Pretreatment glycemic control status is an independent prognostic factor for cervical cancer patients receiving neoadjuvant chemotherapy for locally advanced disease

After Institutional Review Board approval was obtained from both institutions, a search of clinical databases at Sun Yat-sen Memorial Hospital and the People’s Hospital of Shaolin District was performed. All patients with FIGO stage IB2 and IIA2 cervical cancer (histologically confirmed squamous cell carcinoma, adenocarcinoma and adenosquamous carcinoma) who underwent NACT and type III radical hysterectomy from January 1, 2002 to June 30, 2011 were retrospectively reviewed. Pretreatment informed consent was required for all included patients. Patients younger than 16 years as well as patients who had undergone treatment at other hospitals or who had been treated with chemotherapy or radiation therapy for other malignancies or who did not complete the planned cycles of NACT were excluded from the present analysis.

Two to three cycles of NACT were prescribed. Patients underwent type III radical hysterectomy and pelvic lymphadenectomy within 4 weeks after the administration of the last cycle of NACT. CR was defied as no evidence of viable tumor cells in the tumorous area [ 25]. Post-surgical adjuvant radiotherapy was prescribed according to Sedlis criteria [ 15]. DM was defined according to the American Diabetes Association diagnostic criteria or a patient-reported history of diabetes [ 26]. For comparison, patients were classified into three groups: group I (patients without DM), group II (diabetic patients with good glycemic control, hemoglobin A _1c [HbA _1c] levels prior to NACT <7.0%) and group III (diabetic patients with poor glycemic control, HbA _1c levels prior to NACT ≥7.0%) [ 10, 27]. If there were multiple HbA _1c measurements within 6 months before NACT, the mean value was selected for analysis.

The follow-up was scheduled every 3 months for the first 2 years, then every 6 months for the next 3 years, and every year thereafter. Follow-up visits included complete history, physical examination and Papanicolau smear of the vaginal vault. Follow-up information was obtained from office visits or telephone interviews (Additional file 1: Figure S1. Guide for follow-up and telephone interview). When recurrence was suspected based on clinical findings, imaging studies or biopsies of the suspicious lesions were performed on a case by-case basis.

The primary objective of this study was to determine the association between glycemic control status and recurrence-free survival (RFS). The secondary objective was to investigate the impact of glycemic control on cancer-specific survival (CSS), overall survival (OS) and the rate of CR. RFS, CSS and OS were calculated from the date of NACT until the date of events (recurrence OR death from cervical cancer OR death from any cause) or the date of last follow-up. The Kaplan-Meier method was used to estimate the RFS, CSS and OS curves. The log-rank test was used to test for differences between curve estimates. For multiple comparisons of survival curves among the three groups, the Bonferroni adjustment was applied. Cox proportional hazards model using with enter method was used to determine independent prognostic factors. Hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated. The proportional hazards assumption of each variable was verified by Schoenfeld residual plots and no departures from proportionality were observed. Multivariate logistic regression with enter method was used to identify independent variables predicting CR following NACT. Correlations between CR and assessed variables were expressed as odds ratios (ORs) with 95% CI. Variables reaching statistical significance at the P < 0.15 level in the univariate analysis were entered into the multivariate model. All statistical tests were two-sided, and a two-tailed P value <0.05 was considered statistically significant. Statistical analyses were performed using software (STATA 10.0, special edition; StataCorp, College Station, TX, USA).

A total of 388 patients met study criteria. Table 1 summarizes the patient demographics. The median patient age was 52 years (range: 24–80), and the median body mass index (BMI) was 23.2 kg/m ² (range: 19.4–30.1). Eighty-nine (22.9%) patients had DM. Of these diabetic patients, only 35 (39.3%) had good glycemic control. Group II had three patients with type I DM (8.6%), while group III had six patients with type I DM (11.1%) ( P = 0.698). At baseline, compared with group I, group II and group III had more patients with higher levels of serum creatinine, lymph node metastasis, parametrial involvement, positive surgical margins, lymphatic vascular space invasion (LVSI), deep stromal invasion, hypertension and cardiovascular diseases. In contrast, group II and group III had fewer patients achieving CR following NACT. Among patients with DM, the median HbA _1c level was 7.2% (range: 5.1–12.3%). The median HbA _1c level for patients in group II was 6.4% (range: 5.1–6.9%) compared with 8.3% (7.0–12.3%) for patients in group III ( P = 0.0001). In addition, compared with patients in group II, group III had more patients with a higher level of BMI and serum creatinine, non-squamous cell carcinoma, FIGO IIA2 disease, poor differentiated tumor, parametrial involvement, positive surgical margin, LVIS, deep stromal invasion, cardiovascular diseases and history of using metformin use. More patients in group II achieved CR after NACT compared with group III.

