The large intestine helps people process the food that has not been digested and absorbed in the small intestine. It houses billions of bacteria that break down food, especially fibers. It also produces hormones that may help control blood sugar after eating. People with type 2 diabetes, who have difficulty controlling their blood sugar after meals, have a different mix of gut bacteria than people without the disease. But scientists do not know if this unusual mix of bacteria is a result of having diabetes or its treatment, or if it contributes to the condition. They also do not know if hormones released by the large intestine play a role in the disease.

One way to study the role of the large intestine in diabetes is to look at patients who have had all or part of it removed. A few small studies with 21 patients or fewer have looked at how well patients are able to control their blood sugar after removal of all or part of their large intestine. But the results were confusing.

Now, Jensen, Sørensen et al. show that patients who have surgery to remove all or the left part of their large intestine are more likely to develop diabetes. They looked at patient records from a national registry in Denmark to see how many patients developed diabetes during up to 18 years after surgery. In the analysis, they compared post-surgical diabetes diagnoses in 3,793 people who had their whole large intestine removed, 42,486 who had part of it removed, and 694,110 people who had surgery on another part of the body. People who had their rectum removed did not have a greater risk of developing diabetes than people having other surgeries. But the risk was higher among those who had all or the left part of the large intestine taken out. This suggests that the left part of the large intestine helps people control their blood sugar levels.

More studies are needed to confirm that the large intestine plays a role in diabetes and to identify the underlying mechanisms. Understanding how the large intestine helps control blood sugar may help scientists develop new ways of treating or preventing type 2 diabetes.

The gut acts as a key regulator of glucose homeostasis. It influences appetite, gastric emptying, intestinal motility, digestion, and absorption of nutrients, as well as insulin secretion from the pancreas through hormonal secretion and nervous signaling (Holst et al., 2016). Whereas the role of the small intestine in glucose homeostasis has been well elucidated (Holst and Madsbad, 2016), the role of the colon is less clear. The growing interest in the gut microbiota has led to an increased focus on colonic pathophysiology since the vast majority of gut microbes resides in the colon. It has been shown that patients with type 2 diabetes harbor an altered fecal microbiota (MetaHIT consortium et al., 2015; Karlsson et al., 2013; Qin et al., 2012), but it is unknown whether it is causally involved in the pathogenesis of type 2 diabetes (Allin et al., 2015; IMI-DIRECT consortium et al., 2018; MetaHIT Consortium et al., 2016). Whereas some mechanistic studies in mice suggest that obesity and its related phenotypes can be transmitted via the fecal microbiota (Bäckhed et al., 2004; Ridaura et al., 2013), elimination of the gut microbiota by antibiotics in humans, does not seem to have any short-term effect on glucose regulation (Mikkelsen et al., 2015; Reijnders et al., 2016; Vrieze et al., 2014).

In addition to hosting the gut microbiota, the colon also serves as an endocrine organ. Glucagon-like peptide 1 (GLP-1) secreted by L-cells in the small intestine augments insulin secretion from the pancreatic β-cells in response to a meal (Holst et al., 2016). L-cells are, however, not only present in the small intestine but also in the colon, where the cell density increases from the proximal colon to the rectum (Eissele et al., 1992; Gunawardene et al., 2011; Jorsal et al., 2018). Accordingly, it has been shown that biologically active GLP-1 is present in colonic tissue in concentrations, which are similar to those in the small intestine (Deacon et al., 1995), but the physiological role of colonic GLP-1 remains elusive. In addition to GLP-1, colonic L cells also secrete the appetite-reducing hormone peptide YY (PYY) (Svane et al., 2016).

Patients who have undergone resection of the entire colon or parts of it, here denoted colectomy, may serve as an evident opportunity to study the role of the colon in health and disease (Jensen et al., 2015). Glucose regulation in patients with colectomy has been examined in a few previous studies, the largest of which included 21 patients, and although results are not clear-cut, they suggest that colectomy may influence the entero-insulinar axis (Besterman et al., 1982; Nauck et al., 1996; Palnaes Hansen et al., 1997). On one hand, based on studies showing that mice raised in the absence of microorganisms have less body fat and are less insulin resistant, removal of the human gut microbiota by colectomy may be expected to result in a reduced risk of type 2 diabetes (Bäckhed et al., 2004). On the other hand, removal of GLP-1 and PYY producing L-cells by colectomy may be expected to lead to an increased risk of type 2 diabetes.

In the present study, we tested the hypothesis that total and partial colectomy is associated with an altered risk of type 2 diabetes. For this purpose, we used the nationwide Danish National Patient Register to compare the frequency of clinically recorded type 2 diabetes in patients who had a colectomy with patients who had other types of surgeries not involving the gastrointestinal tract.

We identified 8946 individuals with total colectomy and 90,594 individuals with partial colectomy or proctectomy in the Danish National Patient Register. As explained in detail in Materials and methods, we based our main analyses on those who were alive and who did not have a diagnosis of diabetes recorded during the first 1000 days after the surgery. Patients with records of total colectomy at different dates or more than one partial colectomy, patients younger than 30 years, patients who died within the first 1000 days of follow-up, patients diagnosed with diabetes before surgery or within the first 1000 days after surgery, and patients who had a follow-up time shorter than 1000 days were excluded (Figure 1). Thus, we included 46,279 patients who had undergone total colectomy, partial colectomy, or proctectomy. For comparison, we identified a total of 694,110 age, sex, and year of surgery matched non-colectomy patients who had undergone other types of surgery not involving the gastrointestinal tract (Table 1). Mean (SD) follow-up time was 7.1 (5.2), 5.6 (4.6), 6.3 (4.7), 5.7 (4.8), 6.4 (4.9), 5.8 (4.8) years for patients with total colectomy, right hemicolectomy, resection of the transverse colon, left hemicolectomy, sigmoidectomy, and proctectomy, respectively.

Figure 1

Download asset Open asset

The outcome was routinely clinically diagnosed type 2 diabetes (hereafter denoted ‘diabetes’) as recorded in the Danish National Patient Register. Figure 2 shows the time-corresponding cumulative hazards (double Nelson-Aalen plot) of diabetes for colectomy patients vs. non-colectomy patients. The slopes of the curves equalize the hazard ratios (a slope of 1.00 corresponds to a HR of 1.00). The cumulative hazard of diabetes was greater among patients who had undergone total colectomy compared with non-colectomy patients. Correspondingly, the HR of diabetes was 1.40 (95% confidence interval [CI], 1.21 to 1.62; p<0.001) (Figure 3). The analysis of the risk of diabetes after partial colectomy revealed that the highest cumulative hazard was observed among patients who had undergone left hemicolectomy or sigmoidectomy. Accordingly, the HR was 1.41 (95% CI, 1.19 to 1.67; p<0.001) for patients who had undergone left hemicolectomy and 1.30 (95% CI, 1.21 to 1.40; p<0.001) for patients who had undergone sigmoidectomy. Right hemicolectomy was not statistically significantly associated with increased risk of diabetes (HR 1.08 [95% CI, 0.99 to 1.19; p=0.10]). Resection of the transverse colon was infrequent, but showed the same estimate, although with wider confidence interval; HR 1.08 (95% CI, 0.76 to 1.54; p=0.66), whereas proctectomy was not associated with increased risk of diabetes; HR 0.98 (95% CI, 0.91 to 1.07; p=0.71).

