Common surgery weight-loss percentages overstate results, and none predicted six-month blood sugar

Weight loss after bariatric surgery gets reported in more than one way, and the formula a program picks changes the number a patient hears. Researchers at Brigham and Women's Hospital and Harvard Medical School tested the three formulas in common use against a national registry of 182,826 operations, and reported on September 2 that two of the three overstate how much weight was actually lost.

The analysis used 2023 data from MBSAQIP, the accreditation registry that US bariatric programs report their cases into. It covered sleeve gastrectomy and gastric bypass. Of the 182,826 patients, 136,406 had a body mass index recorded both before surgery and at six months. A smaller group, 27,517 patients, also had a hemoglobin A1c at both points. A1c is a blood test that reflects average blood sugar over roughly the previous three months. The remaining patients were lost to follow-up.

Three formulas give three different answers for the same patient

Percent total weight loss, written %TWL, is the plain version. It is the share of starting body weight that is gone. Someone who started at 300 pounds and now weighs 240 has lost 20% of total weight.

The other two work from excess weight, meaning the amount above a reference weight rather than the whole body weight. Percent excess BMI loss (%EBMIL) and percent excess body weight loss (%EBWL) divide by a smaller starting figure, so they produce larger and more flattering percentages for the identical pounds lost.

The researchers first checked whether each formula moved with a patient's starting BMI. Only %TWL was essentially independent of it, with a correlation of -0.016, close to zero. %EBMIL and %EBWL were strongly tied to the starting point, at -0.458 and -0.337. In plain terms, the same physical weight loss produces a much higher excess-weight percentage in a patient who started at a lower BMI, and a lower one in a patient who started very heavy. Two people can lose the same fraction of their body and be handed very different scores.

In 1.46% of patients, %EBMIL came out above 100%. The authors call that value uninterpretable, because it implies losing more than all of the excess weight there was to lose.

No weight number predicted whether blood sugar improved

The second question was whether any weight formula tells you anything useful about diabetes. The team looked at patients whose blood sugar was uncontrolled before surgery and asked which of them reached an A1c below 7.0% at six months.

All three metrics discriminated poorly, scoring between 0.60 and 0.63 on area under the curve. Area under the curve runs from 0.5 to 1.0, where 0.5 is a coin flip and 1.0 is perfect separation. Using nothing but how severe a patient's metabolic disease was at the start scored 0.768, better than any of the weight numbers.

Weight response and blood sugar response pointed in different directions in 47.4% of patients. Kappa, a statistic that measures agreement beyond what chance alone would produce, was 0.119, near the bottom of its range. Balanced accuracy was 59.2%.

The authors conclude that metabolic outcomes should be measured directly rather than inferred from a weight number, and prioritized in patients who have type 2 diabetes.

Six months is early, and follow-up was patchy

This is a registry study, not a trial. It describes what was recorded in routine US practice in one year, and it cannot establish cause. Six months is early for either weight or diabetes outcomes, and the paper says nothing about how these metrics behave at two years or ten.

Follow-up was incomplete by a wide margin. Only 136,406 of 182,826 patients had a paired BMI, and only 27,517 had a paired A1c, so the diabetes analysis rests on roughly 15% of the starting group. Patients who return for follow-up bloodwork may differ from those who do not.

The finding is also about measurement, not about whether surgery works. Nothing here says surgery produced less weight loss or worse diabetes control than previously reported. It says that one common way of expressing the result inflates it, and that a weight number is a weak stand-in for a blood test.