Alex J. Kline, MD; Gary S. Gruen, MD; Hans C. Pape, MD; Iva Early complications following the operative treatment of pilon fractures with and without diabetes. Foot Ankle Int.  2009 Nov; 30(11): 1042-7.   (PubMed ID: 19912712)

WHY did the authors undertake this study?
The worldwide epidemic of diabetes has certainly influenced all medical and surgical fields, orthopedic traumatology not excepted.  These authors have previously demonstrated the increased risk of complications in diabetic patients following operative treatment of ankle fractures, and now aim to examine complication rates associated with operative treatment of pilon fractures.

HOW did they attempt to answer this question?
The primary outcome measures of the study were the development of superficial infection, deep infection, surgical wound healing complications, nonunion and delayed union following operative intervention for intra-articular pilon fractures.  Other examined measures included the development of Charcot arthropathy, the requirement for amputation, requirement for early ankle arthrodesis, and a broad range of patient demographic data. 

Patients who met inclusion criteria were divided into two groups:  diabetics (n=14 fractures) and non-diabetics (n=69 fractures).  The specific surgical intervention was at the discretion of the surgeon, and all patients followed a relatively similar post-operative course.  All information was collected via retrospective chart review.

WHAT were the specific results? 
Statistically significant differences were found with respect to infection rate (both superficial and deep) and non-union/delayed union rate between the two groups.  Diabetic patients had an infection rate of 71% compared to 19% in non-diabetic patients (p < 0.001, odds ratio 10.7 [95% confidence interval 2.9-39.8]), while diabetic patients had a non-union/delayed union rate of 43% compared to 16% in the non-diabetic group (p < 0.001, odds ratio 3.96 [95% confidence interval 1.1-13.7]).

HOW did the authors interpret these results?
From these results, the authors concluded that, similar to ankle fractures, operative intervention of traumatic pilon fractures in diabetic patients carries significant risk for complication. 

Milne WK, Worster A.  Evidence-based Emergency Medicine/Rational Clinical Examination Abstract. Does the clinical examination predict lower extremity peripheral arterial disease?  Ann Emerg Med.  2009 Nov; 54(5): 748-50. (Pubmed ID#: 19185391).

WHY did the authors undertake this study?
This is an interesting and important running segment within the Annals of Emergency Medicine where an emergency department physician will evaluate some portion of the patient clinical exam from an evidence-based standpoint.  An article from the Journal of the American Medical Association (JAMA) related to the topic is reviewed, and then the author provides a “selection of Evidence-Based Medicine Teaching Points”.   In this case the clinical question presented is “Does the clinical examination predict lower extremity peripheral artery disease?” taken from an article of the same title originally published in a 2006 edition of JAMA (Original Pubmed ID#:  16449619).

HOW did they attempt to answer this question?
Measures of sensitivity, specificity, predictive values, likelihood ratios (LR) and confidence intervals (CI) were abstracted from a review of 17 articles meeting inclusion criteria.  Physical examination measures included the presence of a femoral bruit, pulse abnormalities, presence of cool skin and the presence of claudication.  The diagnosis of peripheral arterial disease (PAD) was made with the ankle-brachial index (ABI), duplex sonography and/or angiogram.

WHAT were the specific results?
In asymptomatic patients with PAD, the most useful positive clinical finding was the presence of a femoral bruit  (+LR 4.8; 95% CI 2.4-9.5).  Any pulse abnormality (+LR 3.1; 95% CI 1.4-6.6) and claudication (+LR 3.3; 95% CI 2.3-4.8) also demonstrated positive likelihood ratios. 

In asymptomatic patients without PAD, the most useful negative clinical findings were the absence of claudication (-LR 0.57; 95% CI 0.43-0.76) and the absence of pulse abnormality (-LR 0.44; 95% CI 1.4-6.6)(in other words, the patient has palpable pulses).