Figure 2

Download asset Open asset

Figure 3

Download asset Open asset

We stratified patients with colectomy into a group of colorectal cancer patients and a group with other colorectal diseases (Supplementary file 1 lists the most common diagnoses for the patients with other colorectal diseases). In both groups of patients, total colectomy was associated with increased risk of diabetes (Table 2). The HR of diabetes after total colectomy associated with colorectal cancer was 1.61 (95% CI, 1.22 to 2.11, p<0.001), whereas the HR for total colectomy associated with other colorectal disease was 1.34 (95% CI, 1.13 to 1.59, p<0.001). Corresponding HRs were 1.24 (95% CI, 1.01 to 1.53, p=0.04) and 1.88 (95% CI, 1.41 to 2.50, p<0.001) for left hemicolectomy, and 1.11 (95% CI, 0.997 to 1.24, p=0.06) and 1.47 (95% CI, 1.34 to 1.62, p<0.001) for sigmoidectomy.

If we, instead of following the patients from 1000 days after surgery, followed the patients from the date of surgery, we found a somewhat higher HR of diabetes after total colectomy: HR = 1.72 (95% CI, 1.56 to 1.89), which, however, should be interpreted with caution due to non-proportional hazards. In contrast, HRs of diabetes after left hemicolectomy and sigmoidectomy remained similar to results from the main analyses (Supplementary file 2). When follow-up time began 500 days after surgery or 1500 days after surgery we observed similar HRs of diabetes as compared to the main analyses (Supplementary file 2).

In the present nationwide study, we observed an increased risk of diabetes in colectomy patients. The risk was highest among individuals who had the left part of the colon removed, whereas resection of the rectum was not associated with any change in risk of diabetes. Moreover, we observed an increased risk of diabetes irrespective of whether the patients had co-occurring colorectal cancer or other underlying diseases.

A major strength of the present work is its nationwide nature allowing inclusion of many thousands of patients who have had colectomies and a large number of non-colectomy patients leading to relatively narrow confidence intervals of the hazard ratios observed. Moreover, the data allowed by study design to control for known confounding by sex and age as well as the possible confounding factors implicit in the time period and factors influencing the decision-making to refer to and conduct surgery in general. However, various limitations of our study must be considered in the interpretation of the results as indicating a causal relation of colectomy to later diabetes risk. We did not have information on several other potential confounders, precluding adjustment for them. This implies the possibility that the reported estimates may be positively biased. Colorectal cancer is associated with adiposity (Ning et al., 2010), which is also a major risk factor for diabetes, wherefore adiposity may confound our results. However, based on previously published data (Berentzen et al., 2011), it can be calculated that if differences in BMI were to fully account for our observed HR of 1.4 of diabetes, a BMI difference of 3 units (equaling a difference of 9 kg for individuals 175 cm tall) would be necessary. Another important potential confounder that we did not account for is medication. However, the direction of medication effects on risk of diabetes is likely ambiguous. For example, whereas corticosteroids used for treatment of inflammatory bowel disease have well-documented side-effects resulting in impaired glucose regulation, biologicals such as tumor necrosis factor-α antagonists may alleviate inflammation and therefore potentially improve glucose regulation (Antohe et al., 2012; Parmentier-Decrucq et al., 2009; Solomon et al., 2011; Tam et al., 2007). As the effects of medication on glucose regulation are ambiguous, and since we observed an enhanced risk of diabetes among patients with colorectal cancer and other colorectal diseases, it seems unlikely that our findings are due to differences in medication. Also, confounding by indication, that is, the indication for the colectomy — in this case the underlying disease — may bias our results. Hence, the underlying disease may, either through its effect on the normal colonic function or via other pathways, directly increase the risk of diabetes. Likely, the fact that we achieved similar results when we started the follow-up time 1000 days after the surgery in our main analyses or 1500 days after surgery in the supplemental analyses may speak against such confounding. Last, differences in the life style,for example in physical activity, smoking, dietary habits, including alcohol and coffee drinking, of the colectomy versus the non-colectomy groups may have played a role as confounders.

Another potential limitation of our study is that the outcome was based on data from the Danish National Patient Register, which includes data only on hospital inpatient and outpatient contacts. This implies that patients with undiagnosed diabetes and with diabetes diagnosed outside hospitals, but not recorded later in the hospitals, were missed, and that the observed incidence is an underestimation. On the other hand, it is also possible that the recorded diagnosis of diabetes is wrong. However, as we defined the non-colectomy group as individuals who had undergone various other types of surgery, we presume that the likelihoods of being referred to a hospital or an outpatient clinic, and thus the likelihoods of being registered with a clinical diagnosis of diabetes, are the same in the colectomy and non-colectomy patients. Further, the outcome diabetes based on the diagnosis code E14 (unspecified diabetes) may be subject to misclassification regarding the type of diabetes. The data available does not allow validation of the recorded diagnosis, but to avoid misclassification of type 2 diabetes as type 1 diabetes, we restricted the study population to individuals above 30 years, since type 1 diabetes is much more frequently diagnosed before than after that age. Also, type 1 diabetes is infrequent compared to type 2 diabetes, reducing the risk of significant misclassification in this regard (Diabetesforeningen, 2018 Facts about diabetes in Denmark https://diabetes.dk/diabetesforeningen/in-english/facts-about-diabetes-in-denmark.aspx).

If the observed association between colectomy and excess incidence of diabetes is not simply explained by chance, bias, or confounding, it may represent a causal effect of colectomy on the diabetes risk. Indeed, our prospective design, including a 1000 days delay after surgery, supports causality since the outcome occurred after the exposure and the risk remained equally elevated throughout the follow-up periods. The finding of increased risk following only colectomy including the left part of the colon, but not the remainder part or the rectum, suggests a specific causal effect of that part rather than a general effect of surgery on the colon, which would more likely reflect confounding. Obviously, experimental proof of a direct causal effect of colectomy on diabetes risk in humans is impossible, and epidemiological substitutes, such as instrumental variable analyses, for example by Mendelian randomization, may not be feasible. However, assuming causality may inspire investigations of the possible biological mechanisms that may have induced the association between colectomy and diabetes risk.

