Preexisting Diabetes and Risks of Morbidity
and Mortality
After Gastrectomy for Gastric Cancer
A Nationwide Database Study
Ming-Shian Tsai, MD, Yu-Chiao Wang, MSc, Yin-Hsien Kao, PhD, Long-Bin
Jeng, MD,
and Chia-Hung Kao, MD
INTRODUCTION
G
astric cancer (GC) is among the most common cancers worldwide and causes considerable mortality, despite the declining trend in its incidence.1,2 GC can be attributed toseveral factors, including genetic factors,3,4 red meat diet,5,6
and Helicobacter pylori infection.7 Surgery, if feasible, is the
only currently available treatment for GC.
Studies have reported that diabetes mellitus (DM) is associated with an increased risk of GC.8–11 In addition, a
previous study demonstrated that patients with DM exhibited a significantly higher risk of GC mortality.12 Furthermore, DM
has been proposed to be an independent determinant of a greater risk of perioperative complications and mortality following
various surgeries.13–15 Our previous study indicated that
patients with DM exhibit higher risks of septicemia and acute renal insufficiency after undergoing hepatectomy for hepatocellular carcinoma.16 However, the effects of DM on the risk of
perioperative morbidity and mortality after a gastrectomy for GC remain controversial.
Several previous studies investigating the surgical risk of patients with GC were limited by small sample sizes or were conducted at a single institution.17–20 Herein, we used data
from the Taiwan National Health Insurance Research Database (NHIRD) to evaluate the postoperative risks of mortality and morbidity in patients with DMundergoing gastrectomy for GC
on a nationwide scale. In addition, we investigated the effects of coexisting medical conditions and DM-related comorbidities on postoperative 30-day mortality in patients with
DM.
MATERIALS AND METHODS
Data Resources
The Taiwan National Health Insurance program was established in 1995 and covers _99% of the 23.7 million residents of Taiwan. We used electronic claims data from the NHIRD to establish a longitudinal cohort. All claims records contain data on allmedical services used by beneficiaries of theNHI program, are released by the Bureau of National Health Insurance for public research, and are managed by the National Health Research Institutes. We analyzed data from the 2007–2011 Registry of Catastrophic Illness Database (RCIPD) of the NHIRD in this study. The RCIPD contains records on all confirmed cases of catastrophic illness, such asGC. Patients with catastrophic illness must submit pathology and related laboratory reports to receive a catastrophic illness certificate, which exempts these patients from copayments for all ambulatory visits and hospitalizations. The International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM)was used to code diagnoses of diseases and gastrectomy procedures. The NHIRD files were connected with surrogate identification numbers to protect personal privacy. The research ethics committee of the institutional reviewboard of China Medical University exempted this study from full review (CMU-REC-101–012).
Study Population
Our study was a population-based retrospective cohort study. We identified patients with GC (ICD-9-CM code: 151) who had undergone a partial gastrectomy (ICD-9-CM codes: 43.5, 43.6, 43.7, 43.81, 43.82, and 43.89) or total gastrectomy (ICD-9-CM codes: 43.91 and 43.99) according to the RCIPD records. All patients were aged over 20 years and underwent gastrectomy between 1999 and 2010. The date on which a patient first underwent gastrectomy served as the index date.We identified 6284 patients who were diagnosed with DM (ICD-9-CM code: 250) for the first time before the index date as the DM cohort. The comparison cohort (n¼6269) comprised patients
without a history of DM who were randomly selected from the RCIPD and matched with the patients with DM in a 1:1 ratio according to age (each 5-year span), sex, and index year (supplementary Figure 1, http://links.lww.com/MD/A411). We also used the propensity score matching method to
select the DM and control cohorts. In this analysis, we matched the 2 cohorts in their baseline characteristics and risk factors (including age, sex, index year, monthly income, occupation, and the comorbidities history) in the 1:1 ratio. We were able to identify 3996 patients in DM and non-DM cohorts after propensity score matching (see supplementary Tables 1 and 2,
http://links.lww.com/MD/A411).
