The math of invasive breast cancer risk for LCIS

@elsie37
I have a math background and find the statistics for cancer hard to understand. (Which I'd never looked at, coming from a cancer-free family, until a bad news biopsy last year led to a lumpectomy.)

Here's one example that makes no sense. Two oncologists I saw said that the rule of thumb is that taking an aromatase inhibitor could [note: "could".] lower my risk of recurrence by approximately 45% in general. But that was before the OncotypeDX showed a risk of recurrence of approximately 5.5% if I didn't take them so I didn't have the 'industry accepted' higher risk that the 45% 'could' lower.

The rule of thumb of 45% risk reduction in favor of aromatase inhibitors for ER+, PR+, HER2- breast cancers is usually rounded up to 50% as it's easier to quote.

BUT, one article in the respected New England Journal of Medicine by an oncologist, a leader in his field, noted that the NNT number (aka 'number needed to treat' to prevent projected cancer) for aromatase inhibitors is 49. And his point was that it's a serious ethical question of whether 48 women should be told to take a 'toxic' (his word) drug that they may not even benefit from. And that patients should be fully informed of the side effects and statiscally-derived likely benefit so each can decide whether to take them. And that failing to adequately inform breast cancer patients is further fostering the environment of patient non-compliance or, worse, loss of confidence in physicians at a time when they most need to have it.

What bothers me is that 1 in 49 does not yield a statistic anywhere near the 45% risk reduction talked about. So how does 1 in 49, rounded to 2%, become that near-50%?

I have an M.B.A. in finance and am used to double-checking pro forma numbers and too-rosy projections so this caught my attention and I've not yet found the explanation unless there's some arcane stat to explain it, e.g., the drugs prevent 100% of recurrences for some kind of cancer that accounts for, say, 70+% of cancers to bring the mean (average) risk reduction up to 45-50%?

I'm sure there are benefits (and known downsides) to aromatase inhibitors. But getting clean data points to connect and base a decision on is oddly difficult.

I skipped the drugs. A friend asked if, if I suffer a recurrence, I'll think that not taking them caused it. And, upon thinking it over, my answer would be no. I'd have a 3% risk of recurrence if I'd done the trifecta of radiation, chemo and adjuvant anti-hormone therapy (and get osteoporosis from the latter, at a minimum). So I'd never know what 'caused' the recurrence except that something is very wrong when 1 in 8 women in this country is predicted to get breast cancer in her lifetime in the first place...

@elsie37
I have a math background and find the statistics for cancer hard to understand. (Which I'd never looked at, coming from a cancer-free family, until a bad news biopsy last year led to a lumpectomy.)

Here's one example that makes no sense. Two oncologists I saw said that the rule of thumb is that taking an aromatase inhibitor could [note: "could".] lower my risk of recurrence by approximately 45% in general. But that was before the OncotypeDX showed a risk of recurrence of approximately 5.5% if I didn't take them so I didn't have the 'industry accepted' higher risk that the 45% 'could' lower.

The rule of thumb of 45% risk reduction in favor of aromatase inhibitors for ER+, PR+, HER2- breast cancers is usually rounded up to 50% as it's easier to quote.

BUT, one article in the respected New England Journal of Medicine by an oncologist, a leader in his field, noted that the NNT number (aka 'number needed to treat' to prevent projected cancer) for aromatase inhibitors is 49. And his point was that it's a serious ethical question of whether 48 women should be told to take a 'toxic' (his word) drug that they may not even benefit from. And that patients should be fully informed of the side effects and statiscally-derived likely benefit so each can decide whether to take them. And that failing to adequately inform breast cancer patients is further fostering the environment of patient non-compliance or, worse, loss of confidence in physicians at a time when they most need to have it.

What bothers me is that 1 in 49 does not yield a statistic anywhere near the 45% risk reduction talked about. So how does 1 in 49, rounded to 2%, become that near-50%?

I have an M.B.A. in finance and am used to double-checking pro forma numbers and too-rosy projections so this caught my attention and I've not yet found the explanation unless there's some arcane stat to explain it, e.g., the drugs prevent 100% of recurrences for some kind of cancer that accounts for, say, 70+% of cancers to bring the mean (average) risk reduction up to 45-50%?

I'm sure there are benefits (and known downsides) to aromatase inhibitors. But getting clean data points to connect and base a decision on is oddly difficult.

I skipped the drugs. A friend asked if, if I suffer a recurrence, I'll think that not taking them caused it. And, upon thinking it over, my answer would be no. I'd have a 3% risk of recurrence if I'd done the trifecta of radiation, chemo and adjuvant anti-hormone therapy (and get osteoporosis from the latter, at a minimum). So I'd never know what 'caused' the recurrence except that something is very wrong when 1 in 8 women in this country is predicted to get breast cancer in her lifetime in the first place...

Often the numbers you read are from research of different sources. You might want to consider that research results will be different as it is taken from different research that won’t be the same just as individual people are different. You might think of it this way: Research results will show different outcomes due to the individuals who are studied having many differences on the whole for instance perhaps where they live. You won’t find the exact numbers of such things on the results of different research. For just one small example from a research project might be based on participants being primarily male with ages up to fifty and what were the participants exposed to, what known illnesses were present in their family members, etc. Research cannot include all participants who have exactly the same ancestors, background, other illnesses and exposures. Good research is going to show different results and new questions, but somewhere something will come up in research that can be useful. That may then lead us pointing into another direction to implement toward cause or cure.

In addition, breast cancer projects tend to lump together patients with all sorts of different tumor characteristics at diagnosis. For example, a patient with a Stage 1 diagnosis of a Grade 1 tumor and no node involvement is going to have a different prognosis than a patient with a Stage 2 diagnosis of a Grade 3 tumor with 3 positive nodes. But they are often both included in the same study. Unfortunately, a statistical analysis has certain unavoidable limitations.