The American Association of Oral and Maxillofacial Surgeons Fails to Use Basic Statistics in Research on Wisdom Teeth

I have previously reported how the American Association of Oral and Maxillofacial Surgeons (AAOMS) recently had a press conference on wisdom teeth in Washington, D.C. See http://blog.teethremoval.com/third-molar-multidisciplinary-press-conference/ for more information.

Shortly after the press conference they issued a press release available at http://www.aaoms.org/docs/media/third_molars/press_release.pdf which is titled “Conventional Wisdom about Wisdom teeth Confirmed: Evidence Shows Keeping Wisdom teeth May be More Harmful than Previously Thought.” One of the additional key findings listed in this press release  is

“Most patients (60 percent) with asymptomatic wisdom teeth prefer extraction to retention.”

This finding comes from a recent article in the Journal of Oral and Maxillofacial Surgery, titled “Most Patients With Asymptomatic, Disease-Free Third Molars Elect Extraction Over Retention as Their Their Preferred Treatment.”  The article is by Brian E. Kinard, BS and Thomas B. Dodson, DMD, MPH.  It appears in the December 2010 issue in volume 68, issue 12, on pages 2935-2942.

The article uses a study sample from patients presenting to the Department of Oral and Maxillofacial Surgery at Massachusetts General Hospital between November 2008 and August 2009 for the evaluation and management of their third molars (wisdom teeth) by Thomas B. Dodson.  In the article on Table 8 it is presented that a total of 319 patients were seen during this time who had asymptomatic disease free wisdom teeth. 129 of these patients chose to keep their wisdom teeth and 190 of these patients chose to extract their wisdom teeth. A simple calculation was performed by the authors of 190/319 * 100 = 60%. (it actually equals 59.5611 % but they rounded up). This is how they arrived at their statement quoted above that most patients with asymptomatic wisdom teeth prefer extraction and how they arrived at the titled of the article most patients prefer extraction.

Thomas B. Dodson admits that their bias in this result as “…it is possible for clinicians to present treatment options in a manner, consciously or unconsciously, that directs patients toward the clinician’s preferred treatment.”  This is a valid concern. I also have a problem in that no statistical analysis was done beyond this point with this specific result.

During my undergraduate studies I took several courses on statistics. One popular program to use is MINITAB in addition to knowing how to do hand calculations.

Using this proportion data it is possible to do in MINITAB and by hand a 1 proportion test and determine a confidence interval. A 95% (two sided) confidence interval of this data is (0.539505, 0.649924).  This is calculated by calculating an estimator for the standard error. This estimator is the square root of [(p * (1-p)/n] where p is in this case 190/319 corresponding to the sample proportion which is an estimator of the population proportion and n is the sample size which in this case is 319. We then look up in a t table or use MINITAB to determine the test statistic, which in this case is 1.96745. The 95% confidence interval is then calculated as 190/319 +/- sqrt[(190/319 * (1-190/319)/319] *1.96745  . We then arrive at our 95% confidence interval of 53.9505% to 64.9924% which means we are 95% confident that the true population proportion of those who keep their asymptomatic wisdom teeth falls in this range.

The other important piece of information we need to asses is if we have enough samples in our data. A total of 319 patients were used in this study who had disease free wisdom teeth but it turns out we need more than 319 patients to make an accurate assessment before we can even calculate a 95% confidence interval.

If we look at the Statistics Department at Penn State University we can easily find an educated guess and  conservative method to determine the required sample size.  In this case our educated guess is calculated as [ (1.96)^2 *  0.595611 * (1 – 0.595611) ] / (0.05)^2 which equals 370.113 and our conservative method is calculated as [ (1.96)^2 *  0.5 * (1 – 0.5 ] / (0.05)^2 which equals 384.16. Thus we determine that it is necessary to have at least 371 patients in our study to be able to even come up with a statement as to whether or not we can be 95% confident that patients prefer either to extract or retain healthy impacted wisdom teeth.

The authors only used 319 patients in their study and thus the data does not allow them to make such as statement as they did which AAOMS subsequently reported in a press release.

I encourage Thomas B. Dodson and other oral surgeons to consider using some basic statistics in their research before releasing such information to the the public who should be receiving information that can allow them to make informed decisions about their health.

, , , , , , , , , ,

3 Responses to The American Association of Oral and Maxillofacial Surgeons Fails to Use Basic Statistics in Research on Wisdom Teeth

  1. Hannah December 13, 2010 at 8:34 am #

    It’s unfortunate how they skewed the data, whether intentionally or unintentionally. I hope that they continue their research and testing to make it more accurate.

  2. Daniel October 1, 2013 at 5:46 pm #

    The statistics are actually correct, 59.56% is the true value of patient responses and confidence intervals are not appropriate for that calculation.
    Like any respected peer review journal, JOMS has a statistics editor and I’m willing to say he took more than college statistics.

  3. wisdom October 2, 2013 at 10:43 pm #

    My primary issue with this particular study is the fact that no a priori sample size calculation was performed.

    A total of 319 patients were included in this study and based on calculations as demonstrated in the above post more patients should have been included to avoid Type II errors.

    See also http://ocw.jhsph.edu/courses/statmethodsforsamplesurveys/PDFs/Lecture3.pdf from a Sample Size and Power Estimation Lecture from John Hopkins. I am assuming a 5% margin of error in the above calculation, which is of course open to some debate…

Leave a Reply