Thus, it may be speculated that our findings are explained by removal of GLP-1 producing cells or alterations of the colonic gut microbiota composition and function. Recent studies suggest that modifications in GLP-1 release are directly involved in both development of diabetes (Færch et al., 2015) and diabetes remission after gastric bypass (Holst and Madsbad, 2016). However, the physiological role of GLP-1 secreted from the colon is unclear. Interestingly, the L-cell density increases from the proximal to the distal colon (Eissele et al., 1992; Gunawardene et al., 2011), and accordingly, we observed the highest risk among individuals who had the left part of the colon removed. The bacterial load also increases along the length of the colon (Donaldson et al., 2016), and colectomies involving the left part of the colon therefore likely lead to removal of a larger part of the colonic microbiota than colectomies involving the right part of the colon. Notably however, it has been shown that ileocolonic resection leads to an altered colonization of the neo-terminal ileum (GETAID et al., 2016) suggesting that removal of the colonic gut microbiota may be somewhat compensated for by over-colonization of the remaining parts of the gastrointestinal tract. Such bacterial overgrowth may well change the microbiota-mediated conversion of primary bile acids to secondary bile acids, which through their hormone-like actions impact glucose metabolism (Ridlon et al., 2014). In our study, proctectomy was, with quite narrow upper confidence limits, not associated with an increased risk of diabetes. This may be explained by the fact that feces, and thus bacteria and other stimuli of L-cells, is only present in the rectum during the short time periods preceding defecation, thus hindering continuous signaling between the lumen of the rectum and extra-intestinal tissues. Last, the removal of the entire colon or a part of it might influence the exchanges between the gut epithelium and ingested food and fluids, especially due to an altered bacterial degradation of otherwise indigestible fibers to short-chain fatty acids, which could in turn have an impact upon diabetes development (Allin et al., 2015).

Among the previous studies that have examined glucose regulation in colectomy patients (Besterman et al., 1982; Nauck et al., 1996; Palnaes Hansen et al., 1997), the largest included 10 patients with an ileo-anal reservoir and 11 patients with an ileostomy who underwent an oral glucose tolerance test (Palnaes Hansen et al., 1997). Compared with 10 healthy individuals, colectomy individuals had higher peak levels of plasma glucose and higher peak levels and area under the curve of plasma insulin suggesting impaired glucose tolerance and hyperinsulinemia. However, the healthy controls had lower body weight compared with the patients, and no major differences in plasma GLP-1 levels were observed (Palnaes Hansen et al., 1997). The prior evidence from these studies for an association between colectomy and impaired glucose regulation is thus conflicting. In a previous study also based on the Danish National Patient Register, we found no corresponding difference in cardiovascular disease risk between individuals who had undergone a total colectomy and individuals who had undergone other types of surgeries leaving the gastrointestinal tract intact (Jensen et al., 2015). While we would expect that an increased risk of diabetes would associate with subsequent increased risk of cardiovascular diseases, this discrepancy may be explained by the limited follow-up time, not allowing presence of diabetes yet to be manifested in increased cardiovascular risk.

In conclusion, we observed an increased risk of clinically recorded type 2 diabetes among patients who had undergone total and partial colectomy with the risk being elevated only among individuals who had the left part of the colon removed. The increased risk of diabetes was observed in patients with colorectal cancer as well as in patients with other colorectal diseases. As our findings are based solely on register data, clinical-physiological studies are warranted to establish whether colectomy per se is associated with metabolic alterations indicating an increased risk of type 2 diabetes, and to understand the role of the colon in glucose homeostasis. Such studies could include comparisons of plasma glucose, insulin, and incretin hormone levels during a meal tolerance test before and after different types of colectomy with careful control for likely confounding factors such as those discussed above.

The study is based on data from the Danish National Patient Register [https://sundhedsdatastyrelsen.dk]. Procedures codes from the Danish National Patient Register are provided in Supplementary files 3 and 4. The Danish National Patient Register is protected by the Danish Act on Processing of Personal Data and can be accessed through application to and approval from the Danish Data Protection Agency and the Danish Health Data Authority [https://sundhedsdatastyrelsen.dk/da/forskerservice/ansog-om-data] where the purpose and the feasibility of the intended analysis should be accounted for. Informed consent and assessment of the proposal in scientific ethical committees are not required for registry-based research in Denmark.

1. 1. Allin KH
   2. Nielsen T
   3. Pedersen O

   (2015) Mechanisms in endocrinology: Gut microbiota in patients with type 2 diabetes mellitus

   European Journal of Endocrinology 172:R167–R177.

   https://doi.org/10.1530/EJE-14-0874

   • PubMed
   • Google Scholar
2. 1. Antohe JL
   2. Bili A
   3. Sartorius JA
   4. Kirchner HL
   5. Morris SJ
   6. Dancea S
   7. Wasko MC

   (2012) Diabetes mellitus risk in rheumatoid arthritis: reduced incidence with anti-tumor necrosis factor α therapy

   Arthritis Care & Research 64:215–221.

   https://doi.org/10.1002/acr.20657

   • PubMed
   • Google Scholar
3. 1. Bäckhed F
   2. Ding H
   3. Wang T
   4. Hooper LV
   5. Koh GY
   6. Nagy A
   7. Semenkovich CF
   8. Gordon JI

   (2004) The gut microbiota as an environmental factor that regulates fat storage

   PNAS 101:15718–15723.

   https://doi.org/10.1073/pnas.0407076101

   • PubMed
   • Google Scholar
4. 1. Berentzen TL
   2. Jakobsen MU
   3. Halkjaer J
   4. Tjønneland A
   5. Sørensen TI
   6. Overvad K

   (2011) Changes in waist circumference and the incidence of diabetes in middle-aged men and women

   PLOS ONE 6:e23104.

   https://doi.org/10.1371/journal.pone.0023104

   • PubMed
   • Google Scholar
5. 1. Besterman HS
   2. Adrian TE
   3. Mallinson CN
   4. Christofides ND
   5. Sarson DL
   6. Pera A
   7. Lombardo L
   8. Modigliani R
   9. Bloom SR

   (1982) Gut hormone release after intestinal resection

   Gut 23:854–861.

   https://doi.org/10.1136/gut.23.10.854

   • PubMed
   • Google Scholar
6. 1. Deacon CF
   2. Johnsen AH
   3. Holst JJ

   (1995) Human colon produces fully processed glucagon-like peptide-1 (7-36) amide

   FEBS Letters 372:269–272.

   https://doi.org/10.1016/0014-5793(95)00983-G

   • PubMed
   • Google Scholar
7. Website

   1. Diabetesforeningen

   (2018) Facts about diabetes in Denmark

   Accessed October 18, 2018.

   https://diabetes.dk/diabetesforeningen/in-english/facts-about-diabetes-in-denmark.aspx.
8. 1. Donaldson GP
   2. Lee SM
   3. Mazmanian SK

   (2016) Gut biogeography of the bacterial microbiota

   Nature Reviews Microbiology 14:20–32.

