Ramdass R,  Meyr AJ.  The multiplanar effect of first metatarsal osteotomy on sesamoid position.  J Foot Ankle Surg.  2010 Jan-Feb; 49(1): 63-7.   (PubMed ID:20123290)

WHY did the authors undertake this study?
The standard classification system used to evaluate the sesamoids in perioperative planning of hallux abductovalgus (HAV) is a uniplanar description of a multiplanar deformity.  Additionally, both the sesamoids and first metatarsal are thought to move relative to each other during corrective surgery.  The aim of this study was to evaluate the sesamoid position in multiple planes relative to a stationary anatomic landmark (the 2nd metatarsal) before and after distal first metatarsal HAV surgery.

HOW did they attempt to answer this question?
The primary outcome measures of the study  were radiographic parameters measured before and after HAV surgery.  Measures in the transverse plane included the 1st intermetatarsal angle, hallux abductus angle, tibial sesamoid position, and the tibial sesamoid-2nd metatarsal distance.  Measures in the frontal plane included the sesamoid rotation angle, tibial sesamoid grade and axial tibial sesamoid-2nd metatarsal distance.

WHAT were the specific results?
Statistically significant differences between pre- and post-operative values with improvement towards rectus joint alignment were found for all measures that included the first metatarsal, but no significant change in tibial sesamoid position was found in either the transverse or frontal plane when measured relative to the stationary second metatarsal.

HOW did the authors interpret these results?
From these results, the authors concluded  that the sesamoids may be relatively immobile during HAV corrective surgery, and intervention may best be directed at the relatively mobile 1st metatarsal.

Stubhaug A.  Managing chronic osteoarthritic pain in primary care:  an update.   J Pain Palliat Care Pharmacother.  2009; 23(4): 380-3).

WHY did the authors undertake this study?
Chronic pain from osteoarthritis is one of the most common presenting complaints seen by physicians in both primary care and orthopedic specialties.  Pharmacological and nonpharmacological interventions should form the initial basis of treatment.  The author reviews these interventions with a three-step pharmacotherapy paradigm. 

HOW did they attempt to answer this question?
Epidemiologic and diagnostic considerations of osteoarthritis are first considered.  Differential diagnoses should be considered carefully when several findings are present including patients presenting at a younger age (<45 years), when the complaint is at unusual joints (shoulder, elbow and ankle), febrile conditions are present, the morning stiffness lasts for >30 minutes, and when the pain is progressive and occurs mainly at night.

WHAT were the specific results?
A 3-step pharmacological treatment paradigm is presented.  Step 1 begins with non-pharmacologic interventions (aerobic exercise, joint strengthening, active range of motion, weight loss, etc), acetaminophen (up to 4g/day), topical NSAIDs and topical capsaicin.  Step 2 introduces low dose or weak opioids including oral codeine, oral tramadol, or transdermal buprenorphine.   Step 3 involves high-dose strong opioids, but the author points out that evidence for the beneficial long term use of this intervention is limited.  Oral NSAIDs are recommended at any stage, but only for acute inflammatory flares.

HOW did the authors interpret these results
The author ends by pointing out that routine reassessment (every 3-6 months) is strongly recommended, and that physicians should not hesitate to consider referral to rheumatology, orthopedic and pain management specialties.

Let’s take a closer look at the topic of choosing an appropriate statistical test as it applies to these articles.  The Ramdass article compared pre- and post-operative radiographic measurements, and used a paired Student t test to do so.  But was this the most appropriate statistical technique to use in this situation?

There is little doubt that the t-test is the most commonly utilized statistical test to compare two groups of numbers in the medical literature.  I recently did a review of the Journal of Foot and Ankle Surgery, and found that nearly 30% of articles used some form of t-test.  This was by far the most commonly reported statistical test in the journal, and other similar reviews of other journals in other specialties have demonstrated the same thing.  But just because the test is the most commonly utilized doesn’t mean that it’s always the most appropriate test to use in all these situations.

The question of which statistical test should be used in which situation is really what began my interest in learning more about statistics.   I read a lot of statistics books, but didn’t really have an answer until I came across a book entitled “Primer of Biostatistics” by Stanton Glantz.   He answered my seemingly simple question with a great chart right on the inside cover of his book. 

This chart accounts for the overwhelming majority of statistical tests reported in the foot and ankle literature, and really helps to answer my question of which statistical test should be used in which situation.

So, did the article use the correct statistical test?  The data description certainly fits under the category of “before and after a single intervention in one group of the same individuals”.  Pre- and post-operative radiographs were compared of patients undergoing a single surgical intervention.  But what type of data was it: interval, ordinal or nominal?  In one of the first issues of this journal club we talked about the difference between quantitative and categorical data.  Radiographic measurements (which were the outcome measure of the article) are all quantitative types of data with equal, ordered and constant intervals.   This leads us to the appropriate use of the paired t-test under the condition that the data is coming from a normally distributed population.  If the data is not from a normally distributed population, then a Wilcoxon signed-rank test should be used.

If the data from my study was normally distributed, then I would have used the correct statistical test, but if my data was not normally distributed, then I used the wrong test.  Let’s take a look at an example and see how I did:

The pre-operative 1^st intermetatarsal angle measurement had an average of 12.8 degrees, a standard deviation of 1.6 degrees and a range of 11-18 degrees.  There’s not an exact mathematical formula to tell us if a data set is normally distributed, but we can make some assumptions based on the given data.  Does 12.8 fall near the center of the 11-18 degree range?  Not really, it’s much closer to 11 than it is to 18.  There’s one red flag.   Another little hint is looking at the standard deviation and the range.  Generally speaking, if two standard deviations above and below the mean falls outside of the reported range, then you are probably not dealing with a normally distributed population.  Two standards deviations (2 x 1.6 = 3.2) below the mean (12.8 – 3.2 = 9.6) falls below the lower limit of our range (11).  This is another red flag that we are probably not dealing with a normally distributed population.

As it turns out, it would probably have been more appropriate to use the Wilcoxon signed-rank test instead of the paired t-test for most of my calculations.  This is a very common mistake in the medical literature unfortunately, and one that I fell right into.   The good news however is that the t-test is a fairly robust test that can make up for our mistakes.   If I ran the same data sets through statistical software and this time use the Wilcoxon signed-rank test instead of the paired t-test, I would get all the same statistically significant differences to about the same power.  It’s just something to keep in mind if you are planning a research project.