4 ms·

Scientists develop 10-minute universal cancer test

8 points by charlysl 8y ago

melling 8y ago
“The test has a sensitivity of about 90%, meaning it would detect about 90 in 100 cases of cancer, with 10% false positives.” That’s a large error rate.
- craftyguy 8y ago
  The very next sentence: > It would serve as an initial check for cancer, with doctors following up positive results with more focused investigations. As a cheap, quick way to approximate whether or not a patient has cancer, I think think is a huge improvement over the current method of relying on many expensive, slow tests. That does not scale, and I bet many folks that cannot afford it die because they have no way of knowing, with any percent accuracy, whether they have cancer or not.
- x1798DE 8y ago
  It's particularly bad because the baseline rate for cancer is pretty low. The incidence rate for men and women in the US from [1] is 439 per 100k. So taking into account the baseline rate: Of 439 people with new cancers, 395 (90%) will get a positive result. Of 99,561 people without cancer, 9,956 (10%) will get a positive result. So a total of 10,351 people will get positive results, and 96% of those people will not have cancer, so a positive result means that you have (to a first approximation) a 4% chance of having cancer. If the breakdown from [2] is roughly accurate, taking into account no other risk factors, someone under 50 can expect an even lower baseline rate, the "true positive" rate of a test with 90% sensitivity would range from 0.1% in infants to 25% in people 90+ years of age. 1. https://www.cancer.gov/about-cancer/understanding/statistics https://www.cancer.gov/about-cancer/understanding/statistics 2. https://www.cancerresearchuk.org/health-professional/cancer-statistics/incidence/age#heading-Zero https://www.cancerresearchuk.org/health-professional/cancer-...
  - cuchoi 8y ago
    That's a great point but I imagine doctors won't start running this test in a random sample of the population. They would probably run it in people with symptoms --> makes them a lot more likely to have cancer.
- ericb 8y ago
  I've always wondered, if the error rate for a test like this is high but random, could you just run it 3 times? Does it vary randomly?

8 points by charlysl 8y ago

“The test has a sensitivity of about 90%, meaning it would detect about 90 in 100 cases of cancer, with 10% false positives.” That’s a large error rate.

craftyguy 8y ago
The very next sentence: > It would serve as an initial check for cancer, with doctors following up positive results with more focused investigations. As a cheap, quick way to approximate whether or not a patient has cancer, I think think is a huge improvement over the current method of relying on many expensive, slow tests. That does not scale, and I bet many folks that cannot afford it die because they have no way of knowing, with any percent accuracy, whether they have cancer or not.
x1798DE 8y ago
It's particularly bad because the baseline rate for cancer is pretty low. The incidence rate for men and women in the US from [1] is 439 per 100k. So taking into account the baseline rate: Of 439 people with new cancers, 395 (90%) will get a positive result. Of 99,561 people without cancer, 9,956 (10%) will get a positive result. So a total of 10,351 people will get positive results, and 96% of those people will not have cancer, so a positive result means that you have (to a first approximation) a 4% chance of having cancer. If the breakdown from [2] is roughly accurate, taking into account no other risk factors, someone under 50 can expect an even lower baseline rate, the "true positive" rate of a test with 90% sensitivity would range from 0.1% in infants to 25% in people 90+ years of age. 1. https://www.cancer.gov/about-cancer/understanding/statistics https://www.cancer.gov/about-cancer/understanding/statistics 2. https://www.cancerresearchuk.org/health-professional/cancer-statistics/incidence/age#heading-Zero https://www.cancerresearchuk.org/health-professional/cancer-...
- cuchoi 8y ago
  That's a great point but I imagine doctors won't start running this test in a random sample of the population. They would probably run it in people with symptoms --> makes them a lot more likely to have cancer.
ericb 8y ago
I've always wondered, if the error rate for a test like this is high but random, could you just run it 3 times? Does it vary randomly?