Day et al developed a prediction model for erectile dysfunction in men based on logistic regression analysis. This uses recognized risk factors to identify men who may benefit from therapeutic interventions. The authors are from Pfizer, Inc. (makers of Viagra), in New York City.
Parameters:
(1) age
(2) cigarette smoking
(3) diabetes mellitus
(4) depression
(5) hypertension
(6) prostate
(7) serum cholesterol
Parameter |
Finding |
Points |
---|---|---|
age of the patient in years |
|
|
cigarette smoking |
no |
0 |
|
yes |
1 |
diabetes mellitus |
absent |
0 |
|
present |
1 |
depression |
absent |
0 |
|
present |
1 |
hypertension |
absent |
0 |
|
present |
1 |
prostate disease |
absent |
0 |
|
present |
1 |
serum cholesterol |
not elevated |
0 |
|
elevated |
1 |
where:
• The specific aspects of prostate disease were not defined.
• Diabetes mellitus was not subdivided. Different levels of smoking, depression and hypertension were not identified.
• Prostate disease could range from mild prostatitis to radiation for prostate cancer.
• The threshold for an elevated serum cholesterol was not stated. Either a value > 200 mg/dL or > 250 mg/dL might be used.
Z =
= (0.066 * (age in years)) + (0.486 * (points for smoking)) + (1.239 * (points for diabetes)) + (0.781 * (points for depression)) + (0.48 * (points for hypertension)) + (0.727 * (points for prostate disease)) + (0.42 * (points for high serum cholesterol)) - 3.462
where:
• The published equation used "+ 3.462". However, when this is used it can be seen that the probability estimates are too high. Subtracting the value results in more reasonable results (the values for age alone are similar to that shown on page 135).
probability of erectile dysfunction =
= 1 / (1 + EXP((-1) * Z))
Limitations:
• The lack of specific cutoffs for some of the risk factors would seem to make the estimates subject to interobserver variability.
• Erectile dysfunction is not characterized as to its severity.
Performance:
• The sensitivity was 82% and specificity 61%, with an overall accuracy of 74%. These numbers might be acceptable for a screening test.
Purpose: To screen a man for the risk of having erectile dysfunction using the prediction model of Day et al.
Specialty: Urology
Objective: risk factors
ICD-10: F52, N48,