Hemoglobin A1c as a marker to stratify diabetes risk following pancreaticoduodenectomy.

BACKGROUND: Pancreatic cancer has been shown to cause diabetes mellitus, and diabetes mellitus has been shown to be a risk factor for pancreatic cancer. The effect of pancreaticoduodenectomy on risk for development of diabetes mellitus is unclear. This study used hemoglobin A1c to determine the incidence of diabetes mellitus development following pancreaticoduodenectomy based on preoperative risk of diabetes mellitus.
METHODS: Retrospective review of patients undergoing pancreaticoduodenectomy was performed with comparison of preoperative diabetes mellitus status and hemoglobin A1c with development of diabetes mellitus postoperatively. Risk ratios were calculated to determine the risk for diabetes mellitus development.
RESULTS: Among 90 patients who met inclusion criteria, 26.7% developed new-onset or worsening diabetes mellitus following pancreaticoduodenectomy. Of those with hemoglobin A1c ≤ 5.6%, only 7.7% of patients developed diabetes mellitus. Patients at risk for diabetes mellitus preoperatively had 4.0 times greater risk for development of diabetes mellitus following pancreaticoduodenectomy.
CONCLUSION: Hemoglobin A1c levels should be used to identify patients at risk for new-onset or worsening diabetes mellitus following pancreaticoduodenectomy.

INTRODUCTION

The association between pancreatic ductal adenocarcinoma (PDAC) and diabetes mellitus (DM) has been studied dating back to 1973 [1]. Over the following decades, reports suggested that patients with type 2 DM had a nearly 200% increased risk for development of PDAC [[2], [3], [4]]. More recent evidence suggests at least some evidence of glucose intolerance at the time of PDAC diagnosis in up to 80% of patients [5]. This is likely an overestimate, however, as PDAC may cause a paraneoplastic reaction leading to DM [6,7]. In fact, new-onset DM can be an early marker of pancreatic malignancy [[8], [9], [10]].

Patients with PDAC are treated with aggressive surgery when possible, with pancreaticoduodenectomy (PD) being the standard of care for patients with lesions in the head of the pancreas. With a significant reduction in pancreatic volume, this may lead to a significant decrease in the number of insulin-secreting β cells, causing concern that this operation could lead to new-onset or worsening DM [11]. For benign disease, studies show that up to 18% of patients undergoing PD and 14%–31% undergoing distal pancreatectomy develop DM postoperatively [12,13]. Other studies, which include both benign and malignant indications, note a 4%–40.4% incidence of new-onset DM postoperatively [14,15]. The mechanism behind this increase is unclear, as many patients have improved insulin resistance following surgery [16].

The objective of this study is to evaluate whether preoperative hemoglobin A1c (HbA1c) values in patients undergoing PD can help predict the incidence of new-onset or worsening DM. Patients are stratified by HbA1c to determine the effect of preoperative diabetes risk on development of DM postoperatively. We hypothesize that higher preoperative HbA1c will place patients at higher risk for postoperative development of DM.

MATERIAL AND METHODS

We performed a retrospective review of all patients who underwent a PD from January 2011 to December 2017 at a single academic center. Patients with both benign and malignant indications for PD were included. Patients who did not have documented follow-up within the hospital system within 6 months from the date of operation were excluded, as determination of change in DM status was impossible. Patients who underwent total pancreatectomy based on intraoperative margin status were also excluded. Information on patient demographics, DM status, surgical indications, and perioperative outcomes was also obtained through medical record review.

DM was defined by the following criteria: HbA1c value of ≥ 6.5%, documentation of DM diagnosis in preoperative history and physical examination, or treatment with insulin or oral antihyperglycemic agents preoperatively. Patients without a preoperative diagnosis of diabetes but with an HbA1c of 5.7%–6.4% were considered at risk for diabetes in accordance with the American Diabetes Association recommended classification [17]. Worsening DM was defined by an increase in HbA1c level of at least 0.5% or by an escalation in treatment (eg, initiation of oral antihyperglycemic therapy or transition from oral medications to insulin). Improvement was defined as decrease in HbA1c or cessation of need for insulin or oral medications. Diabetes in the postoperative period was defined similarly. Patients were not considered to have postoperative DM unless this diagnosis was made greater than 30 days from the date of operation. Blood glucose and use of insulin in the immediate postoperative period were not recorded, as hyperglycemia in the immediate postoperative period is common and not associated with long-term development of DM [18].