HOW did the authors interpret these results?
Although the diagnosis of PAD cannot be made solely with the physical examination, it does provide valuable information that can lead to the diagnosis.  PAD is more likely in the setting of femoral bruits, pulse abnormalities, and claudication, whereas PAD is less likely in the absence of bruits and pulse abnormalities.                                                                                                                      

Let’s take a closer look at the topic of odds ratio analysis as it apply to these articles, specifically what information this test is giving you as a critical reader and how to interpret it.  The Kline et al article attempted to determine a difference in a variable (complication rate) between two different groups of patients (diabetics and non-diabetics) undergoing a similar intervention (operative treatment of pilon fractures).  Using infection rate as an example, they concluded that during their observation period the infection rate of diabetic patients was 71% whereas the infection rate of non-diabetic patients was 19%.  Seventy-one percent sure seems like a bigger number than 19%, but are these two numbers really “different” from a statistical standpoint?  To answer this question the authors (appropriately) chose to perform an odds-ratio analysis, and calculated the following statistical endpoint:  “[p < 0.001, odds ratio 10.719 (95% confidence interval 2.914-39.798)]”.  OK, I can see that the p-value is less than 0.05, but what are all these numbers really telling us?  Let’s break them down one at a time.

First, the calculated value for the odds ratio was 10.719.  An appropriate interpretation of this number for this study can be summarized by the following sentence:  “Diabetic patients undergoing operative intervention for pilon fractures are 10.719 times as likely to have a post-operative infection compared to non-diabetic patients undergoing operative intervention for pilon fractures.”  Does “10.719 times” give you clinically more information than “10.7 times” or even “11 times”?  Not really, but we do want to pay attention to this number in relation to “1”.    If our calculator (or statistical software) spits out “1” for the odds ratio value, it tells us that the two groups have equal odds for the given variable.  In our situation you could say that “diabetic patients undergoing operative intervention for pilon fractures are as likely to have a post-operative infection compared to non-diabetic patients”.  An odds ratio value that is close to “1” means that both groups (diabetic and non-diabetic patients) have equal odds of having the given variable (post-operative infection). 

However, the greater that the odds ratio value is above “1”, then the higher the odds are of the first group having the given variable compared to the second group.  And the lower that the odds ratio value is below “1”, then the lower the odds are of the first group having the given variable compared to the second group.  An odds ratio value can never be lower than zero, but can theoretically be infinitely high. 

Second, we are told that the 95% confidence interval of the odds ratio value is between 2.914 and 39.798.  An appropriate interpretation of this number for this study can be summarized by the following sentence:  “Given this data, we can be 95% sure that the actual value of the odds ratio is between 2.914 and 39.798.”  We can never be 100% sure what the exact value is of the odds ratio for this situation because we don’t have information from the operative intervention of all diabetic pilon fractures and all non-diabetic pilon fractures.  All we have in this study is information from the operative intervention of 14 diabetic fractures and 69 non-diabetic fractures.  So based on this limited data, we can be confident that 95 times out of 100, the actual number will be within that range.  This concept comes across in a lot of statistical analyses; We can never know what the “exact” number is, but we can be pretty confident in our estimation of the number.

Once again, does “2.914-39.798” give you clinically more information than “2.9-39.8” or even “3-40”?  Not really, but we do want to pay attention to this number in relation to “1”.  The range of the confidence interval does not include 1, so we know that there will be a degree of statistical significance to the odds ratio.  Any time the confidence interval of an odds ratio does not include 1 (whether all above or all below), we can be certain that there is statistical significance to that odds ratio.  Put in another way, any time the confidence interval of an odds ratio does include 1, we can be certain that there is not statistical significance to that odds ratio.

And finally, we are told that the p-value for the odds ratio is less than 0.001.  This confirms the statistical significance that we implied from the range of the confidence interval.  It also allows us to say that based on the given data, greater than 99.9 times out of 100, 71% is actually different from a statistical standpoint than 19%.   That’s a lot of math to tell us that a diabetic patient is more likely to have a post-operative infection!