Screening athletes for causes of sudden cardiac death (SCD) is difficult. The conditions that cause SCD are rare and sometimes difficult to diagnose, especially in athletes under 35 years old. Many hundreds of athletes must be screened to find even a single case of a potentially lethal heart condition. In many series of SCD in athletes, hypertrophic cardiomyopathy (HCM) is a leading cause (1). There are characteristics of HCM that can be detected by the history (syncope, family history), physical exam (a systolic murmur louder on Valsalva maneuver) and by electrocardiogram (EKG). In addition, it can be detected easily on echocardiography (echo), even a quick screening echo. Therefore, since HCM can be deadly and since it can be detected using information available at a routine screening, the diagnosis must be made. Would an additional tool, such as Bayesian analysis, be helpful in diagnosing HCM during an athlete screening session?
Bayes Theorem is useful when trying to make a diagnosis under uncertainty. It closely follows clinical medicine; as information is received, the probability of the diagnosis is revised either upward or downward. Bayes Theorem requires three inputs: a pretest probability (or prevalence of disease), the sensitivity of a finding or a test and the specificity of the finding or test. It works as follows. A patient comes to a doctor’s office and a disease is suspected. The history is taken and the patient reports a symptom. The doctor knows the prevalence of the disease in the population and looks up the sensitivity and specificity of the finding in picking out the disease entity. The numbers are put in Bayes formula and a posttest probability is calculated. In other words, the patient is suspected of having a disease, reports a symptom consistent with the disease and the doctor’s suspicion of the disease then increases. If the patient has a physical finding consistent with a disease, then the posttest probability just calculated now becomes the pretest probability and the sensitivity and specificity of the physical finding are used to find a new posttest probability. If the patient then has an abnormal EKG, then previous posttest probability and the sensitivity and specificity of the EKG abnormality are used to revise the estimate of disease probability, and so on. As new information is obtained, the previous probability of disease is used to come up with a new estimate- exactly as is done in an office setting- as new information comes to light, the probability of disease goes either up or down. This approach was used by Diamond and Forrester (2) to calculate the probability of coronary artery disease and has been used many times since then. This same method can help diagnose HCM during screening of athletes.
The prevalence of HCM is well known (3). The generally accepted prevalence is that 1 person out of 500 people (0.2%) in the population will have HCM. This has recently been revised and the new estimate is 1 person out of 200 people (0.5%) may have HCM. For the purposes of the screening tool, both numbers are used to provide a range of probabilities.
A detailed history and physical are the cornerstones of the cardiovascular screening of athletes. Every athlete fills out a standard American Heart Association questionnaire and a physical examination is performed. For the purpose of diagnosing HCM, two items on the questionnaire are of interest. A prior history of unexplained syncope may be associated with HCM. This has been studied and it has been determined that the sensitivity for unexplained syncope in diagnosing HCM is 35% and the specificity is 85% (4). Since HCM is a genetic disease and runs in families, the family history is very important. A family history of unexplained SCD has a sensitivity of 42% and a specificity of 79% in diagnosing HCM (4). A family history of HCM carries a sensitivity of 44% and a specificity of 99% (5). Findings on physical examination can also determine the presence of HCM. The classic murmur of HCM is a harsh systolic murmur that gets louder with Valsalva maneuver. The sensitivity of a systolic murmur, louder with Valsalva, is 65% while the specificity is 96% (4). The history and physical examination may not be able to definitively diagnose HCM (the sensitivities are quite low), but if these factors are present, the probability of HCM increases and additional testing is warranted.
The next test during an athletic screening is the EKG. While controversial and not performed routinely in all parts of the world, the EKG should be done if one suspects HCM. Many patients with HCM have abnormal and bizarre EKGs. An EKG is abnormal if it meets the findings of the 2013 Seattle criteria and the updated 2017 International criteria. The sensitivity of an abnormal EKG in diagnosing HCM using the Seattle/International criteria is 93% with a specificity of 96% (6).
Lastly, an echo is often done during screening of athletes. Usually an echo is performed if there is a reasonable probability that a condition which may cause SCD is present. An echo is often used to rule in or rule out a diagnosis of HCM. While the differentiation between HCM and an athlete’s heart on echo can be difficult, at screening one needs to determine if the heart is hypertrophied or not and whether additional testing is necessary. There are many criteria used to diagnose HCM on echo, but three criteria are the generally accepted starting points in making the diagnosis: interventricular septal wall to posterior wall ratio greater than or equal to 1.3, systolic anterior motion (SAM) of the mitral valve and maximal interventricular septal thickness >1.5 cm (7). These three parameters are easily obtained on echo during a screening session for athletes; they don’t require additional expertise by the echo tech or echo reader. An interventricular septal wall to posterior wall ratio greater than or equal to 1.3 has a sensitivity of 76% and a specificity of 93% (7). Systolic anterior motion of the mitral valve has a sensitivity of 82% and a specificity of 99% (7). Maximal interventricular septal thickness >1.5 cm has a sensitivity of 87% and a specificity of 97% (8).
Table of Sensitivities and Specificities in Diagnosing HCM
|
Sensitivity
|
Specificity
|
Unexplained syncope
|
0.35
|
0.82
|
Family History of unexplained SCD
|
0.42
|
0.79
|
Family History of HCM
|
0.44
|
0.99
|
Systolic murmur increased w/Valsalva
|
0.65
|
0.96
|
Abnormal EKG - Seattle/International Criteria
|
0.93
|
0.96
|
Septal/Posterior wall ratio => 1.3
|
0.76
|
0.93
|
SAM
|
0.82
|
0.99
|
Interventricular septum > 1.5 cm
|
0.87
|
0.97
|
How does the Bayes calculator work? Currently, it is a spreadsheet and the relevant factors (ex, family history HCM) are set to a default of 0. If a factor is positive, then the 0 is replaced by a 1 and a new posttest probability is displayed.
