“It’s tough to make predictions, especially about the future”
The quote about predictions is from noted philosopher (and Yankees Hall of Famer) Yogi Berra. One person who can help us make predictions, Thomas Bayes, was a Presbyterian minister. He was born in 1701 in England and studied theology and logic at the University of Edinburgh. He worked as a minister but continued to dabble in questions of logic until he died in 1761. After he passed away, his family found a paper that he wrote and brought it to the attention of the Royal Society. The paper described what came to be known as Bayes Theorem. Bayes Theorem launched a new way of thinking and Bayesian probability has become the core of modern statistics. In 2018, the University of Edinburgh dedicated their informatics department to Reverend Thomas Bayes. Bayes Theorem gives a way to work out conditional probabilities. A conditional probably is the probability of an event A, given that condition B is true; what is the probability of A given B. In other words, it takes a belief, evaluates new evidence and gives a new, improved belief. Scientists believe that human brains use a type of Bayesian thinking to deliberate and decide. Bayesian reasoning has exploded with uses in many diverse areas including physics, ecology, psychology, artificial intelligence and even self-driving cars. One of the fields where Bayes Theorem is most useful is in medicine.
Consider the case of a 50-year-old man who comes to his doctor’s office for evaluation. What is the probability of coronary artery disease (CAD) for this man? In 1979, the New England Journal of Medicine published a landmark study introducing Bayes Theorem to the medical community. The article asked exactly this question; how can the probability of CAD be calculated in a 50-year-old male patient? From population studies and autopsies, the probability of CAD for a 50-year-old male is 7% (for comparison the probability of CAD in a 50-year-old woman is 1.6%, for a 65-year-old man it is 12%). Next, we learn of a new finding, our 50-year-old man has been having chest pain. Suppose he is having symptoms that are typical of angina pectoris. We know that being 50 years old and a male he has a probability of CAD of 7%. Now we have new information, he has chest pain, typical angina. We can evaluate with Bayes equation and find that his probability of CAD is now 90%. If his symptoms are more atypical, then his probability of CAD is only 53%. If he has atypical chest pain, but has a positive stress test, then his probability goes back up to 70%. Contrast that with a 30-year-old man who has a low baseline probability (0.03%). Even if the 30-year-old has typical chest pain, his probability of CAD is only 26%. This is the power of Bayes, the probability of a disease can be revised up and down based on each piece of new information.
The most important factor in determining the risk for CAD is baseline characteristics (pre-test probability). Age and sex give a crude estimate. More sophisticated calculators have been developed that incorporate more variables (for example diabetes, blood pressure, cholesterol, smoking, family history). These include the Framingham risk score, the American College of Cardiology (ACC) risk calculator, the Pooled Cohort equations and the most recent iteration, the PREVENT equations. These tools are accurate but they have drawbacks. They give a 10-year estimate for the risk for a heart attack (rather than providing a risk for CAD right now) and are not useful in all populations (for example, they are less accurate in lower socioeconomic groups). In addition, they may not be able to predict a first heart attack. A study analyzed 465 patients under 65 years old without known CAD who presented with a heart attack. The patients’ information was entered into the ACC risk estimator and the PREVENT calculator. The researchers found that 45% of the patients would have classified as low risk with ACC calculator, meaning no further testing or statin advised. For the PREVENT calculator, 61% would have been classified as low risk. Predicting the future is hard.
The next most important factor is symptoms (for example, chest pain). Remember that our 50-year-old man (baseline 7%) having typical chest pain, has a probability of CAD of 90%. The characterization of the chest pain is vitally important as it will shift the probability (and therefore the work up and treatment) of CAD significantly. Typical angina pectoris (chest pain) is exertional, described as a tightness or squeezing or pressure or heaviness or dull aching located in the area of the breast bone. The pain can radiate to the left arm, neck, jaw or back. It will usually last a few minutes and be relieved by rest. Atypical pain is sharp or stabbing, located on the right side and brought on by deep breathing or change in position. There is a substantial population of patients who continue to have typical angina despite having no severe blockage on heart catheterization. This problem has vexed clinicians for more than 40 years. After years of research, we now have a better understanding of this syndrome (microvascular dysfunction) and a catchy moniker, ANOCA (Angina, NOrmal Coronary Arteries). Catheterization will evaluate for blockage in the main heart arteries, but catheterization only shows about 10% of the total heart circulation. The rest consists of small vessels, the microvasculature. The small vessels are the primary site of regulation of blood flow to the heart muscle. When the heart needs more oxygen (for example while running) the microvasculature dilates and the flow increases. With ANOCA, that regulation is abnormal and patients will have classic chest pain when the demand is greater than the supply. ANOCA patients have predictable triggers of chest pain such as exertion or emotional stress. It is more common in women, in smokers, and people with hypertension or high cholesterol. ANOCA is common, occurring in about 25% of all heart catheterizations and affecting about 3 to 4 million Americans. It is diagnosed after the large heart arteries are found to be normal. A stress test with PET imaging or heart catheterization with specialized testing will diagnose ANOCA.
So, if you have chest pain, you can calculate your probability of CAD by entering your variables into Bayes equation, or enter the data into a Bayes app (it does exist!) or checking a table. However, the prudent course would be to call your doctor and ask what tests are necessary. If you continue to have classic angina despite normal arteries on heart catheterization, then seek out a cardiologist specializing in ANOCA (they do exist!).




