Bariatric Surgery versus Intensive Medical Therapy for Diabetes — 3-Year Outcomes
2014 · paper · Clinical trials
Reports three-year results of a randomized trial of 150 obese patients with uncontrolled type 2 diabetes, comparing intensive medical therapy alone with added gastric bypass or sleeve gastrectomy. Target blood-sugar control was met by 5% (medical), 38% (bypass) and 24% (sleeve).
2 findings
Substance Overall baseline BMI is printed as "36.0±3.5" (abstract) and "36±3.5" (Results, 'In the current analysis'), but Fig. 1D gives group baseline means of 36.4, 36.1 and 37.1, all above 36.0.
- p. 1 Abstract, Results
meters) was 36.0±3.5
- p. 6 Figure 1D, 'Value at Visit' baseline column
36.4
Only the authors can say what is right here.
Judge's note: Checked the abstract crop and the upright Fig. 1D. The baseline 'Value at Visit' entries are 36.4 (medical), 36.1 (sleeve) and 37.1 (bypass). A pooled mean cannot fall below the smallest group mean, so 36.0 cannot come from the same patients. Fig. 1A baseline HbA1c means (9.0/9.5/9.3) match Table 1 exactly, which suggests the figure uses the 137-patient analysis cohort. The paper does not say that the overall BMI figure uses a different population. Only the authors can say which value is right.
Second check: another model agreed
Report a problem with this finding
Substance The text says "with 43% of patients requiring insulin at baseline" in 'the current analysis', but Table 2 gives baseline insulin use of 21/40, 22/48 and 22/49 (65/137, 47%).
- p. 3 Results, Study Patients
with 43% of patients requiring insulin at baseline
- p. 5 Table 2, Insulin row, At Baseline
21 (52)
Only the authors can say what is right here.
Judge's note: Read the p3 text and the Table 2 crop. The same sentence uses the 137-patient cohort ('49 patients (36%)' = 49/137). Table 2's baseline insulin counts sum to 65/137 = 47.4%. No other population is stated for the 43%.
Second check: another model agreed
Report a problem with this finding
5 minor (typos and the like; most checked by the judge only)
Reference or label The Fig. 1 legend says "P values are for the comparison between each surgical group and the medical-therapy group in Panels A, C, and D", but Panels C (P=0.02) and D (P=0.006) also show bracketed P values joining the sleeve-gastrectomy and gastric-bypass curves.
- p. 5 Figure 1 legend
between each surgical group and the medical-therapy group in Panels A, C, and D
- p. 6 Figure 1C and 1D, bracketed P values between the sleeve-gastrectomy and gastric-bypass curves
P=0.006
Judge's note: In the upright Fig. 1, Panel C has a bracket labelled P=0.02 spanning the sleeve and bypass points at 36 months, and Panel D has a bracket labelled P=0.006 spanning the same two curves. Each panel also shows two P<0.001 values against medical therapy. The legend does not mention the surgical-vs-surgical comparisons.
Second check: another model agreed
Report a problem with this finding
Wording The Discussion says "More than 90% of surgical patients had glycemic control without the use of insulin". Table 2 supports >90% not on insulin (90/97), but Table 1 shows only 65% of surgical patients at HbA1c ≤7%.
- p. 7 Discussion, first paragraph
than 90% of surgical patients had glycemic con-trol without the use of insulin
- p. 5 Table 2, insulin at 3 years
3 (6)†
- p. 3 Table 1, ≤7%
31 (65)
Only the authors can say what is right here.
Judge's note: Table 2 at 3 years shows insulin use of 3/48 and 4/49, so 92.8% of surgical patients were not on insulin. Table 1 gives ≤7% for 31/48 and 32/49 (64.9%), and ≤6% for 18 and 12. The sentence as worded attributes 'glycemic control' to >90%, which the tables do not support. It probably means 'without insulin', but the authors should clarify.
Second check: another model agreed
Report a problem with this finding
Wording The 1-year recap "gastric bypass (42% of patients) or sleeve gastrectomy (37%) than after intensive medical therapy (12%) (P<0.001 for both comparisons)" gives proportions and P values that do not correspond under the declared chi-square test at these group sizes.
- p. 10 Discussion, first paragraph (recap of the 1-year STAMPEDE results)
intensive medical therapy (12%) (P<0.001 for both comparisons)
- p. 3 Table 1 footnote / relapse denominators (Table 1, p3)
4/5 (80)
Only the authors can say what is right here.
Judge's note: Rests on outside knowledge: the 1-year group sizes in ref 7 (about 41-50 per group). Table 1 relapse denominators give 5, 21 and 18 patients at target at 1 year. With 5/41, 21/50 and 18/49, a chi-square/two-proportion z-test gives about z=3.13 (P≈0.002) and z=2.66 (P≈0.008). Using this paper's 3-year Ns (40/48/49) also gives P>0.001. The paper names the chi-square test for this end point. The authors should state which test and P values from ref 7 apply.
Second check: another model agreed
Report a problem with this finding
Wording The Fig. 1 legend says "Shown are the percentage change in glycated hemoglobin levels (Panel A), the percentage change ... (Panel B)", but the axes are labelled "Change from Baseline (percentage points)" and plot absolute changes.
- p. 5 Figure 1 legend
Shown are the percentage change in glycated hemoglo-
- p. 6 Figure 1A/1B y-axis label
Change from Baseline (percentage points)
Judge's note: In the upright Fig. 1, the y-axes of Panels A and B read 'Change from Baseline (percentage points)'. The plotted 36-month values (about -2.5 for surgery, -0.5 for medical therapy) match Table 1's 'Change from baseline — percentage points'. The legend's 'percentage change' wording misdescribes what is plotted.
Second check: another model agreed
Report a problem with this finding
Typo Excessive weight gain is defined as "(>5% from baseline)" in the text but as "a 5% increase in body weight over baseline" in the Table 3 footnote.
- p. 7 Results, Adverse Events
Excessive weight gain (>5% from baseline)
- p. 9 Table 3, footnote ‡
Excessive weight gain was defined as a 5% increase in body weight over baseline.
Judge's note: Both wordings checked on p7 and the p9 footnote. The counts (7 of 43, 16%) agree. The only difference is whether the threshold is strict, which is a minor wording inconsistency.
Second check: none (judge only)