vast majority of patients don’t seem to have a significant amount of pain, and the literature I have read has not been very supportive of the practice.
Today I looked at the literature again, and was a little irritated at what I found. One particular study was Mody et al, which concluded “Compared with no anesthetic, a 1% lidocaine paracervical block did not result in a statistically significant decrease in perceived pain with IUD insertion.”
So clearly most would read this and say that it doesn’t matter if you put in a paracervical block or not.
But’s here’s the kicker; the summary of their data is the following: ” Twenty-six women received the paracervical block before IUD insertion, and 24 received no local anesthesia. Groups were similar in age, parity, ethnicity, education and complications. Women who received the paracervical block reported a median VAS score of 24.0 mm with IUD insertion, and women who did not receive local anesthetic reported a median VAS score of 62.0 mm with IUD insertion; p=.09.”
So patients who got a block had pain of 2 on a scale of 2 to 10, and patinets who did not have a block had pain of 6 on the same scale. So did the paracervical block really not work?
I would say it did. What didn’t work was the study, or at least most likely so. The p value, or likelihood that the outcome was do to statistical chance alone, was 0.09. In translation, there was a 9% chance that the difference between the groups was due to chance alone, and a 91% chance that it was due to an actual effect of the paracervical block. Since we arbitrarily say that a p of 0.05 is statistically significant, the authors say that the study was negative.
But isn’t this quite misleading? It is 10 to 1 likely that the paracervical block actually made the procedure less painful, and we just didn’t reach a p of < 0.05 because there weren’t enough patients in the study to adequately separate the groups mathematically (lack of power).
P values are important, but we can’t consider them to be everything. Statistical significance is a continuous variable, not a nominal one. If we say that p=0.049 means that the study shows a difference and p=0.051 means there was no difference, we are just being foolish, and in the end quite ignorant of the actual mathematics that goes into how the p value was created.
Here’s how I would like to see such a study concluded : “Compared with no anesthetic, a 1% lidocaine paracervical block showed a strong trend towards decreasing pain with IUD insertion, that did not reach statistical significance. Further study with a greater number of patients is warranted.”
If you read their conclusion, you would probably take from that that paracervical blocks don’t work. But if you read mine, you take a different message, which is actually supported by the data, which would be “It is very likely that paracervical blocks decrease pain with IUD insertion”, which of course makes sense
Mody SK; Kiley J; Rademaker A; Gawron L; Stika C; Hammon C. Pain control for intrauterine device insertion: a randomized trial of 1% lidocaine paracervical block. Contraception. 2012; 86(6): 704-9.