After a median follow-up time of 39 months (range: 4–67 months), recurrence was observed in 72 patients, including 35 (11.7%) patients in group I, 10 (28.6%) patients in group II and 27 (50%) patients in group III. The differences in recurrence rate between the three groups were significant ( P < 0.0001). The median time to tumor recurrence was 18 months (range: 6–45 months). RFS at 5 years was 87.3%, 67.9% and 45.2% for patients in group I, group II and group III, respectively. Fig. 1a shows the Kaplan-Meier curves for RFS by diabetic status. Differences between the three survival curves were significant (log-rank test: χ ² = 62.21, P < 0.0001). A post hoc Bonferroni analysis revealed that the differences in RFS between group I and group II (log-rank test: χ ² = 7.44, P = 0.006), group I and group III (log-rank test: χ ² = 58.76, P < 0.0001), and group II and group III (log-rank test: χ ² = 5.85, P = 0.016) were statistically significant. The cumulative CSS at 5 years was 83.5%, 61.5% and 30.3% for patients in group I, group II and group III, respectively. Fig. 1b displays Kaplan-Meier curves for CSS and significant differences were noted between the curves (log-rank test: χ ² = 69.18, P < 0.0001). A post hoc Bonferroni analysis showed that the differences in RFS between group I and group II (log-rank test: χ ² = 7.56, P = 0.006), group I and group III (log-rank test: χ ² = 67.82, P < 0.0001), and group II and group III (log-rank test: χ ² = 6.28, P = 0.012) were statistically significant. The estimated 5-year OS was 60.5%, 28.5% and 9.5% for patients in group I, group II and group III, respectively. Kaplan-Meier curves for OS are presented in Fig. 1c, and significant differences were noted between the curves (log-rank test: χ ² = 92.63, P < 0.0001). In the post hoc Bonferroni analysis, differences in OS between group I and group II (log-rank test: χ ² = 16.43, P < 0.0001), group I and group III (log-rank test: χ ² = 91.40, P < 0.0001), and group II and group III (log-rank test: χ ² = 6.11, P = 0.013) were significant.

The results of Cox regression analyses are detailed in Table 2 and Table 3. HbA _1c ≥ 7.0% prior to NACT was identified as an independent predictor of RFS (HR = 3.33, 95% CI 1.89–5.88, P < 0.0001), CSS (HR = 3.60, 95% CI 1.96–6.61, P < 0.0001) and OS (HR = 4.35, 95% CI 2.64–7.17, P < 0.0001).

To investigate the survival impact of glycemic control for diabetic patients, we excluded patients without DM and performed a further subgroup analyses. The results of Cox proportional hazard analyses are summarized in Table 4 and Table 5. When the level of HbA _1c was treated as a dichotomous variable, HbA _1c ≥ 7.0% prior to NACT exhibited an independent negative effect on RFS (HR = 2.18, 95% CI 1.02–4.63, P = 0.044) and OS (HR = 2.29, 95% CI 1.20–4.35, P = 0.012). When the HbA _1c level was examined as a continuous variable (Additional file 2: Table S1), it was independently associated with RFS (HR = 1.39, 95% CI 1.13–1.71, P = 0.002), CSS (HR = 1.28, 95% CI 1.04–1.59, P = 0.021) and OS (HR = 1.27, 95% CI 1.08–1.50, P = 0.004).

CR following NACT has been validated as an accurate surrogate endpoint of survival for LACC patients. Given its significance, a further logistic regression analysis was conducted to identify variables that could predict CR after NACT. Table 6 presents the results. Both good glycemic control (OR = 0.06, 95% CI 0.02–0.17, P < 0.0001) and poor glycemic control (OR = 0.04, 95% CI 0.02–0.11, P < 0.0001) were identified as independent predictors of decreasing incidence of CR after NACT.