Descriptive statistics were calculated to describe new diagnosis of or worsening DM postoperatively. Categorical variables were analyzed with the χ2 test unless the expected cell frequency was less than 5. In these scenarios, the Fisher exact test was used. Relative risk was also calculated for pairwise group comparisons. The P values for pairwise group comparisons were adjusted for multiplicity using Holm-Bonferroni multiple comparison procedure [19]. Patients were stratified based on diagnosis and preoperative HbA1c. This study was approved by the Institutional Review Board of the University of Utah.

RESULTS

A total of 173 PDs were performed during this study period. Eighty-three patients (48.0%) were excluded for lack of 6-month follow-up (n = 46), lack of preoperative HbA1c value (n = 32), death within 30 days (n = 4), or intraoperative conversion to total pancreatectomy (n = 1). A total of 90 patients met the study inclusion criteria. Median patient age at time of resection was 63 (interquartile range 56–69). Twenty-one patients (23.3%) had a known diagnosis of DM prior to evaluation. Of these, 12 patients (57.1%) were prescribed insulin, 8 patients (38.1%) were using oral antihyperglycemic agents, and 1 patient (4.8%) was not taking medications for DM. There were 7 patients (7.8%) with an HbA1c ≥ 6.5% preoperatively with no known diagnosis of DM. Another 23 patients (25.6%) were at risk for DM preoperatively. Most of the PDs (66.7%) were performed for a malignant diagnosis. Median follow-up time for all patients was 613 days (interquartile range 395–1,079 days).

In the group of patients with DM (either known or based on HbA1c) prior to evaluation, 14 (50.0%, standard error [SE] 9.4%) had worse DM postoperatively. Of the 23 patients at risk for DM preoperatively, 7 (30.4%, SE 9.6%) were diagnosed with DM postoperatively. Only 3 patients (7.7%, SE 4.3%) of the 39 patients with preoperative HbA1c ≤ 5.6% were diagnosed with DM following resection (Table 1). This diagnosis was made for 1 patient 22 months following resection by a rise in HbA1c from 5.3 to 10.8. This patient underwent PD for a benign cystic lesion of the pancreas. The new DM diagnosis was made in the other 2 patients without preoperative DM risk by initiation of oral antihyperglycemic medications within the first 6 months following resection. Diagnosis of new or worsening DM was made by comparison of pre- and postoperative HbA1c in 14 patients (50.0%). New initiation of medications for DM or addition of insulin to patient's medication regimens occurred in 21 patients (75.0%) with new or worsening DM. Postoperative HbA1c laboratory values were available for 29 patients (32.2%). (See Figure)

Table 1

Risk of new-onset or worsening DM, stratified by preoperative DM status

	Patients	Known DM preoperatively	New DM postoperatively	Worsening DM postoperatively
No risk (HbA1c ≤ 5.6%)	39	0	3 (7.7%, 4.3%)	–
At risk (HbA1c 5.7%–6.5%)	23	0	7 (30.4%, 9.6%)	–
DM (HbA1c ≥ 6.5% or known diagnosis)	28	21	–	14 (50.0%, 9.4%)

Reported as total number of patients (percent of patients, SE).

Figure

Risk of new or worsening DM, stratified by preoperative DM status. Error bars represent standard errors.

Preoperative HbA1c values were associated with patient's risk for development of new onset or worsening DM. Patients at risk for diabetes (HbA1c 5.7%–6.4%) were significantly more likely to develop DM following PD than those with no risk for DM (HbA1c ≤ 5.6%) (RR = 4.0, P = .019). Patients with known DM did not have significantly greater risk for worsening of glycemic control than those at risk for DM (RR = 1.6, P = .158) but had significantly worse glycemic control postoperatively compared with those at no risk for DM (RR = 6.5, P < .001).

In all groups, there was no significant difference in newly diagnosed or worsening DM when stratified by underlying pancreatic diagnosis (66.7% for malignant vs 33.3% for benign, P = 1.00). New or worsening DM was also not associated with age at resection (50% if < 30 years of age, 16.7% if 31–50, 31.3% if 51–75, and 0% if 75 +, P = .154). There was no difference related to smoking status (26.3% in nonsmokers versus 23.3% in smokers, P = .760) either. The median time to diagnosis of new-onset DM following resection was 12 months, with only 2 patients receiving a diagnosis greater than 24 months following surgery. The median time to diagnosis of worsening DM was 18 months (Figure, Table 2).

Table 2

Patients with both preoperative and postoperative HbA1c values with absolute and percent change

Preop	Postop	Change
4.8	5.4	0.6	12.5%
5.0	5.6	0.6	12.0%
5.3	10.8	5.5	103.8%
5.3	5.8	0.5	9.4%
5.4	5.3	− 0.1	− 1.9%
5.4	5	− 0.4	− 7.4%
5.5	5.4	− 0.1	− 1.8%
5.6	5.7	0.1	1.8%
5.6	5.9	0.3	5.4%
5.6	6.0	0.4	7.1%
5.7	6.5	0.8	14.0%
5.8	4.5	− 1.3	− 22.4%
5.8	5.1	− 0.7	− 12.1%
6.0	5.5	− 0.5	− 8.3%
6.1	7.5	1.4	23.0%
6.2	6.2	0	0.0%
6.2	7.1	0.9	14.5%
6.2	5.8	− 0.4	− 6.5%
6.2	8.0	1.8	29.0%
6.7	5.6	− 1.1	− 16.4%
6.7	7.7	1	14.9%
6.8	7.7	0.9	13.2%
6.9	6.6	− 0.3	− 4.3%
6.9	7.6	0.7	10.1%
6.9	8.3	1.4	20.3%
7.3	9.3	2	27.4%
7.6	7.2	− 0.4	− 5.3%
9.7	13.2	3.5	36.1%
10.7	15.4	4.7	43.9%

DISCUSSION

This study demonstrates the utility of preoperative HbA1c values in predicting development of new-onset or worsening DM following PD. Few prior studies used HbA1c to stratify patient cohorts by presence and severity of preoperative DM despite this marker being the criterion standard for diagnosis of DM [20]. This cohort had similar rates of DM development following PD compared to other published reports (11.1%) [[12], [13], [14], [15]]. Patients at risk for DM were 4.0 times more likely to develop DM postoperatively than patients with preoperative HbA1c ≤ 5.6%, and patients with known DM were 6.5 times more likely to develop worsening glycemic control following PD. Patients with preoperative HbA1c ≤ 5.6% were very unlikely to develop DM following PD (7.7% of patients). No other variables in this study were helpful for predicting risk of postoperative DM development.

Using HbA1c ≥ 6.5% to define DM, 25.0% of patients with DM preoperatively were unaware of their diagnosis. Although the duration of DM in patients is not known in this study, these patients are unlikely to have had long-standing DM diagnoses. This supports other evidence that new-onset DM may be an early marker of PDAC. The relative risk of being diagnosed with PDAC is 50% greater in patients diagnosed with DM within 4 years of the diagnosis of malignancy [8]. Another study shows that 46% of patients are diagnosed with new-onset DM at the same time as a PDAC diagnosis and 88% of patients are diagnosed within 2 years of the PDAC diagnosis [21].

This study was limited by a small sample size. HbA1c was not available preoperatively for all patients, requiring many patients to be excluded from this analysis. Available data for many patients, such as body mass index and other patient comorbidities, were also limited based on the retrospective nature of this study. Many patients were also excluded secondary to short follow-up times. Many patients treated at this institution live significant distances away and receive follow-up care at hospitals with shorter travel times. Although the sample size could be much larger, excluding patients without at least 180 days of documented follow-up is critical to the validity of the results presented. The median time to new-onset DM development following PD was 12 months. Some previous studies failed to follow postoperative outcomes beyond 30 days, a major limitation considering that up to 45% of patients who develop endocrine insufficiency following pancreatic resection do so at least 90 days after surgery [22]. Immediate postoperative glucose was not included in this study, as many patients develop temporary hyperglycemia following surgery [18,23]. Further studies looking at the relationship of DM and pancreatic surgery should ensure that longer-term outcomes are available.

This study describes a cohort of patients who underwent PD and the subsequent incidence of new-onset or worsening DM following resection, stratified by preoperative HbA1c values. An improved understanding of the relationship between pancreatectomy and DM will help to identify patients at risk for DM postoperatively, which may lead to opportunities to optimize these patients prior to surgery and to encourage patients to engage in healthy behaviors including weight loss, exercise, and dietary modification postoperatively. This knowledge can also help surgeons more accurately discuss the risks of PD with patients before undergoing an operation.

Author contribution

CS and JB conceived of the project idea. JB and HS were responsible for data collection. JB performed the data analysis and wrote the initial draft of the manuscript. All authors contributed to revisions and completion of the final manuscript.

Conflict of interest

None of the authors have any financial disclosures or conflicts of interest.

Funding sources

This study is unfunded.