Eddy Edson
Well-Known Member
- Relationship to Diabetes
- In remission from Type 2
Some nerdifying, triggered by somebody querying their retinopathy screening result ...
It's very common to see people getting a positive result on a periodic screen for mild background non-proliferative diabetic retinopathy (NPDR), which goes away the next time they get tested. You often see people interpreting that as the retinopathy going away, but it's more likely to indicate a false positive on the first test.
You can back-of-envelope the chances of a false positive using the formula for "Positive Predictive Value" (PPV):
Given a positive test result, the PPV is the probability that the result is actually a true positive. So false positive probability is 1 - PPV.
For fundus retinopathy screening:
- Prevalence of mild background NPDR (for T2D's, in this) depends a lot on BG/BP/cholesterol, age, duration of diabetes etc but a reasonable back-of-envelope seems to be around 30%. Note that adding in more severe forms of retinopathy would increase that by quite a lot.
- Sensitivity is the probability of accurately detetcting a positive case. It depends a lot on methodological and definitional details, equipment, operator skill etc, but for standard fundus screening for mild NPDR it's apparently around 80%. (Sensitivity increases as retinopathy becomes more severe, and easier to detect accurately).
- Specificity is the probability of accurately detecting a negative case, apparently around 85%.
Plugging those numbers in to the formula you get PPV = 70%. Which means that if you get a positive result for "r1", mild background NPDR, there's a 30% chance it's a false positive.
If you think in terms of repeated tests over the years, even without any development of retinopathy, the probability of getting at lease one false positive result over 5 annual tests is 1-0.70^5 = 84%. In other words, it would be unusual *not* to get a false positive result over this period. Over 10 years, the probability would be 97% - almost certain.
How acceptable is this? It's probably OK, because mild NPDR isn't serious by itself, doesn't require any particular treatment, and is handled just by keeping the usual risk factors under control, which is the goal for general good health anyway. Ideally, the result would just serve to make good management a bigger priority. (What is not OK is nannying messaging which overstates the seriousness and definiteness of the screening result.)
Also, if the screening finds more serious forms of retinopathy which require some more intensive type of management, the chances of a false positive are much lower, because sensitivity and specificity increase with severity of retinopathy.
It's very common to see people getting a positive result on a periodic screen for mild background non-proliferative diabetic retinopathy (NPDR), which goes away the next time they get tested. You often see people interpreting that as the retinopathy going away, but it's more likely to indicate a false positive on the first test.
You can back-of-envelope the chances of a false positive using the formula for "Positive Predictive Value" (PPV):
Given a positive test result, the PPV is the probability that the result is actually a true positive. So false positive probability is 1 - PPV.
For fundus retinopathy screening:
- Prevalence of mild background NPDR (for T2D's, in this) depends a lot on BG/BP/cholesterol, age, duration of diabetes etc but a reasonable back-of-envelope seems to be around 30%. Note that adding in more severe forms of retinopathy would increase that by quite a lot.
- Sensitivity is the probability of accurately detetcting a positive case. It depends a lot on methodological and definitional details, equipment, operator skill etc, but for standard fundus screening for mild NPDR it's apparently around 80%. (Sensitivity increases as retinopathy becomes more severe, and easier to detect accurately).
- Specificity is the probability of accurately detecting a negative case, apparently around 85%.
Plugging those numbers in to the formula you get PPV = 70%. Which means that if you get a positive result for "r1", mild background NPDR, there's a 30% chance it's a false positive.
If you think in terms of repeated tests over the years, even without any development of retinopathy, the probability of getting at lease one false positive result over 5 annual tests is 1-0.70^5 = 84%. In other words, it would be unusual *not* to get a false positive result over this period. Over 10 years, the probability would be 97% - almost certain.
How acceptable is this? It's probably OK, because mild NPDR isn't serious by itself, doesn't require any particular treatment, and is handled just by keeping the usual risk factors under control, which is the goal for general good health anyway. Ideally, the result would just serve to make good management a bigger priority. (What is not OK is nannying messaging which overstates the seriousness and definiteness of the screening result.)
Also, if the screening finds more serious forms of retinopathy which require some more intensive type of management, the chances of a false positive are much lower, because sensitivity and specificity increase with severity of retinopathy.
Last edited:
