A recent case series of diabetic ketoacidosis (DKA) associated with colonoscopy in sodium–glucose cotransporter 2 inhibitor (SGLT2i)-treated diabetes (1) prompted a clinical alert update to include colonoscopy (2). That update advised colonoscopy cancellation when capillary ketone concentrations are >1.0 mmol/L (if blood gas analysis is unavailable) when SGLT2i have not been withheld for 72 h. However, those guidelines were formulated in the absence of data regarding the normal range for capillary ketone concentrations at the time of colonoscopy. Unnecessary colonoscopy cancellation carries risks, including delays in possible cancer detection, and psychological consequences, whereas a ketone cutoff that is too high increases the risk of DKA.
To define a ketone concentration that may serve as an empirical determinant of the need for further investigation, we now report a nondiabetic reference interval for capillary ketone concentrations at the time of colonoscopy from a multicenter observational study conducted between June and December 2020 at four tertiary health services (Alfred, Austin, Eastern, and Western Health) in Melbourne, Australia. Multisite approval was granted by the Alfred Human Research Ethics Committee.
Inclusion criteria for the reference population were community-dwelling, normoglycemic adults undergoing colonoscopy. Exclusion criteria were a history of diabetes, pancreatitis, pancreatic cancer, pancreatic surgery, hemochromatosis, cystic fibrosis, starvation ketosis (defined as fasting >72 h), pregnancy, or the discovery during recruitment of a fasting capillary glucose >5.5 mmol/L (100 mg/dL).
Participants with type 2 diabetes were also recruited to compare capillary ketone concentrations with the reference interval population. Except for physician-adjudicated type 2 diabetes, exclusion criteria were the same as those for the reference interval population. These participants were assigned to two prespecified subgroups: those treated with and without SGLT2i.
One hundred fifty-one normoglycemic participants formed the reference interval population (3). Of the 142 participants with type 2 diabetes, 105 were assigned to the non-SGLT2i-treated subgroup (102 not usually treated with SGLT2i and three who had taken their last SGLT2i doses 99–168 h before colonoscopy). Thirty-seven in the SGLT2i-treated subgroup had taken their last SGLT2i dose 59 ± 17 h (mean ± SD) before the colonoscopy (range 14–77 h).
Participants were predominantly Caucasian, with ethnicity being similar across the three groups. Those with diabetes were older (non-SGLT2i-treated, 67.1 ± 11.0 years; SGLT2i-treated, 64.4 ± 0.6 years; reference population, 53.8 ± 14.0 years; P < 0.001 overall) and had higher BMI (31.0 ± 6.6, 33.4 ± 7.3, and 27.4 ± 5.7 kg/m2, respectively; P < 0.001 overall). The percentage of males was similar (53%, 70%, and 60%; P = 0.18, χ2 test), as was fasting duration (21.3 ± 11.7, 18.2 ± 7.7, and 18.6 ± 8.8 h; P = 0.18, ANOVA). Use of standard versus extended bowel preparations was similar among the groups. Diabetes treatments (except for SGLT2i), duration, and HbA1c were also similar in both groups.
Point-of-care capillary glucose and ketone concentrations were measured in all participants <90 min before colonoscopy, using Nova Biomedical meters or Abbott Freestyle Optium Neo meters (two study sites each). Nova Biomedical StatStrip and Abbott ketone test strips measure β-hydroxybutyrate; the measurement range for both is 0–8.0 mmol/L with a coefficient of variation <5.5% (4). Venous blood gases were also measured when ketones were >1.0 mmol/L.
The nondiabetic reference interval for capillary ketone concentrations was 0.0–1.7 mmol/L using the central 95% of the data set. Thirteen participants(9%) had ketone concentrations >1.0 mmol/L, and seven (5%) had >1.5 mmol/L. The median (interquartile range) capillary ketone concentration was 0.4 mmol/L (0.2, 0.7), with no sex difference. Figure 1 displays capillary ketone concentrations in the reference population without diabetes compared with the subgroups with diabetes (overall P = 0.051, Kruskal-Wallis test). In the non-SGLT2i-treated diabetes subgroup, median capillary ketone concentration was 0.3 mmol/L (0.2, 0.6). Five participants (5%) had capillary ketone concentrations >1.0 mmol/L, and 1 (1%) had >1.5 mmol/L. These proportions were not significantly different from the normoglycemic reference group. In the SGLT2i-treated diabetes subgroup, median capillary ketone concentration was 0.5 mmol/L (0.2, 1.1). A greater proportion in this subgroup had capillary ketone concentrations >1.0 mmol/L: nine (24%) >1.0 mmol/L (P < 0.05 vs. reference population and non-SGLT2i, χ2 test) and four (11%) >1.5 mmol/L (NS vs. reference group, P < 0.05 vs. non-SGLT2i).
Only one participant (in the normoglycemic reference population) had venous pH <7.30 (pH 7.27). She was clinically well, and the colonoscopy and recovery period were uneventful. No participant had a venous bicarbonate <18 mmol/L.
The key finding of this study is that the reference range for capillary ketone concentrations in normoglycemic people undergoing colonoscopy is 0.0–1.7 mmol/L. This 1.7-mmol/L cutoff is clearly relevant when assessing a person with SGLT2i-treated diabetes at the time of colonoscopy. Colonoscopy management guidelines for SGLT2i-treated patients should be revised to reflect the normal capillary ketone concentration upper limit of 1.7 mmol/L.
Acknowledgments. The authors are very grateful for the assistance of all the study participants. They gratefully acknowledge the assistance of Amanda Nicoll, Head of Gastroenterology, Eastern Health; Kirsty Eddy and Jacinta Holman, Endoscopy Suite Nurse Unit Managers, Eastern Health; Josh Szental, Department of Anaesthesia, Pain and Perioperative Medicine, Western Health; Allison Scholey, Endoscopic Services, Western Health; Helen Chen, gastroenterology research nurse, Alfred Health; Nadia Scicluna, secretary for Department of Endocrinology and Diabetes, Alfred Health; and the endoscopy booking staff and day procedure staff at all sites.
Duality of Interest. L.A.B. was an investigator in DECLARE-TIMI 58. E.I.E.’s institution has received research funding from Novo Nordisk, Sanofi, Eli Lilly, Gilead, and Bayer. E.I.E. has served on the advisory boards of Sanofi, Eli Lilly, and Pfizer, with the consultancy fees for these donated toward diabetes research at E.I.E.’s institution. E.I.E. has received a travel grant from Boehringer Ingelheim to attend an American Diabetes Association virtual conference. R.W. has received speaker fees from Amgen. No other potential conflicts of interest relevant to this article were reported.
Author Contributions. P.S.H. designed and coordinated the study, researched data, and wrote the manuscript. R.W. contributed to study design, coordinated a site, researched data, and wrote the manuscript. E.I.E. and A.M. contributed to study design and revised the manuscript. S.S.-M. and R.H. each coordinated a site, researched data, and revised the manuscript. S.B., A.F., T.P.H., S.N.K., M.N., and S.R. researched data and revised the manuscript. L.A.B. contributed to study design, performed statistical analyses, and revised the manuscript. P.S.H. is the guarantor of this study and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.