Exposure Variables
We analyzed the sociodemographic information recorded before the index date, namely age, sex, monthly income, and occupation, as well as comorbidities, namely hypertension (ICD-9-CM codes: 401–405), coronary artery disease (CAD, ICD-9-CM codes: 410–414), chronic obstructive pulmonary disease (COPD, ICD-9-CM codes: 490–496), stroke (ICD-9-CM codes: 430–438), hyperlipidemia (ICD-9-(ICD-9-CM code: 272), liver cirrhosis (ICD-9-CM codes: 571.2, 571.5, and 571.6), and chronic kidney disease (CKD, ICD-9-CM codes: 580–589). The monthly income of individuals could be deduced
according to their premium paid to the insurance bureau. For those with monthly income less than NT$15,000, the premium was NT$900 per month. On the contrary, those who had monthly income higher than NT$25,000 would pay more than NT$1500 per month as the premium.
We calculated other potential risk factors of 30- and 90-day postoperative mortality present in the DM cohort between the date ofDMdiagnosis and the index date, namely the type ofDM (type 1 ICD-9-CM codes: 250.x1 and 250.x3; type 2 ICD-9-CM codes: 250.x0 and 250.x2) by defining patients with ketoacidosis (ICD-9-CM code: 250.1), coma (ICD-9-CM codes: 250.2
and 250.3), renal manifestations (ICD-9-CM code: 250.4), eye involvement (ICD-9-CM code: 250.5), and peripheral circulatory disorder (ICD-9-CM code: 250.7).
The 30- and 90-day postgastrectomy mortality in the DM and non-DM cohorts was the primary outcome. The secondary outcomes included 10 major surgical postoperative complications that occurred within 30 and 90 days after operation,
namely septicemia CM code: 038), pneumonia (ICD-9-CM codes: 480–487), stroke (ICD-9-(ICD-9-CM codes: 430–438), acute renal failure (ARF) (ICD-9-CM code: 584), wound infection (ICD-9-CM code: 998.59), wound dehiscence
(ICD-9-CM codes: 998.83 and 998.3), acute myocardial infarction CM code: 410), intraabdominal abscess
(ICD-9-CM code: 682.2), intestinal obstruction (ICD-9-(ICD-9-CM codes: 560.30, 560.39, and 560.8), and other intestinal complications (ICD-9-CM codes: 534, 564.2, 564.3, and 997.49).
Statistical Analysis
We described the distributions of age, sex, monthly
income, and occupation, as well as hypertension and compared these distributions between the DM and non-DM cohorts by using the x2 test and Student t test. Adjusted odds ratios (ORs)
and 95% confidence intervals (CIs) were calculated using multivariate logistic regression to assess the association between DM and the risk of postoperative complications and mortality following a gastrectomy after we adjusted for age, sex, monthly income, occupation, and comorbidity history. The
impact of differences in DM severity on 30- and 90-day postoperative mortality was determined using multivariate logistic
regression, and DM severity was partitioned into 7 categories, namely ketoacidosis, coma, renal manifestations, eye involvement, peripheral circulatory disorders, type 1 DM, and type 2
DM. We also used multivariate logistic regression to examine the association between postoperative mortality (30- and 90-day) and coexisting hypertension, hyperlipidemia, ischemic heart disease, mental disorders, COPD, CKD, liver cirrhosis, hepatitis B, hepatitis C, stroke, and liver cancer in the DM cohort, and compared this association with that in the non-DM cohort.
We also analyzed the association between the adjusted OR of mortality (30- and 90-day) and the interaction between DM and total gastrectomy by using multivariate logistic regression.
Logistic regression analysis was applied to calculate the propensity score and estimate the probability of theDMassignment
base on baseline variables including index year, sex, age, residential area, parental occupation, and history of comorbidity. Finally, we used the logistic regression model stratification of the matched pairs to estimate the risk of postoperative complications and mortality between DM and non-DM cohorts. All analyses were performed using the SAS 9.4 package (SAS Institute Inc, North Carolina). P<0.05 in 2-tailed tests indicated significance.
RESULTS
TheDMcohort and the non-DMcohort comprised 6284 and 6269 patients, respectively, who received a gastrectomy for GC during the study period (1999–2010). Table 1 shows the demographic characteristics and comorbidities of the 2 cohorts.
Because we matched the 2 cohorts in age and sex, no significant difference in age and sex existed between the cohorts (Table 1). However, patients with DM tended to have a lower monthly income than did patients withoutDM(Table 1). As expected, the prevalence of comorbidities, namely hypertension, CAD, COPD, stroke, hyperlipidemia, liver cirrhosis, and CKD, was significantly higher in the DM cohort than in the non-DM cohort.
Next, we analyzed the postgastrectomy risks of morbidity and mortality in both cohorts. Multiple logistic regression
model analysis revealed that, compared with the patients without DM, the patients with DM did not exhibit excessive risks of day postgastrectomy mortality or morbidities (including 30-and 90-day postoperative mortality), as shown in supplementary Table 1, http://links.lww.com/MD/A411.
To further verify our observation, we used the second
method, propensity matching method, to select theDMand non-DM cohorts (n¼3996, respectively) from our database. As expected, there existed no differences in baseline characteristics and comorbidities between DM and non-DM cohorts after
propensity matching (supplementary Table 2, http://links.lww.-com/MD/A411). In the analysis after propensity score matching, we still did not find any statistically significant difference in postoperative complications and mortality (including 30 and 90 days) between the cohorts (supplementary Table 3, http://
links.lww.com/MD/A411), which was consistent with the results obtained by the first analysis.
We further investigated whether DM-related complications are associated with the 30- and 90-day postgastrectomy risk of mortality (Tables 2 and 3). Patients who experienced a DMrelated coma before undergoing a gastrectomy were associated
with a significantly higher risk of 30-day postoperative mortality compared with the patients in the non-DM cohort (adjusted OR 2.46, 95%CI 1.10–5.54, Table 2). On the contrary, no difference was observed in the 30-day mortality rate between the patients withoutDMand patients with otherDM-related morbidities, such as ophthalmological involvement, peripheral circulatory disorder, ketoacidosis, and nephropathy (Table 2). Compared with
the non-DM cohort, the risk of 90-day postoperative mortality is higher in DM patients with eye involvement (adjusted OR 1.50, 95%CI 1.08–2.06), peripheral circulatory disorder (adjusted OR 1.69, 95%CI 1.23–2.33),coma (adjustedOR2.02, 95%CI 1.24– 3.30), and renal manifestations (adjusted OR 1.31, 95%CI 1.01– 1.70) respectively. (Table 3)
We then investigated whether the risk of postoperative mortality is associated with coexisting morbidities in patients with DM (Tables 4 and 5). Most of the comorbidities, including hypertension, CAD, COPD, stroke, hyperlipidemia, and CKD, were not associated with an increased postoperative mortality risk in the diabetic patients, compared with the non-DM patients. However, patients with DM and concomitant cirrhosis exhibited a significantly higher 30-day postoperative mortality risk than did those without DM (adjusted OR 2.08, 95% CI 1.01–4.27, Table 4). But in risk of 90-day postoperative
mortality, the DM cohort with particularly comorbidity (including hypertension, hyperlipidemia, CAD, CKD, COPD, stroke,
and liver cirrhosis) had no statistically significant higher than non-DM cohort (P>0.05; Table 5).
