Impact of type 2 diabetes mellitus on short-term and long-term outcomes of patients with esophageal squamous cell cancer undergoing resection: a propensity score analysis

Study design

The study included patients with histologically proven ESCC who underwent conventional open esophagectomy with 2-field lymphadenectomy at the First Affiliated Hospital of Sun Yat-sen University, the Second Hospital of Lanzhou University, or Sun Yat-sen University Cancer Center between January 2001 and December 2010. Patients were excluded if they (1) had received preoperative or/and postoperative chemotherapy and/or radiotherapy; (2) presented with synchronous primary tumors and/or had previous malignant diseases; (3) underwent R1 or R2 resection; or (4) underwent subtotal/total gastrectomy or had other complications, including hyperthyroidism or active tuberculosis.

A diagnosis of diabetes mellitus is based on a patient's history of diabetes mellitus and/or the use of antidiabetic mediations in the hospital medical record (all diabetes diagnoses should be made on the basis of classification and diagnosis of Diabetes released in 2015) [17]. Blood sugar levels were rigorously controlled by insulin injections during the perioperative period.

To minimize the effect of potential confounders on selection bias, propensity score-matching analysis was performed. Factors related to combined treatment were entered into the propensity model; these factors included age, sex, smoking status, drinking status, weight loss rate, body mass index (BMI), tumor location, tumor grade, pT category, pN category, pathological tumor/node/metastasis (pTNM) stage, and complications. Patients with and without T2DM were matched one-to-one using the optimal matching method. The adjusted comparisons by propensity scores were based on data from the patients with T2DM. After adjustment for these factors, overall survival (OS) rates, baseline characteristics, and were re-analyzed to explore the prognostic factors for the patients with and without T2DM.

Preoperative evaluation included a complete history, a physical examination, complete blood cell counts, serum biochemistry, chest X-rays, barium swallow test, computed tomography (CT) scans of the chest and abdomen, ultrasonic esophagoscopy, and mandatory cardio-pulmonary assessments.

Follow-up

Postoperative follow-up and CT scans were performed every 3 months during the first year, every 6 months during the next 2 years, and once every year thereafter. Contrast CT scans of the lower neck, thorax, and upper abdomen, including the liver and adrenal glands, were routinely performed.

Survival status was verified using the best available methods in December 2015, including reviews of outpatient records and direct telecommunication with each patient or his/her family. T2DM history, weight loss, BMI, smoking status, and drinking status were determined by reviewing each patient's medical records at the first visit to a hospital. Weight loss rate was defined as the following formula [18]:

weight1: body weight measured on the first day of admission.

weight2: body weight measured before leaving the hospital.

Because few reports have defined weight loss rate, this parameter was empirically divided into four categories: 0, > 0 to ≤ 5%, > 5% to ≤ 10%, and > 10%.

Complications were estimated from clinical, laboratory, and imaging data. The pTNM stage of each patient was defined according to the criteria of the American Joint Committee on Cancer staging system (2010).

Statistical analysis

All statistical analyses were performed using the SPSS 19.0 software package (IBM, Chicago, IL, USA). OS was defined as the time from the date of surgery to the date of death or final clinical follow-up. All patients lost to follow-up or being alive at the last follow-up were censored. Associations between T2DM and clinicopathologic characteristics were assessed using the χ² or Fisher's exact test. Survival rates were calculated with Kaplan–Meier curves and compared with log-rank tests. Univariate and multivariate cox regression analyses were controlled for the variables mentioned above. A two-sided P value < 0.05 was considered statistically significant. Receiver operating characteristic (ROC) curve was constructed using Medcalc 12.0 (Medcalc Company, Brussels, Belgium). Propensity score-matched analysis was performed using R software (TIBCO, Silicon Valley, CA, USA).

Patient characteristics

Between January 2001 and December 2010, a total of 2489 patients with ESCC underwent open esophagectomy at the three institutions. The study cohort consisted of 862 patients with ESCC who underwent surgery as primary therapy; 638 (74.0%) of them were male (Fig. 1). Before propensity score matching, the mean age of all patients was 60.1 ± 9.2 years. The median weight loss rate and BMI of these patients were 0.0% (range, 0–18.8%) and 22.06 kg/m² (range, 14.98–32.96 kg/m²), respectively. After propensity score matching, mean age was 61.8 ± 9.0 years, the median weight loss rate was 3.1% (range, 0–18.8%), and the median BMI was 22.75 kg/m² (range, 14.98–32.96 kg/m²). The characteristics of the 280 matched T2DM and non-T2DM patients are shown in Table 1. All further statistical analyses were performed on this population.

Patient characteristics by T2DM

Of the 862 patients, 280 (32.5%) had T2DM and 582 (67.5%) did not. These two groups were well matched in weight loss rate, tumor location, tumor grade, pT category, pN category, and pTNM stage. However, the patients with T2DM were significantly older (P = 0.012), had higher percentage of smoking (P < 0.001), had a significantly higher percentage of female (P = 0.024), had a significantly higher BMI (P = 0.006), had significantly higher probability of no complications (P = 0.018), and had higher rates of postoperative anastomotic leakage (P = 0.001) than the patients without T2DM. After propensity score matching, all characteristics in the two groups were well matched, expect for the probability of no complication (P = 0.004) and postoperative anastomotic leak (P < 0.001) (Table 1).

Overall survival and T2DM

The median and mean OS were 45.5 and 39.0 months, respectively. The 1-, 3-, and 5-year OS rates for the entire patient cohort were 86.6%, 53.2%, and 30.2%, respectively (Fig. 2a). Log-rank test showed that the OS rate tended to be higher in ESCC patients with T2DM than in those without T2DM, but the difference between them was not significant (P= 0.563, Fig. 2b).

The modeled area under curve (AUC) for weight loss rate was 0.593 (P < 0.001, 95% confidence interval [CI] 0.551–0.634; Fig. 3). The diagnostic value of the weight loss rate was significantly associated with AUC, with the cutoff points for weight loss rate being 5.05%; the sensitivity and specificity of cutoff points were 0.616 and 0.595, respectively.

Subgroup analysis of patients with low weight loss rate (≤ 5.05%) showed that survival was significantly longer in patients with T2DM than in those without T2DM (P= 0.003, Fig. 4a). In contrast, analysis of patients with high weight loss rate (> 5.05%) showed that survival was significantly shorter in patient with T2DM than in those without T2DM (P= 0.001, Fig. 4b).

Univariate and multivariable analyses

Univariate analysis showed that T2DM was not significantly associated with patient survival. In contrast, sex (hazard ratio [HR] 0.769; P = 0.030), weight loss rate (HR 1.168; P = 0.005), tumor grade (HR 1.489; P < 0.001), pT category (HR 1.642; P < 0.001), pN category (HR 1.820; P< 0.001), and pTNM stage (HR 2.270; P < 0.001) were significantly associated with the survival of patients with ESCC (Table 2). In multivariate analysis, only tumor grade (HR 1.373; P< 0.001) and pN category (HR 1.474; P= 0.001) were significantly associated with patient survival (Table 2).