   https://doi.org/10.1038/nrmicro3552

   • PubMed
   • Google Scholar
9. 1. Eissele R
   2. Göke R
   3. Willemer S
   4. Harthus HP
   5. Vermeer H
   6. Arnold R
   7. Göke B

   (1992) Glucagon-like peptide-1 cells in the gastrointestinal tract and pancreas of rat, pig and man

   European Journal of Clinical Investigation 22:283–291.

   https://doi.org/10.1111/j.1365-2362.1992.tb01464.x

   • PubMed
   • Google Scholar
10. 1. Færch K
    2. Torekov SS
    3. Vistisen D
    4. Johansen NB
    5. Witte DR
    6. Jonsson A
    7. Pedersen O
    8. Hansen T
    9. Lauritzen T
    10. Sandbæk A
    11. Holst JJ
    12. Jørgensen ME

    (2015) GLP-1 response to oral glucose is reduced in prediabetes, screen-detected type 2 diabetes, and obesity and influenced by sex: the ADDITION-PRO study

    Diabetes 64:2513–2525.

    https://doi.org/10.2337/db14-1751

    • PubMed
    • Google Scholar
11. 1. GETAID
    2. Mondot S
    3. Lepage P
    4. Seksik P
    5. Allez M
    6. Tréton X
    7. Bouhnik Y
    8. Colombel JF
    9. Leclerc M
    10. Pochart P
    11. Doré J
    12. Marteau P

    (2016) Structural robustness of the gut mucosal microbiota is associated with Crohn's disease remission after surgery

    Gut 65:954–962.

    https://doi.org/10.1136/gutjnl-2015-309184

    • PubMed
    • Google Scholar
12. 1. Gunawardene AR
    2. Corfe BM
    3. Staton CA

    (2011) Classification and functions of enteroendocrine cells of the lower gastrointestinal tract

    International Journal of Experimental Pathology 92:219–231.

    https://doi.org/10.1111/j.1365-2613.2011.00767.x

    • PubMed
    • Google Scholar
13. 1. Holst JJ
    2. Gribble F
    3. Horowitz M
    4. Rayner CK

    (2016) Roles of the gut in glucose homeostasis

    Diabetes Care 39:884–892.

    https://doi.org/10.2337/dc16-0351

    • PubMed
    • Google Scholar
14. 1. Holst JJ
    2. Madsbad S

    (2016) Mechanisms of surgical control of type 2 diabetes: glp-1 is key factor

    Surgery for Obesity and Related Diseases 12:1236–1242.

    https://doi.org/10.1016/j.soard.2016.02.033

    • PubMed
    • Google Scholar
15. 1. IMI-DIRECT consortium
    2. Allin KH
    3. Tremaroli V
    4. Caesar R
    5. Jensen BAH
    6. Damgaard MTF
    7. Bahl MI
    8. Licht TR
    9. Hansen TH
    10. Nielsen T
    11. Dantoft TM
    12. Linneberg A
    13. Jørgensen T
    14. Vestergaard H
    15. Kristiansen K
    16. Franks PW
    17. Hansen T
    18. Bäckhed F
    19. Pedersen O

    (2018) Aberrant intestinal microbiota in individuals with prediabetes

    Diabetologia 61:810–820.

    https://doi.org/10.1007/s00125-018-4550-1

    • PubMed
    • Google Scholar
16. 1. Jensen AB
    2. Ajslev TA
    3. Brunak S
    4. Sørensen TI

    (2015) Long-term risk of cardiovascular and cerebrovascular disease after removal of the colonic microbiota by colectomy: a cohort study based on the danish national patient register from 1996 to 2014

    BMJ Open 5:e008702.

    https://doi.org/10.1136/bmjopen-2015-008702

    • PubMed
    • Google Scholar
17. 1. Jorsal T
    2. Rhee NA
    3. Pedersen J
    4. Wahlgren CD
    5. Mortensen B
    6. Jepsen SL
    7. Jelsing J
    8. Dalbøge LS
    9. Vilmann P
    10. Hassan H
    11. Hendel JW
    12. Poulsen SS
    13. Holst JJ
    14. Vilsbøll T
    15. Knop FK

    (2018) Enteroendocrine K and L cells in healthy and type 2 diabetic individuals

    Diabetologia 61:284–294.

    https://doi.org/10.1007/s00125-017-4450-9

    • PubMed
    • Google Scholar
18. 1. Karlsson FH
    2. Tremaroli V
    3. Nookaew I
    4. Bergström G
    5. Behre CJ
    6. Fagerberg B
    7. Nielsen J
    8. Bäckhed F

    (2013) Gut metagenome in european women with normal, impaired and diabetic glucose control

    Nature 498:99–103.

    https://doi.org/10.1038/nature12198

    • PubMed
    • Google Scholar
19. 1. Lynch SV
    2. Pedersen O

    (2016) The human intestinal microbiome in health and disease

    New England Journal of Medicine 375:2369–2379.

    https://doi.org/10.1056/NEJMra1600266

    • PubMed
    • Google Scholar
20. 1. Lynge E
    2. Sandegaard JL
    3. Rebolj M

    (2011) The danish national patient register

    Scandinavian Journal of Public Health 39:30–33.

    https://doi.org/10.1177/1403494811401482

    • PubMed
    • Google Scholar
21. 1. MetaHIT consortium
    2. Forslund K
    3. Hildebrand F
    4. Nielsen T
    5. Falony G
    6. Le Chatelier E
    7. Sunagawa S
    8. Prifti E
    9. Vieira-Silva S
    10. Gudmundsdottir V
    11. Pedersen HK
    12. Arumugam M
    13. Kristiansen K
    14. Voigt AY
    15. Vestergaard H
    16. Hercog R
    17. Costea PI
    18. Kultima JR
    19. Li J
    20. Jørgensen T
    21. Levenez F
    22. Dore J
    23. Nielsen HB
    24. Brunak S
    25. Raes J
    26. Hansen T
    27. Wang J
    28. Ehrlich SD
    29. Bork P
    30. Pedersen O

    (2015) Disentangling type 2 diabetes and metformin treatment signatures in the human gut microbiota

    Nature 528:262–266.

    https://doi.org/10.1038/nature15766

    • PubMed
    • Google Scholar
22. 1. MetaHIT Consortium
    2. Pedersen HK
    3. Gudmundsdottir V
    4. Nielsen HB
    5. Hyotylainen T
    6. Nielsen T
    7. Jensen BA
    8. Forslund K
    9. Hildebrand F
    10. Prifti E
    11. Falony G
    12. Le Chatelier E
    13. Levenez F
    14. Doré J
    15. Mattila I
    16. Plichta DR
    17. Pöhö P
    18. Hellgren LI
    19. Arumugam M
    20. Sunagawa S
    21. Vieira-Silva S
    22. Jørgensen T
    23. Holm JB
    24. Trošt K
    25. Kristiansen K
    26. Brix S
    27. Raes J
    28. Wang J
    29. Hansen T
    30. Bork P
    31. Brunak S
    32. Oresic M
    33. Ehrlich SD
    34. Pedersen O