Next, we addressed whether diabetes is associated with different 30- and 90-day mortality risks in patients undergoing different extent of gastrectomy (ie, total gastrectomy or not). Based on the ICD coding, we were able to identify the subsets of patients undergoing total gastrectomy or not. Supplementary
Tables 4 and 5, http://links.lww.com/MD/A411 illustrate the joint effects of total gastrectomy and DM on the OR of 30- and 90-day mortality. Not surprisingly, we found that total gastrectomy was associated in marginal increase in the 30-day
mortality rate (adjusted OR 1.39, 95% CI 0.93–2.09; supplementary Table 4, http://links.lww.com/MD/A411) and significantly
elevated risk of 90-day mortality (adjusted OR 1.30, 95% CI 1.06–1.60; supplementary Table 5, http://links.lww.com/ MD/A411) in nondiabetic cohorts. However, DM is not associated with increased 30- or 90- day mortality no matter whether the patients underwent total gastrectomy or not.
DISCUSSION
According to our research, this study was the first nationwide population-based study to investigate the effects of DM on
the risk of surgical mortality and morbidity in patients undergoing resection for GC. We observed that the patients with
DM did not exhibit increased risks of major surgical complications or mortality compared with the patients without DM
after we adjusted for covariates. However, patients with a history of a DM-related coma exhibited an _2-fold increased risk of postoperative mortality. However, other preoperative DM-related complications were not significantly associated
with an increased risk of postoperative mortality after a gastrectomy. The study results, which were obtained using
a nationwide data set, indicated that GC surgery can be performed safely in patients with DM. However, attention should be devoted to perioperatively monitoring and managing patients with a history of DM-related neurologic symptoms.
Overall, our data indicated that the 30-day postoperative mortality rates were 1.63% for the non-DM cohort and 1.80% for the DM cohort; these rates are 2-fold to 3-fold higher than the mortality rates reported in randomized controlled studies.
21,22 Several studies have demonstrated that the overall
inhospital mortality rate can be <1% when experienced surgeons perform the operation, irrespective of the extent of lymph
node dissection.19,21,22 By contrast, a case series of 434 patients
The results of the present study were based on nationwide data and, therefore, may more accurately reflect the results of GC surgery in general practice.
Postoperative infection is among the major complications
that can lead to severe sequelae and death following gastrectomy.
23 Severe postoperative infection can be due to pneumonia,
bloodstream infection related to the central line, urinary tract infection, infected intraabdominal fluid accumulation, and anastomotic insufficiency.20,21 Patients with DM are generally
susceptible to infection,24 and previous studies have proposed
that DM is a prognostic factor of postoperative infection events.25,26 However, our data indicated that preoperative
DM is not related to postoperative infection events; this finding is consistent with that of another study reporting that postoperative blood sugar, but not preexisting DM, predicts postoperative
infection.27 We recommend that future studies
investigate whether our finding can be explained partly by greater postoperative glycemic control, which reduces the risk of postoperative septicemia.28
Postoperative ARF is not an uncommon complication after
major operations such as gastrectomy. Depending on the definition of ARF, the incidence of ARF varies among studies,
ranging between 1% and 30% after cardiac bypass surgery29
and between 5.8%and 6.7%after noncardiac surgery.30,31 In this
study, the incidence of ARF after gastrectomy was 0.49% in the non-DM cohort and 0.68% in the DM cohort. Theoretically, DM might predispose patients to ARF development through various pathophysiological mechanisms, such as renal hemodynamic instability,32 endothelial cell injury,33 and inflammatory processes. 34 Our previous study suggested that DM is related to
an increased incidence of posthepatectomyARF,16 which was not
observed in the present study. Our results may be explained by improved perioperative management techniques, including the maintenance of hemodynamic stability during operation, amelioration of surgery-related inflammation, avoidance of vasoactive
or nephrotoxic agents, and prevention of postoperative
dehydration. Postoperatively, cardiovascular events are often difficult to diagnose andmay be overlooked because atypical symptomsmay
be masked by analgesics and physicians may not be alert. One study estimated that only 53% of patients with myocardial infarction exhibit a clinical sign or symptom that triggers a physician’s awareness.35 No significant difference in myocardial
infarction between the DMand non-DMcohorts was observed in this study, possibly because myocardial infarction is underdiagnosed in patients with DM, and perioperative care has become
sophisticated. Patients with DM have blunted pain perception, causing diagnosing myocardial infarction to be difficult.
