Sunday, April 05, 2015

Pre-Diabetes and the Diabetes Risk Prediction Tool

Until recently the USPSTF did not recommend screening for diabetes in asymptomatic adults with BP of 135/80 or less. There’s now a draft recommendation, however, that basically recommends screening everyone over 45 plus lots of people under 45:

… 45 years or older, overweight or obesity, or a first-degree relative with diabetes. Women with a history of gestational diabetes or polycystic ovarian syndrome … African Americans, American Indians/Alaska Natives, Asian Americans, Hispanics/Latinos, and Native Hawaiians/Pacific Islanders …

In other worlds, only skinny euros under 45 avoid screening.

Which makes me feel a bit better about our typically shotgun “workplace health” program. It’s been doing my fasting glucose for about 3 years. A glucose that’s consistently about 100-103 mg/dl (jumped from 82 in 2011 to 99 15 months later, which makes me wonder about something immune beating on beta cells). [2][3]

I should be under 100, typically under 90, so that’s not a good number. It’s in the “pre-diabetes” range, and the current enthusiasm is to start treatment with meds in the hope of delaying the onset of “true” diabetes. Of course there’s also the theoretical benefit of weight loss and exercise, but few people manage that. Unfortunately we know weight loss programs rarely work, and the reports of med efficacy smell dubious to me, so it’s not entirely clear how useful the “pre-diabetes” diagnosis really is. The diagnosis, of course, is likely to raise one’s health insurance costs, though ObamaCare helps somewhat.

Since I’m already a skinny fitness nut my family doc wanted to completely ignore this, presumably on the grounds that there’s nothing useful to do about it. A wise recommendation, but I’m not built that way. 

Poking around the net I found a BMJ article on a statistical model that tries to put some personalized precision medicine context around that fasting glucose. The researchers settled on 7 factors [1] that seemed to predict 3 year conversion to diabetes (yeah, only 3 year range). So I ran the Diabetes risk prediction tool on myself. 

I got 91 points; the instrument recommends consideration of preventive steps for scores of 146 and up. Which vindicates my FP’s intuition (maybe she does the instrument in her head?).

Figuring out what 91 points means for 3 year conversion to Diabetes is a lot harder. I didn’t find anything that mapped scores onto risk quartiles! The data on 3 y progression ti diabetes per quartile was, very roughly:

  • 1st quarter: 10%
  • 2nd quartile: 20%
  • 3rd quartile: 30%
  • 4th quartile: 60%

I’m guessing my odds are probably in the 25-30% range over 3 years. So pretty high over 10 years, but the bottom line is that there’s really not enough data to justify taking an (invariably) icky medicine.

So I’ll just keep playing with my glucometer [2]…

[1] The HbA1c (glycosylated Hb) value is weird in this study. Normal HbA1c is usually given as 4.0-5.9%, but in this study anything above 4 starts to pile on bad points. This could be either a sign that there’s something funky with their model, or a sign that we should redefine the normal range for HbA1c (I’m 5.2, “normal”, but on their chart it lines up with a fasting glucose of about 105).

Lastly, the paper text says “baseline fasting glucose was by far the most important predictor” which does make me wonder if the other 6 factors mean anything.

[2] Since I’m kind of curious about what’s going on I bought a glucometer on Amazon (I paid $28 or so, it’s $11 today. They make their money on the proprietary strips.) My 1-2h post-prandial glucose is identical to my fasting glucose — 103. “Normal” is less than 140. So that’s weird. Next time i’ll redo the standard sample test — maybe the glucometer is dead. Or maybe the FBG/PPG ratio is an interesting predictor  …

[3] I do idly wonder about getting a TSH (thyroid stimulating hormone).

See also

No comments: