Paperonium

← Back to the plain-language case

Abstract reports a 13.2-point TIR gain for baseline GMI >8.0; Table 1 implies 22.3 points

Diabetes Care (2022) Β· doi:10.2337/dc22-1217 Β· PMC9862393

Summary

The abstract states that time in target range (TIR, 70–180 mg/dL) improved by 13.2 percentage points after the switch from predictive low glucose suspend (PLGS) to automated insulin delivery (AID) in people with baseline GMI >8.0 (N = 4,064). Table 1 gives this group a TIR of 32.5% on PLGS and 55.7%, 54.5% and 54.1% in AID months 1 to 3. That is a gain of 23.2, 22.0 and 21.6 points, or 22.3 points when the three AID months are averaged, which is the convention the abstract's other figures follow. The Results ("by >20 percentage points") and the Conclusions ("22–23 percentage point change") agree with the table, so the abstract is the only place with 13.2. The abstract figure understates the largest stratified effect by about 9 points (41%), and it makes that effect look smaller than the under-18 effect (14.0) when the table shows it is about 1.6 times as large.

What the paper reports

Abstract, Results:

Compared with PLGS, AID increased TIR on average from 58.4 to 70.5%.

Glycemic improvements were most pronounced in those <18 years old (TIR improvement 14.0 percentage points) and those with baseline GMI >8.0 (TIR improvement 13.2 percentage points).

Results, second paragraph:

Glycemic outcomes improved most in those with baseline GMI >8.0. These improvements were highly selective: individuals in an optimal state of glycemic control (baseline GMI ≀6.9) had TIR improved by ∼4.0 percentage points and had rate of hypoglycemia on AID compared with PLGS reduced by 0.9 percentage points, while those with baseline GMI >8.0 had TIR improved by >20 percentage points, but time below range also increased slightly (Table 1).

Conclusions (the paper's discussion section), fourth paragraph:

The group with the highest GMI, >8.0, showed the greatest percent improvement in TIR with a 22–23 percentage point change, and they achieved this change with no clinically significant increase in time spent in a state of hypoglycemia. Percent time >250 mg/dL was reduced by nearly one-half along with a reduction in time >180 mg/dL of 22–23 percentage points.

Table 1, "By baseline GMI", % TIR (columns PLGS | AID month 1 | AID month 2 | AID month 3; data are means Β± SD):

≀6.9 (N = 4,450) | 80.6 Β± 7.3 | 84.2 Β± 6.8 | 83.8 Β± 7.3 | 83.6 Β± 7.6
6.9–7.4 (N = 5,359) | 65.9 Β± 5.4 | 75.4 Β± 7.0 | 74.8 Β± 7.7 | 74.4 Β± 8.1
7.4–8.0 (N = 5,018) | 51.8 Β± 5.0 | 67.5 Β± 7.8 | 66.6 Β± 8.5 | 66.2 Β± 8.9
>8.0 (N = 4,064) | 32.5 Β± 9.5 | 55.7 Β± 11.0 | 54.5 Β± 11.5 | 54.1 Β± 12.0

What the paper's own numbers show

Method. First I worked out the convention the abstract uses for "improvement", using the overall rows of Table 1. The abstract's overall figures (TIR 58.4 to 70.5%, GMI 7.5 to 7.1, time <70 mg/dL 1.66 to 1.46%) equal the PLGS value and the mean of the three AID months. Then I applied the same convention to every TIR row stratified by baseline GMI and by age. For the under-18 group, which the abstract reports as one figure, I pooled the 0–12 and 13–18 rows weighted by their N. I also computed each month separately, and the same quantities for time >180 mg/dL. Every number below is printed by recompute.py.txt.