    (2016) Human gut microbes impact host serum metabolome and insulin sensitivity

    Nature 535:376–381.

    https://doi.org/10.1038/nature18646

    • PubMed
    • Google Scholar
23. 1. Mikkelsen KH
    2. Frost M
    3. Bahl MI
    4. Licht TR
    5. Jensen US
    6. Rosenberg J
    7. Pedersen O
    8. Hansen T
    9. Rehfeld JF
    10. Holst JJ
    11. Vilsbøll T
    12. Knop FK

    (2015) Effect of antibiotics on gut microbiota, gut hormones and glucose metabolism

    PLOS ONE 10:e0142352.

    https://doi.org/10.1371/journal.pone.0142352

    • PubMed
    • Google Scholar
24. 1. Nauck MA
    2. Siemsglüss J
    3. Orskov C
    4. Holst JJ

    (1996)

    Release of glucagon-like peptide 1 (GLP-1 [7-36 amide]), gastric inhibitory polypeptide (GIP) and insulin in response to oral glucose after upper and lower intestinal resections

    Zeitschrift Fur Gastroenterologie 34:159–166.

    • PubMed
    • Google Scholar
25. 1. Ning Y
    2. Wang L
    3. Giovannucci EL

    (2010) A quantitative analysis of body mass index and colorectal cancer: findings from 56 observational studies

    Obesity Reviews 11:19–30.

    https://doi.org/10.1111/j.1467-789X.2009.00613.x

    • PubMed
    • Google Scholar
26. 1. Palnaes Hansen C
    2. Andreasen JJ
    3. Holst JJ

    (1997) The release of gastric inhibitory peptide, glucagon-like peptide-I, and insulin after oral glucose test in colectomized subjects

    Scandinavian Journal of Gastroenterology 32:473–477.

    https://doi.org/10.3109/00365529709025084

    • PubMed
    • Google Scholar
27. 1. Parmentier-Decrucq E
    2. Duhamel A
    3. Ernst O
    4. Fermont C
    5. Louvet A
    6. Vernier-Massouille G
    7. Cortot A
    8. Colombel JF
    9. Desreumaux P
    10. Peyrin-Biroulet L

    (2009) Effects of infliximab therapy on abdominal fat and metabolic profile in patients with Crohn's disease

    Inflammatory Bowel Diseases 15:1476–1484.

    https://doi.org/10.1002/ibd.20931

    • PubMed
    • Google Scholar
28. 1. Qin J
    2. Li Y
    3. Cai Z
    4. Li S
    5. Zhu J
    6. Zhang F
    7. Liang S
    8. Zhang W
    9. Guan Y
    10. Shen D
    11. Peng Y
    12. Zhang D
    13. Jie Z
    14. Wu W
    15. Qin Y
    16. Xue W
    17. Li J
    18. Han L
    19. Lu D
    20. Wu P
    21. Dai Y
    22. Sun X
    23. Li Z
    24. Tang A
    25. Zhong S
    26. Li X
    27. Chen W
    28. Xu R
    29. Wang M
    30. Feng Q
    31. Gong M
    32. Yu J
    33. Zhang Y
    34. Zhang M
    35. Hansen T
    36. Sanchez G
    37. Raes J
    38. Falony G
    39. Okuda S
    40. Almeida M
    41. LeChatelier E
    42. Renault P
    43. Pons N
    44. Batto JM
    45. Zhang Z
    46. Chen H
    47. Yang R
    48. Zheng W
    49. Li S
    50. Yang H
    51. Wang J
    52. Ehrlich SD
    53. Nielsen R
    54. Pedersen O
    55. Kristiansen K
    56. Wang J

    (2012) A metagenome-wide association study of gut microbiota in type 2 diabetes

    Nature 490:55–60.

    https://doi.org/10.1038/nature11450

    • PubMed
    • Google Scholar
29. 1. Reijnders D
    2. Goossens GH
    3. Hermes GD
    4. Neis EP
    5. van der Beek CM
    6. Most J
    7. Holst JJ
    8. Lenaerts K
    9. Kootte RS
    10. Nieuwdorp M
    11. Groen AK
    12. Olde Damink SW
    13. Boekschoten MV
    14. Smidt H
    15. Zoetendal EG
    16. Dejong CH
    17. Blaak EE

    (2016) Effects of gut microbiota manipulation by antibiotics on host metabolism in obese humans: a randomized double-blind placebo-controlled trial

    Cell Metabolism 24:341.

    https://doi.org/10.1016/j.cmet.2016.07.008

    • PubMed
    • Google Scholar
30. 1. Ridaura VK
    2. Faith JJ
    3. Rey FE
    4. Cheng J
    5. Duncan AE
    6. Kau AL
    7. Griffin NW
    8. Lombard V
    9. Henrissat B
    10. Bain JR
    11. Muehlbauer MJ
    12. Ilkayeva O
    13. Semenkovich CF
    14. Funai K
    15. Hayashi DK
    16. Lyle BJ
    17. Martini MC
    18. Ursell LK
    19. Clemente JC
    20. Van Treuren W
    21. Walters WA
    22. Knight R
    23. Newgard CB
    24. Heath AC
    25. Gordon JI

    (2013) Gut microbiota from twins discordant for obesity modulate metabolism in mice

    Science 341:1241214.

    https://doi.org/10.1126/science.1241214

    • PubMed
    • Google Scholar
31. 1. Ridlon JM
    2. Kang DJ
    3. Hylemon PB
    4. Bajaj JS

    (2014) Bile acids and the gut microbiome

    Current Opinion in Gastroenterology 30:332–338.

    https://doi.org/10.1097/MOG.0000000000000057

    • PubMed
    • Google Scholar
32. 1. Sjölander A
    2. Greenland S

    (2013) Ignoring the matching variables in cohort studies - when is it valid and why?

    Statistics in Medicine 32:4696–4708.

    https://doi.org/10.1002/sim.5879

    • PubMed
    • Google Scholar
33. 1. Solomon DH
    2. Massarotti E
    3. Garg R
    4. Liu J
    5. Canning C
    6. Schneeweiss S

    (2011) Association between disease-modifying antirheumatic drugs and diabetes risk in patients with rheumatoid arthritis and psoriasis

    JAMA 305:2525–2531.

    https://doi.org/10.1001/jama.2011.878

    • PubMed
    • Google Scholar
34. 1. Svane MS
    2. Jørgensen NB
    3. Bojsen-Møller KN
    4. Dirksen C
    5. Nielsen S
    6. Kristiansen VB
    7. Toräng S
    8. Wewer Albrechtsen NJ
    9. Rehfeld JF
    10. Hartmann B
    11. Madsbad S
    12. Holst JJ

    (2016) Peptide YY and glucagon-like peptide-1 contribute to decreased food intake after Roux-en-Y gastric bypass surgery

    International Journal of Obesity 40:1699–1706.

    https://doi.org/10.1038/ijo.2016.121

    • PubMed
    • Google Scholar
35. 1. Tam LS
    2. Tomlinson B
    3. Chu TT
    4. Li TK
    5. Li EK

    (2007) Impact of TNF inhibition on insulin resistance and lipids levels in patients with rheumatoid arthritis

    Clinical Rheumatology 26:1495–1498.

    https://doi.org/10.1007/s10067-007-0539-8

    • PubMed
    • Google Scholar
36. 1. Vrieze A
    2. Out C
    3. Fuentes S
    4. Jonker L
    5. Reuling I
    6. Kootte RS
    7. van Nood E
    8. Holleman F
    9. Knaapen M
    10. Romijn JA
    11. Soeters MR
    12. Blaak EE
    13. Dallinga-Thie GM
    14. Reijnders D
    15. Ackermans MT
    16. Serlie MJ
    17. Knop FK
    18. Holst JJ
    19. van der Ley C
    20. Kema IP
    21. Zoetendal EG
    22. de Vos WM
    23. Hoekstra JB
    24. Stroes ES
    25. Groen AK
    26. Nieuwdorp M

    (2014) Impact of oral vancomycin on gut microbiota, bile acid metabolism, and insulin sensitivity

    Journal of Hepatology 60:824–831.

    https://doi.org/10.1016/j.jhep.2013.11.034

    • PubMed
    • Google Scholar

Thank you for submitting your article "Risk of type 2 diabetes after colectomy" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Prabhat Jha as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

The authors present data from a record linkage study in Denmark which has been used to investigate the association between having different types of partial or complete colectomy and future diagnosis of diabetes. The purpose behind such a study is the investigation of the role of the colon in the aetiology of diabetes.

Summary:

The observation made in this paper of an association between colectomy and incident diabetes is potentially important but there are questions about possible confounding that need to be addressed in greater depth and also needs to be mentioned in the Abstract. The authors should follow the suggestion to restructure the Discussion to focus first on the question of whether or not there is a real association before proceeding to discuss whether or not it is likely to be causal and what biological mechanisms may underlie it.

As with all such observational studies the first question is whether there is an association between colectomy and diabetes incidence and secondly whether that association is explained by chance, bias or confounding. The authors have done a good job in largely ruling out the role of chance since the study has been conducted in the context of a national health data registry thus allowing them to include many thousands of patients who have had colectomies and a large number of control patients. The confidence intervals of the hazard ratios observed are generally relatively narrow. However, the problems of confounding are much more difficult to deal with and the paper needs to focus more on how this may affect the results.

Essential revisions:

1) The researchers have included all patients who have had partial or total colectomy within the national system of Denmark within a specific time period, thus the possibility of selection bias is limited. The national register includes data from hospital inpatient and outpatient contacts and therefore there is limited possibility of information bias either. The outcome is clinically diagnosed type 2 diabetes according to the ICD-10 classification given in the register. We think it would be worth describing this throughout as "clinically diagnosed type 2 diabetes" instead of simply saying that the research team as studied the incidence of type 2 diabetes (T2D). The addition of "clinically diagnosed" makes two points; firstly that there is a distinction between all diabetes and that which is diagnosed. This is important because any association with diagnosed T2D could be a true association with diabetes or alternatively an association with an increased likelihood of it being diagnosed. Secondly it makes the point that the diagnosis of T2D as opposed to any other form of diabetes is clinical and thus subject to misclassification. The Discussion section of the paper describes the potential for the case ascertainment process to miss cases, but it doesn't describe the potential for misclassification.

2) The major limitation of the analysis, as acknowledged by the authors, is the lack of information on potential confounding variables. However, given these are individuals within the hospital system we were surprised that basic data such as BMI were not available. Can the authors confirm such data are not captured by the Danish hospital system?

3) The non-colectomy patients are highly selected using individual matching on age, sex and date of surgery, resulting in identical means, SDs and% s in those with and without colectomy (Table 1), so it is unclear what the point was of also adjusting for these variables in the Cox regression models.

4) Of greater concern is the fact that no other potential confounders were included in the model. The authors highlight this as an issue in the fourth paragraph of the Discussion and discuss why the result may not be explained by adiposity or medication; however, even if there is a true association, the inability to adjust for multiple potential confounders (not just BMI and medication, but also activity, diet and smoking) means that the reported estimates are likely to be positively biased, so the statement that the work "enables a precise estimation of the association" is highly questionable. The lack of data on and adjustment for confounders is a major limitation which should also be stated in the Abstract.

5) In the Discussion, it is noted: "However, we observed a similar risk of type 2 diabetes in patients with colorectal cancer or other colorectal diseases suggesting that adiposity does not account for our results." However, adiposity may also be a risk factor for the other colorectal diseases (whatever they might be). In any case, the authors really cannot get around the lack of such data and the interpretation of the results should be conducted with this in mind.

6) Figure 2 claims to plot cumulative hazard for colectomy vs. non-colectomy patients, but cumulative hazard is a function of time, so it is not clear what is being presented here, or whether this type of plot is actually useful. It would be preferable to see Kaplan-Meier plots of cumulative survival (or failure) over time for the colectomy and non-colectomy groups. The statement “Compared with the control patients, the cumulative hazard of type 2 diabetes was greater among patients who had undergone total colectomy (Figure 2)” also needs to be clarified.

7) Having considered whether the association is true and not a reflection of chance, bias and confounding, the next question is whether it is likely to be causal. Of the criteria used to investigate the likelihood of causality, biological plausibility is one of the weakest but within the Discussion section of this paper is given rather too much prominence and is considered almost at the beginning of the discussion and is quite lengthy. A restructuring of the Discussion might be considered so that it is clear to begin with that the observation is an association between colectomy and incidence of clinically diagnosed T2D. Alternative explanations for the observed association should then be considered before the issues of possible causality are discussed, of which biological plausibility is only one.

8) Even if it is cited in the Introduction (first paragraph), the role of the gut in the absorption of nutrients is not considered a possible factor in influencing the development of diabetes. However, the removal of the entire colon or a part of it might influence the exchanges between the gut epithelium and ingested food and fluids, and hence modify the absorption of nutrients which could in turn impact upon diabetes development. This should be indicated as a possible explanation for why colectomy might raise diabetes risk.

9) The final section of the Discussion mentions the need for "clinical-physiological studies" to establish whether colectomy per se is associated with an increased risk of diabetes. It is not clear what the authors mean by this and this should be clarified. They also state that an implication of this paper is that clinicians should be aware of this risk to improve the long-term health of this patient group. Given that the risk increase is relatively small and the authors have been unable to deal with confounding by a whole series of individual-level confounding factors, it does seem a bit of a leap to suggest that a major implication of this study is clinical. It is more likely that the implications lie in understanding the role of the colon in glucose homeostasis.

Essential revisions:

1) The researchers have included all patients who have had partial or total colectomy within the national system of Denmark within a specific time period, thus the possibility of selection bias is limited. The national register includes data from hospital inpatient and outpatient contacts and therefore there is limited possibility of information bias either. The outcome is clinically diagnosed type 2 diabetes according to the ICD-10 classification given in the register. We think it would be worth describing this throughout as "clinically diagnosed type 2 diabetes" instead of simply saying that the research team as studied the incidence of type 2 diabetes (T2D). The addition of "clinically diagnosed" makes two points; firstly that there is a distinction between all diabetes and that which is diagnosed. This is important because any association with diagnosed T2D could be a true association with diabetes or alternatively an association with an increased likelihood of it being diagnosed. Secondly it makes the point that the diagnosis of T2D as opposed to any other form of diabetes is clinical and thus subject to misclassification. The Discussion section of the paper describes the potential for the case ascertainment process to miss cases, but it doesn't describe the potential for misclassification.

We accept the arguments for making this aspect more explicit. However, we suggest that it would be even more precise to say, ‘clinically recorded type 2 diabetes’. In the present nationwide study population, the diagnosis of diabetes will always be the result of a clinical procedure carried out at the hospital contact, where it is recorded or earlier, e.g. by the general practitioners. We use the information about type 2 diabetes when it is recorded for the first time in the patient register, i.e. at a hospital contact. Therefore, we have specified that the study is based on ‘clinically recorded type 2 diabetes’ in the title (which has consequently been slightly modified), the Abstract, the end of the Introduction, the beginning of the Results section, the Materials and methods section, and in the headings of the tables and the figure legends, but we suggest that it makes the text rather heavy-going by using the full phrase throughout the manuscript.

We have added a description of the possible misclassification regarding the type of diabetes in the Discussion section:

“Further, the outcome diabetes based on the diagnosis code E14 (unspecified diabetes) may be subject to misclassification regarding the type of diabetes. […] Also, type 1 diabetes is infrequent compared to type 2 diabetes, reducing the risk of significant misclassification in this regard ("Diabetesforeningen. Facts about diabetes in Denmark (https://diabetes.dk/diabetesforeningen/in-english/facts-about-diabetes-in-denmark.aspx)").”

2) The major limitation of the analysis, as acknowledged by the authors, is the lack of information on potential confounding variables. However, given these are individuals within the hospital system we were surprised that basic data such as BMI were not available. Can the authors confirm such data are not captured by the Danish hospital system?

Unfortunately, we do not have access to individual hospital records, where such information may be included. We can confirm that data on BMI is not captured in the Danish National Patient Register. The register contains information such as hospital departments, diagnoses and operation codes, and date for hospital arrival and departure, outpatients contacts, and surgical procedures. However, no information on anthropometrics is registered.

3) The non-colectomy patients are highly selected using individual matching on age, sex and date of surgery, resulting in identical means, SDs and% s in those with and without colectomy (Table 1), so it is unclear what the point was of also adjusting for these variables in the Cox regression models.

The sampling procedure implies that the most efficient and unbiased comparison is based on each colectomy patient with the matched non-colectomy patients. The non-colectomy patients matched to other colectomy patients different in sex, age and date of surgery (of potential great influence on risk of diabetes) may not be valid for that comparison. To maintain these strengths of the matched comparisons, we added the matching variables to the regression models. Accordingly, in a methods paper from Sjölander and Greenland (2013) conclude that ‘ignoring the matching variables in a cohort study can leave bias if there are additional confounders, even with adjustment for the additional confounders. This bias can be avoided by adjusting for the matching variables.’

We have modified the text in the Materials and methods section (including a reference to the above-mentioned paper) accordingly:

“To optimize the efficiency and minimize bias (Sjolander & Greenland, 2013) of the comparisons of the colectomy and non-colectomy patients, the Cox regression models included adjustment for the matching variables, sex, age and year of surgery.”

4) Of greater concern is the fact that no other potential confounders were included in the model. The authors highlight this as an issue in the fourth paragraph of the Discussion and discuss why the result may not be explained by adiposity or medication; however, even if there is a true association, the inability to adjust for multiple potential confounders (not just BMI and medication, but also activity, diet and smoking) means that the reported estimates are likely to be positively biased, so the statement that the work "enables a precise estimation of the association" is highly questionable. The lack of data on and adjustment for confounders is a major limitation which should also be stated in the Abstract.

We agree that there may be a risk of uncontrolled confounding that leads to positively biased estimates, and we also agree that there may be other known, as those mentioned, or even unknown confounders beyond those focused upon in our Discussion.

With the statement about ‘precise estimation’, we refer to the narrowness of the confidence intervals. However, we acknowledge that it may be mistaken for ‘accuracy’ (or ‘validity’), which may well be affected by the possible biases and confounding. Accordingly, we have deleted the statement, and instead, based on the general comment above, we say:

“A major strength of the present work is its nationwide nature allowing inclusion of many thousands of patients who have had colectomies and a large number of matched non-colectomy patients leading to relatively narrow confidence intervals of the hazard ratios observed.”

Additionally, we have added this sentence to the Abstract:

“Although we were not able to adjust for several potential confounders, our findings …”.

5) In the Discussion, it is noted: "However, we observed a similar risk of type 2 diabetes in patients with colorectal cancer or other colorectal diseases suggesting that adiposity does not account for our results." However, adiposity may also be a risk factor for the other colorectal diseases (whatever they might be). In any case, the authors really cannot get around the lack of such data and the interpretation of the results should be conducted with this in mind.

We acknowledge this argument and accordingly, we have deleted this sentence from the Discussion.

6) Figure 2 claims to plot cumulative hazard for colectomy vs. non-colectomy patients, but cumulative hazard is a function of time, so it is not clear what is being presented here, or whether this type of plot is actually useful. It would be preferable to see Kaplan-Meier plots of cumulative survival (or failure) over time for the colectomy and non-colectomy groups. The statement “Compared with the control patients, the cumulative hazard of type 2 diabetes was greater among patients who had undergone total colectomy (Figure 2)” also needs to be clarified.

Kaplan-Meier plots of cumulative survival (or failure) probabilities would not be interpretable in the type of data analyzed here. There are two options for making such plots, both of which will be based on invalid assumptions. If patients who die during follow-up without having had the diagnosis of diabetes recorded are censored in the analysis, then the estimated probabilities in the Kaplan-Meier plots will assume that the patients who die are still alive with the same diabetes risk as before, which makes no sense. If a combined end-point of diabetes or death is used instead, the comparisons between the colectomy and non-colectomy patients will be biased because of different mortality in these groups, especially for the patients after total colectomy as shown in Supplementary file 5. These problems of using conditional cumulative probabilities are avoided by estimating Nelson-Aalen cumulative hazards over time as we have done it in Supplementary file 5. One option to show the time-dependency of the emergence type 2 diabetes (as recorded for the first time) could be to show such graphs for all types of colectomy and their respective non-colectomy groups. However, we think such graphs will be overwhelming and difficult to read, and they will add little to answering the key question about the difference in risk (hazard) over time across the colectomy and non-colectomy groups. By plotting the time-corresponding cumulative hazard (double Nelson-Aalen plot) as done in Figure 2, we get an optimal representation of the hazard ratios over time, because the slope of the curves equalizes the hazard ratios (a slope of 1.00 is a hazard ratio of 1.00). At the same token, these plots show that the fundamental assumption of the Cox regression analysis of proportional hazards over time is fulfilled by the curves being approximately straight lines. Therefore, we would suggest that we maintain the presentations as done in Figure 2. To secure the correct interpretation by the readers, we have added this text:

– Methods section: “The cumulative hazard of diabetes was estimated using the Nelson-Aale estimator. […] A linear relationship between the time-corresponding Nelson-Aalen estimators indicates fulfillment of the assumption of proportional hazards.”

– Results section: “Figure 2 shows the time-corresponding cumulative hazard (double Nelson-Aalen plot) of diabetes for colectomy patients vs. non-colectomy patients. The slopes of the curves equalize the hazard ratios (a slope of 1.00 corresponds to a HR of 1.00).”

– Figure 2 legend: “The cumulative hazard was estimated using the Nelson-Aalen estimator. […] The linearity attests to the fulfillment of the assumption of proportional hazards in the Cox regression models.”

Furthermore, we have clarified the statement: “The cumulative hazard of diabetes was greater among patients who had undergone total colectomy compared with non-colectomy patients.”

7) Having considered whether the association is true and not a reflection of chance, bias and confounding, the next question is whether it is likely to be causal. Of the criteria used to investigate the likelihood of causality, biological plausibility is one of the weakest but within the Discussion section of this paper is given rather too much prominence and is considered almost at the beginning of the discussion and is quite lengthy. A restructuring of the Discussion might be considered so that it is clear to begin with that the observation is an association between colectomy and incidence of clinically diagnosed T2D. Alternative explanations for the observed association should then be considered before the issues of possible causality are discussed, of which biological plausibility is only one.

Whereas we of course agree that biological plausibility is just one, and usually also a weak, indicator of causality, we do suggest that it is the possible biological explanations of the observations that make them particularly interesting from a biomedical point of view. However, we have complied with the request by restructuring the discussion, which now underlines the importance of confounding and bias first, followed by a discussion of possible causality and ending with a paragraph on biological plausibility, which has been shortened.

8) Even if it is cited in the Introduction (first paragraph), the role of the gut in the absorption of nutrients is not considered a possible factor in influencing the development of diabetes. However, the removal of the entire colon or a part of it might influence the exchanges between the gut epithelium and ingested food and fluids, and hence modify the absorption of nutrients which could in turn impact upon diabetes development. This should be indicated as a possible explanation for why colectomy might raise diabetes risk.

We agree that especially removal of the entire colon and perhaps the right part may affect the absorption, also because of compensatory mucosal adaptations in the small intestine. We find it less likely that isolated removal of the left colon or the sigmoid colon would have such general effects on nutrient absorption, but there may be effects on specific nutrients, e.g. short-chain fatty acids, produced in the colon by bacterial degradation of the otherwise indigestible fibers. We have added this sentence to the Discussion:

“Last, the removal of the entire colon or a part of it might influence the exchanges between the gut epithelium and ingested food and fluids, especially due to an altered bacterial degradation of otherwise indigestible fibers to short-chain fatty acids, which could in turn have an impact upon diabetes development (Allin et al., 2015).”

9) The final section of the Discussion mentions the need for "clinical-physiological studies" to establish whether colectomy per se is associated with an increased risk of diabetes. It is not clear what the authors mean by this and this should be clarified. They also state that an implication of this paper is that clinicians should be aware of this risk to improve the long-term health of this patient group. Given that the risk increase is relatively small and the authors have been unable to deal with confounding by a whole series of individual-level confounding factors, it does seem a bit of a leap to suggest that a major implication of this study is clinical. It is more likely that the implications lie in understanding the role of the colon in glucose homeostasis.

We have added this sentence to the Discussion:

“Such studies could include comparisons of plasma glucose, insulin, and incretin hormone levels during a meal tolerance test before and after different types of colectomy with careful control for likely confounding factors such as those discussed above.”

We have deleted the last sentence in the Discussion about the relevance of the findings for the clinicians.

Author details

1. Anders B Jensen

   1. Novo Nordisk Foundation Center for Protein Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
   2. Institute for Next Generation Healthcare, Mount Sinai Health System, New York, United States

   Contribution

   Data curation, Software, Formal analysis, Funding acquisition, Visualization, Methodology, Writing—original draft, Writing—review and editing

   Contributed equally with

   Thorkild IA Sørensen

   Competing interests

   No competing interests declared

2. Thorkild IA Sørensen

   1. Novo Nordisk Foundation Center for Basic Metabolic Research, Section of Metabolic Genetics, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
   2. Department of Public Health, Section of Epidemiology, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

   Contribution

   Conceptualization, Formal analysis, Supervision, Visualization, Methodology, Project administration, Writing—review and editing

   Contributed equally with

   Anders B Jensen

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0003-4821-430X

3. Oluf Pedersen

   Novo Nordisk Foundation Center for Basic Metabolic Research, Section of Metabolic Genetics, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

   Contribution

   Writing—review and editing

   Competing interests

   No competing interests declared

4. Tine Jess

   Center for Clinical Research and Prevention, Bispebjerg and Frederiksberg Hospital, Copenhagen, Denmark

   Contribution

   Supervision, Writing—review and editing

   Competing interests

   No competing interests declared

5. Søren Brunak

   Novo Nordisk Foundation Center for Protein Research, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark

   Contribution

   Conceptualization, Resources, Software, Supervision, Funding acquisition, Methodology, Project administration, Writing—review and editing

   Contributed equally with

   Kristine H Allin

   For correspondence

   soren.brunak@cpr.ku.dk

   Competing interests

   No competing interests declared

6. Kristine H Allin

   1. Novo Nordisk Foundation Center for Basic Metabolic Research, Section of Metabolic Genetics, Faculty of Health and Medical Sciences, University of Copenhagen, Copenhagen, Denmark
   2. Center for Clinical Research and Prevention, Bispebjerg and Frederiksberg Hospital, Copenhagen, Denmark

   Contribution

   Funding acquisition, Visualization, Writing—original draft, Writing—review and editing

   Contributed equally with

   Søren Brunak

   For correspondence

   kristine.allin@regionh.dk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-6880-5759

• 2,392

  Page views

• 203

  Downloads

• 20

  Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.