Take for example an athlete whose only positive finding is a prior history of unexplained syncope, the physical is normal and the EKG is normal. In this case, the baseline probability of HCM goes from 0.2% - 0.5% to 0.4% - 1%. This is still quite a low probability and if one is wondering whether to do an echo, it may acceptable to skip additional testing.
|
1/500
|
1/200
|
Prevalence of HCM
|
0.002
|
0.005
|
Unexplained syncope
|
1
|
1
|
Family History of unexplained SCD
|
0
|
0
|
Family History of HCM
|
0
|
0
|
Systolic murmur increased w/Valsalva
|
0
|
0
|
Abnormal EKG - Seattle/International Criteria
|
0
|
0
|
|
| |
| | |
Post Test Probability
|
0.004
|
0.010
|
Now if the same athlete has a history of unexplained syncope and an abnormal EKG by Seattle/International criteria, then the probability of HCM goes to 8%- 18%. This is now in an intermediate range and doing an echo would be prudent.
|
1/500
|
1/200
|
Prevalence of HCM
|
0.002
|
0.005
|
Unexplained syncope
|
1
|
1
|
Family History of unexplained SCD
|
0
|
0
|
Family History of HCM
|
0
|
0
|
Systolic murmur increased w/Valsalva
|
0
|
0
|
Abnormal EKG - Seattle/International Criteria
|
1
|
1
|
|
| |
| | |
Post Test Probability
|
0.083
|
0.185
|
If an echo is done and the interventricular septum is greater than 1.5 cm, the posttest probability goes to 72-87%, all but cinching a diagnosis of HCM.
|
1/500
|
1/200
|
Prevalence of HCM
|
0.002
|
0.005
|
Unexplained syncope
|
1
|
1
|
Family History of unexplained SCD
|
0
|
0
|
Family History of HCM
|
0
|
0
|
Systolic murmur increased w/Valsalva
|
0
|
0
|
Abnormal EKG - Seattle/International Criteria
|
1
|
1
|
|
| |
| | |
Post Test Probability
|
0.083
|
0.185
|
| | |
| | |
Echo
| | |
Septal/Posterior wall ratio => 1.3
|
0
|
0
|
SAM
|
0
|
0
|
Interventricular septum > 1.5 cm
|
1
|
1
|
| | |
| | |
| | |
Post Test Probability (+ echo)
|
0.724
|
0.868
|
Post Test Probability ( - echo)
|
0.016
|
0.040
|
The calculator may be helpful in another way. For example, take an athlete who has a family history of HCM and an abnormal EKG. The posttest probability is quite high (67-84%). If, however, the athlete has a negative echo, the posttest probability of a negative echo drops to 27 to 48%. While the probability at this time is lower, this athlete should not be ignored, the probability of HCM is still close to 50%. This scenario represents an athlete who should be followed over time and have repeated echoes. Even though he doesn’t have HCM at present, he is at risk and may develop it in the future. Young athletes with abnormal EKGs and normal echoes may represent an early phase of the disease and it may be evident years later (9). The calculator may help identify these athletes.
Other causes of SCD in athletes are even less prevalent than HCM. In addition, they do not have the same clues on history and physical examination. The EKG can have subtle abnormalities or it can be normal. The echo can be diagnostic, but oftentimes additional testing (stress testing or cardiac MRI for example) is necessary to diagnose one of these conditions. These tests are not part of a screening evaluation for athletes. Take arrhythmogenic right ventricular dysplasia (ARVD) as an example. The prevalence is 1 in 5000 (ten times higher than HCM) (10). Sensitivities and specificities are known for common EKG abnormalities in ARVD (11). The pretest probability of ARVD is very low due to the low prevalence, about 0.02%. Even if the EKG is abnormal, the posttest probability is only 1.3%, still quite low. The Bayesian approach may not be helpful.
ARVD
|
1/5000
|
Prevalence
|
0.000200
|
Positive Family History- ARVD in first degree relative
|
0
|
Inverted T waves in V1, V2 and V3 or beyond in patient > 14 yrs old, w/o RBBB
|
1
|
Epsilon wave in V1, V2, V3, w/o RBBB
|
1
|
|
|
|
|
| |
Post Test Probability
|
0.013598
|
Using the Bayes calculator may be helpful in diagnosing HCM during a routine athlete screening and in identifying athletes who need to be followed closely over time. The tool uses commonly available variables during a typical athlete screening session and takes only a few minutes to use. However, it has not been prospectively validated.
Steven Georgeson, MD FACC FACP
Medicor Cardiology
Atlantic Medical Group
Bridgewater NJ USA
References:
1. Circ 1996;94:850-856
2. NEJM 1979;300:1350-1358
3. NEJM 2018;379:655-668
4. Heart 2004;90:570-575
5. Am J Card 2014;114:1383-1389
6. Br J Sports Med 2018;52:667-673
7. JACC Imaging 2008;1:787-800
8. JACC 1993;22:498-505
9. NEJM 2008;358:152-161
10.NEJM 2107;376:61-72
11.Circ 2009;120:477-487