Survival impact of pre-treatment neutrophils on oropharyngeal and laryngeal cancer patients undergoing definitive radiotherapy

We identified 356 patients with squamous cell carcinoma of the oropharynx or larynx diagnosed between 2006 and 2015 with available follow up and treatment outcomes who were treated with curative-intent radiation therapy (RT) within the University of Colorado Hospital system (n = 336) and the University of New Mexico Hospital (n = 20) systems. In order to ensure adequate assessment of response to treatment without potential confounding of post-treatment changes on imaging scans, patients were excluded if they had fewer than 6 months of follow up from completion of treatment [29]. Patients must have received minimum of two blood draws for complete blood count with differential during the period stretching from 1 month prior to initiation of treatment, including chemotherapy and radiation, through 1 month post-treatment. Additionally, patients were excluded if the radiation treatment course was prolonged by greater than 1 week past the planned completion date. A total of 196 patients met these criteria.

All patients received external beam radiation to a minimum dose of 54 Gray (Gy) (range 54–70 Gy, median 69.3 Gy, standard deviation (SD) 3.98 Gy) using 1.8–2.25 Gy/fraction. A total of 5 patients received less than 60 Gy. The median number of radiation treatments was 33 (range 27–35) over an average of 48 days. Potentially relevant patient and treatment characteristics were included. Age at diagnosis was analyzed categorically based on the approximate median age of 60 years. Race was categorized as white or other. Performance status was categorized according to the Karnofsky Performance Score (KPS) as ≤80 or >80 [30]. T-classification and N-classification were recorded according to the American Joint Committee on Cancer (AJCC) classification 2002 and 2010 editions based on date of diagnosis. Tobacco use was divided into <10 or ≥10 pack years [31]. Patients were considered post-operative if they had received any of the following: subtotal resection, gross total resection, or neck dissection (ipsilateral or bilateral). Chemotherapy was analyzed as a binary variable (yes/no). Additional analyses included steroid use as a binary variable with >100 mg oral equivalents of dexamethasone versus <100 mg and chemotherapy regimens stratified into none, cisplatin weekly, cisplatin every 3 weeks (q3), carboplatin plus paclitaxel and cisplatin plus cetuximab.

The cohort was divided into tertiles by neutrophil count (tertile 1: ≤3.2 × 1000 mm³ (n = 65), tertile 2: 3.2–4.4 × 1000 mm³ (n = 66) and tertile 3: >4.4 × 1000 mm³ (n = 65), and Kaplan–Meier analysis was performed. For multivariate analysis (MVA), the neutrophil and lymphocyte counts were assessed as continuous variables. The primary endpoints were OS, CSS, LC and DC. Survival was determined from the date of diagnosis to the date of death or most recent follow-up. Cancer-specific survival was defined as cancer survival unrelated to other causes of death. Local failure included recurrence or progression in the primary tumor bed and surrounding tissues or regional lymph nodes. Distant failure was defined as a site of disease outside of the tumor bed and regional lymph nodes that was not present during initial therapy.

Three time-periods were defined during the treatment course for collection of absolute neutrophil count (ANC) and absolute lymphocyte count (ALC): (1) 1 month prior to initiation of any treatment (pre-treatment), (2) 1 week to 1 month post-treatment (post-treatment) and (3) containing all time periods from pre-treatment to post-treatment (overall). For the collection of pre-treatment counts, values were taken from the day of treatment initiation whenever possible (190 of 196 patients). For the remaining time points, the lowest value, or nadir, of the ANC and ALC during each respective time period were collected. Analyses were performed at each time period for the ANC, ALC and NLR. The data was then combined over all time-periods and the lowest ANC and ALC observed were selected to represent the overall nadir for analysis. Additionally, the absolute difference in neutrophil count from pre-treatment to post-treatment was collected for analysis.

All statistical analyses were performed using SPSS V23.0 (SPSS Inc., Chicago, IL). Pearson Chi square tests were used to assess associations between categorical variables and blood counts. Median follow up was calculated using the reverse KM method [31]. OS, CSS, LC and DC were first examined using the KM method. Univariate survival analysis (UVA) was performed with the log-rank test and unadjusted Cox proportional hazards models to estimate hazard ratios (HR), with HR >1 corresponding to worse OS, CSS, LC and DC. Patient and clinical variables were selected a priori. Multivariate Cox regression analysis was performed on all blood counts found to be significant (p < 0.05) on UVA using OS, CSS, LC and DC as outcomes with a significance level of p < 0.05.

Univariate analysis p values of each patient characteristic for all endpoints are presented in Table 2. OS was significantly worse for patients with tobacco use ≥10 pack-years, T-3–4 and HPV-negative status. CSS was worse among those with KPS ≤80, and T-classification 3–4. A decrease in LC was seen in patients with T-classification 3–4 and HPV-negative status. There were no factors that independently correlated to decreased DC. Factors that did not influence any endpoints include age, primary site, gender, race, nodal status, post-operative status and receipt of chemotherapy. Pearson Chi square analysis of chemotherapy regimens with overall neutrophil nadir did not indicate a significant correlation (0.039, p = 0.60). Similarly, no correlation was observed between receipt of steroids and overall neutrophil nadir during the treatment period (r = 0.18, p = 0.14). Univariate analysis of RT dose as a continuous variable did not reveal a significant correlation with OS or CSS (HR, 1.00; 95% CI 1.00–1.01; p = 0.41 and HR, 1.00; 95% CI 0.99–1.00; p = 0.91, respectively).

Two-year OS progressively diminished with higher pre-treatment neutrophil nadir with tertiles 1, 2 and 3 at 93.0, 86.0 and 78.5%, respectively (p = 0.03). These results are shown in Fig. 1a. On both UVA and MVA, all neutrophil time-points (pre-treatment, post-treatment and overall nadir) correlated a higher neutrophil count to worse OS (Table 3). The pre-treatment NLR was additionally predictive of OS on both UVA (HR, 1.07; 95% CI 1.01–1.14; p = 0.02) (Table 3) and MVA (HR, 1.09; 95% CI 1.01–1.17; p = 0.03) (Table 4). A summary of UVA for all outcome measures is shown in Table 5. The MVA for pre-treatment neutrophil count and OS including all patient characteristics is shown in Additional file 1: Table S2.

Similar to the OS findings, 2-year CSS progressively diminished with higher pre-treatment neutrophil nadir with tertiles 1, 2 and 3 at 86.0, 76.0 and 73.5%, respectively (p = 0.009) Fig. 1b). This was consistent on MVA with pre-treatment neutrophil count portending poor CSS (HR, 1.32; 95% CI 1.10–1.59; p = <0.01) (Table 4).

Pre-treatment neutrophil count in tertiles 1 through 3 indicated that 2-year LC was 92.5, 84.5 and 82.5%, respectively, which was not statistically significant (Fig. 1c). On both UVA and MVA, a higher post-treatment neutrophil count correlated to worse LC (HR, 1.47; 95% CI 1.16–1.86; p = <0.01 (Table 3) and HR, 1.58; 95% CI 1.21–2.07; p = <0.01 (Table 4), respectively).

Between tertiles 1 and 2, 2-year DC for pre-treatment neutrophil count was 96 and 93%, respectively, while tertile 3 was 79% (p = <0.01) (Fig. 1d). On MVA, a higher pre-treatment neutrophil count was also shown to predict worse DC (Table 4). On both UVA and MVA, a higher overall NLR correlated to worse DC (Tables 3, 4, respectively).