Episode 1 of Bedside Rounds (which was untitled at this point), in which we use the physical exam to solve a 2000 year-old medical mystery, and learn a little about biostatistics in the process.
Our story starts, as all good stories do, with a paleopathological mystery – the cause of the death of the Roman emperor Hadrian. Hadrian is remembered today for building his eponymous wall in Britain as well as being a humanist and one of the “good emperors”. He ruled for almost 20 years before falling ill at the age of 60 according to Cassius Dio, a historian from the third century. The emperor had long suffered from nosebleeds, but they began to intensify.
“He became consumptive” – referencing body wasting – “as a result of his great loss of blood, and this led to dropsy,” an old fashioned word for edema, or swelling, wrote Cassius. He tried “charms and magic rites” that would relieve the edema temporarily, but inevitably he would fill up again with water. Hadrian became depressed and “longed for death”.
As Cassius reported,
“Finally, he abandoned his careful regimen and by indulging in unsuitable foods and drinks met his death, shouting aloud the popular saying” – and I absolutely love this – “many physicians have slain a king.”
So epistaxis (doctor word for nosebleeds) and edema – admittedly not a lot to go on as far as history goes, but that hasn’t kept physicians from trying. He had been diagnosed by paleopathologists as having alternatively renal failure, or heriditary telangiectasia, a genetic disorder that leads to heavy. But more recently, Dr. Petrakis added a bit of physical exam to Emperor Hadrian’s case. Dr. Petrakis was observing a bust of Hadrian at the Athens National Museum and by chance saw a prominent physical exam finding. He was able to find three additional busts of Hadrian, which all exhibited the same sign. And this physical sign led him to his diagnosis.
What Dr. Petrakis saw was bilateral diagonal earlobe creases, otherwise known as Frank’s sign. A deep crease in the lower lobule of the ear was first noted by Dr. Frank in 1973 in a group of 20 patients who were less than 60 years old and also had cardiac chest pain. Since then, over 50 studies have been published on the diagonal earlobe crease that suggest variably suggest Frank’s sign is an independent predictor of heart disease, independent of other risk factors such as cholesterol, blood pressure or smoking. Of course, science being science, other studies have found that there is no predictive relationship – that Frank’s sign might in fact just be a sign of normal aging.
So how useful is Frank’s sign to clinicians? A 2011 meta-analysis looked at all studies evaluating the earlobe crease as a predictor of coronary artery disease, and found that it was statistically significant. Just how significant it is will take a brief foray into biostatistics, specifically likelihood ratios. A likelihood ratio is the probability of finding a clinical sign in patients with a disease divided by the probability of the same finding in patient’s without a disease. So, for example, if 80% of people with coronary artery disease had Frank’s sign, but only 10% of people without CAD had it, the likelihood ratio would be 8. LRs greater than one argue for the diagnosis; LRs less than 1 argue against it. If the LR is one, you might as well flip a coin. This is all important because we can combine the likelihood ratio with the pre-test probability of a disease to find out how much more like a physical finding makes it that someone actually has the disease. In reality, I never actually run these calculations, but instead use a table that correlates LR with probability changes.
In the case of the diagonal earlobe crease, the meta-analysis found that the likelihood ratio was 2.37, or roughly 17% more likely to have coronary artery disease. In other words, let’s say the middle aged man in my clinic has a 40% change of having coronary artery disease. If, when I examine him, I notice Frank’s sign, I can presume that his chance of having CAD has increased to ~60%. Is this enough to change my management? Probably not. But then again, the likelihood ratio for predicting heart attacks from traditional risk factors such as smoking, high blood pressure, and high cholesterol, was only 1.07 – barely better than a coin flip.
So that takes us back to the poor Emperor Hadrian. His nosebleeds and edema suggest, among other things, heart failure from CAD, renal failure, or, sure, decompesated heart failure from bleeding too much from Osler-Weber-Rendu syndrome. But that extra 17% probability of coronary artery disease from his earlobe creases when he was a much young
Petrakis NL, Diagonal Earlobe Creases, Type A Behavior and the Death of Emperor Hadrian. West J Med. Jan 1980; 132(1): 87–91.
Mark N, Buckley S, The Diagonal Earlobe Crease: Historical Trivia or a Useful Sign of Coronary Artery Disease? http://www.clinicalcorrelations.org/?p=4927