Quantity Location Reported Recomputed from the paper's own numbers
Overall TIR, PLGS β†’ AID Abstract; Table 1 58.4 β†’ 70.5%; "average difference" 12.06 mean of AID months 70.50%; change 12.10 (consistent)
Overall GMI and time <70 mg/dL, PLGS β†’ AID Abstract; Table 1 7.5 β†’ 7.1; 1.66 β†’ 1.46% 7.47 β†’ 7.08; 1.66 β†’ 1.46% (consistent)
TIR change, age <18 Abstract 14.0 13.93 (N-weighted; 0–12: 14.13, 13–18: 13.80) (consistent)
TIR change, age <18 Results "β‰₯13" 13.0 to 15.2 in every month for both groups (consistent)
TIR change, baseline GMI >8.0 Abstract 13.2 22.27 (54.77 βˆ’ 32.5); months 1–3: 23.2 / 22.0 / 21.6
TIR change, baseline GMI >8.0 Results ">20" 21.6 to 23.2 (consistent)
TIR change, baseline GMI >8.0 Conclusions, para. 4 "22–23" 22.27 (consistent)
Time >180 mg/dL reduction, GMI >8.0 Conclusions, para. 4 "22–23" 22.53; months 1–3: 23.4 / 22.3 / 21.9 (consistent)
TIR change, other GMI strata Table 1 (text gives only "∼4.0" in Results and "3–4" in Conclusions for ≀6.9) β€” ≀6.9: 3.27; 6.9–7.4: 8.97; 7.4–8.0: 14.97
Ratio of GMI >8.0 gain to <18 gain Abstract vs Table 1 13.2 / 14.0 = 0.94 22.27 / 13.93 = 1.60
Any baseline-GMI TIR change (month 1, 2, 3 or mean) within 0.15 of 13.2 Table 1 β€” none

The abstract's figure for the GMI >8.0 group cannot be obtained from Table 1 by any time point or summary used elsewhere in the paper. The abstract's other numbers can. Under the abstract's own convention, the value is 22.3. That agrees with the Results and with the Conclusions, which both describe the same group and the same outcome.

A small side observation, not part of the main finding: the abstract gives overall time >180 mg/dL on AID as 28.1%, while the mean of the three AID months is 28.0% (Table 1's average difference of βˆ’11.86 gives 28.04%). The difference is 0.1 points and has no bearing on any conclusion.

Figure

The abstract's 13.2-point figure for baseline GMI >8.0 conflicts with Table 1, the Results and the Conclusions

Figure 1. The conflicting passages side by side. Left (orange): the abstract's figure for the GMI >8.0 group. Right (blue): the Table 1 values for the same group, with the month-by-month changes and their mean recomputed from the table. Their mean (22.3) is 9.1 points larger than the abstract figure. Bottom: the Results and Conclusions describe the same group and outcome, and both agree with the table, not with the abstract. All quoted text is verbatim from the article as archived in PMC.

Recomputed TIR changes by baseline GMI and for under-18s, with the two values the abstract reports

Figure 2. Change in % TIR from PLGS to AID (mean of AID months 1–3 minus PLGS), recomputed from Table 1 for the four baseline-GMI strata and for the pooled under-18 group (blue circles). The two values the abstract reports are shown as orange diamonds. For under-18s, the abstract's 14.0 agrees with the recomputed 13.9. For GMI >8.0, the abstract's 13.2 is 9.1 points below the recomputed 22.3. It is also below the recomputed value for the next stratum down (GMI 7.4–8.0, 15.0). The axis starts at zero.

Why it matters

Explanations considered

Status

Suggested correction

In the abstract, replace

those with baseline GMI >8.0 (TIR improvement 13.2 percentage points)

with

those with baseline GMI >8.0 (TIR improvement 22.3 percentage points)

The value 22.3 is the mean of AID months 1–3 minus PLGS from Table 1 ((55.7 + 54.5 + 54.1) / 3 βˆ’ 32.5 = 22.27), the same convention the abstract uses for its other figures, and it agrees with the Results (">20") and the Conclusions ("22–23"). If the authors computed this value from the underlying patient-level data, their exact figure may differ slightly in the last decimal. The authors may also wish to reconsider the wording "most pronounced in those <18 years old ... and those with baseline GMI >8.0". The table shows the GMI >8.0 gain is about 1.6 times the under-18 gain, so the two gains are not comparable in size.