Balanced fluid therapy, blood pressure control, and painmanagement are essential for preventing postoperative cardiovascular
events. Moreover, clinicians can currently use numerous medications that lower the risk of cardiovascular events.35 Additional
studies are required to clarify our observations.
In the present study, although total gastrectomy seemed to be associated with a decreased 30-day mortality in DM cohort, further analysis of 90-day mortality showed the trend disappeared. Moreover, total gastrectomy is associated with
increased 90-day mortality in nondiabetic cohort (supplementary Table 5, http://links.lww.com/MD/A411). It is welldocumented that total gastrectomy is associated with significant
higher risks of postoperative morbidities and mortality, in comparison with subtotal gastrectomy.36,37 Total gastrectomy
usually carries more unfavorable factors for postoperative recovery, such as technical difficulty, poorer oral intake due to lack of food reservoir, and more extensive lymphadenectomy, as well as possible splenectomy.37,38 Based on our database,
we could not analyze the distribution and/or percentage of these unfavorable factors in the diabetic and nondiabetic cohorts. Further studies are therefore warranted.
We were unable to analyze some important preoperative
parameters, which may affect the postoperative outcomes, such as cancer stages, performance status, activity of daily life, and American Society of Anesthesiologists score. Presumably, the DM group should have worse preoperative physical status than the non-DM group because the prevalence of comorbidities was higher in the DM group. Despite these unfavorable factors, we still could not demonstrate poorer surgical outcomes after
gastrectomy in the DM group. Our finding suggested that DM per se may not affect the surgical results for those patients undergoing gastrectomy for GC.
We did not investigate the impact of cancer stage on 30-and 90-day surgical mortalities in the present study, due to lack of such data in NHIRD. The stage of cancer is a critical factor affecting not only long-term survival but also surgical outcomes. Although diabetes is associated with increased risk of
GC,8–11 it remains unknown whether diabetic patients tend to
have more advanced GC. Theoretically, surgery could be more difficult in patients with more advanced GC, which usually presented with larger tumor size, poorer nutritional conditions, and more extensive lymph node metastasis. The patients with more advanced GC therefore tend to experience more extensive surgical resection. As shown in supplementary Tables 4 and 5, http://links.lww.com/MD/A411, patients undergoing total gastrectomy exhibited higher surgical mortality risks than those
who did not have total gastrectomy. Our findings supported that surgical risks may be higher in patients with later stages of GC. However, it has been shown that extended lymph node dissection did not increase the mortality or morbidity rate of resection for GC in experienced hands.39 Further study will be necessary
to address the effects of cancer stage on surgical mortality after resection of GC, especially in diabetic patients.
Some of the limitations should be mentioned which may be the unknown confounding factors or unmeasured biases relevant to this study: the NHIRD does not record a patient’s individual data of the smoking amount, alcohol use, liver function, blood loss, and gastrectomy reconstruction method; this is a retrospective cohort study (unlike prospective studies); although the data from the inpatient database were strictly audited for the reimbursement purpose, we could not use the individual chart review for every subject to confirm the diagnoses’ accuracy of DM, GC, and complications; a potentially
selection bias should be considered because the study patients were identified by the inpatient admission with severe diseases or increased access to hospital care (such as these patients with mild DM and patients without hospital cares might have been
excluded in this study). Although the above study limitations could cause the underestimation or overestimation of the risk of postoperative complications in DM patients, they should not produce the key bias based on the universal and populationbased health insurance data.
In summary, we found that the risk of 90-day mortality
after gastrectomy for GC is higher in patients with certain DMrelatedmanifestations than thosewithoutDM. This nationwide
study suggested that DM-related eye involvement, coma, peripheral circulatory disease, and renal manifestations are prognostic of poor surgical outcomes after gastrectomy for GC.
