Actuarial survival curve according to histological nodal involvement (P<.001).
Actuarial survival curve according to natural killer cell activity (50:1 dilution) (P<.001).
González FM, Vargas JA, López-Cortijo C, Castejón R, Gorriz C, Ramírez-Camacho R, Millán I, Durántez A. Prognostic Significance of Natural Killer Cell Activity in Patients With Laryngeal Carcinoma. Arch Otolaryngol Head Neck Surg. 1998;124(8):852-856. doi:10.1001/archotol.124.8.852
For laryngeal carcinoma, the present TNM clinical staging system does not seem completely satisfactory as a guide for providing a prognosis for survival. We believe that natural killer cell activity would probably have a role in a more reliable system. Therefore, we analyzed the disease outcome with previously untreated epidermoid carcinoma of the larynx, evaluating classic clinical and pathologic factors, as well as natural killer cell activity in peripheral blood samples of the patients.
To determine the level of natural killer cell activity in patients with laryngeal carcinoma and to analyze the prognostic value of this finding when associated with other clinical and pathologic variables.
Prospective cohort study of surveillance of laryngeal cancer. Mean follow-up of 42.8 months.
Tertiary care referral center and ambulatory and hospitalized care.
We compared 81 men (mean age, 62.4 years; range, 35-89 years) with laryngeal carcinoma with 44 healthy men serving as control subjects (mean age, 57.6 years; range, 37-82 years).
Natural killer cell activity was significantly reduced in patients who had died of cancer-related causes in comparison with tumor-free survivors. Overall actuarial survival differed significantly in histologically assessed nodal involvement and low natural killer cell activity. Use of the Cox proportional hazards regression model showed that the factors that seem to have a prognostic effect on survival are histologically determined nodal involvement and low natural killer cell activity.
These results support the prognostic significance of the determination of pretreatment natural killer cell activity in peripheral blood samples from patients with laryngeal carcinoma and suggest that assessment of adding such a determination to the current tumor staging system is advisable.
AN EFFECTIVE system for cancer classification and staging must be structured so that information obtained about treatments and outcome for a group of patients with similar lesions will be reliable, accurate, and reproducible1 so results can be analyzed and validated for their prognostic significance.2 For laryngeal carcinoma, the present TNM staging system, which is based on clinical factors, such as the anatomic site, local and regional extension, and the presence of distant metastases, does not seem completely satisfactory as a guide for therapeutic decisions or for providing a prognosis for survival. The pathologic pTNM system for postsurgical classification is an attempt to improve the correlation between staging and prognosis, but it is no substitute for pretreatment classification.3 To achieve a reliable system, it will be necessary to incorporate additional variables, and natural killer (NK) cell activity would probably have a role in this new approach.
Natural killer cell activity is defined as the lytic capacity developed by NK cells against tumor cells and virally infected cells without previous sensitization in a non–major histocompatibility complex–restriction fashion.4 Natural killer cells are large granular lymphocytes that do not rearrange T-cell or immunoglobulin antigen-receptor genes and are identified by their expression of CD16 and/or CD56 membrane antigens.5 Animal studies have shown that NK cell activity is important in the immune system's surveillance against cancer.6
We conducted this study to determine the level of NK cell activity in patients with laryngeal carcinoma and to analyze the prognostic value of this finding when associated with other clinical and pathologic variables.
The study included 81 men with previously untreated epidermoid carcinoma of the larynx (mean age, 62.4 years; range, 35-89 years) who were studied in our department. All patients had a history of cigarette smoking (more than 1 pack a day) and consumed alcohol in moderate to heavy amounts. Supraglottic localization was found in 62 patients (76%) and glottic localization in 19 (23%). Tumors were staged according to the unified criteria of the American Joint Committee on Cancer Staging7 and the International Union Against Cancer.8 Staging was distributed as follows: I, 2 patients; II, 17 patients; III, 25 patients; and IV, 37 patients. There was no correlation between clinical and pathologic findings for local and regional extension except for advanced tumors (T4) (Table 1) and extensive nodal involvement (N3) (Table 2). None of the subjects showed evidence of distant metastases at the time of the study. All patients were treated with laryngeal surgery with associated neck dissection (70 patients) alone or in combination with postoperative radiotherapy (56 patients). The degree of tumor differentiation was as follows: G1 tumors, 52 patients; G2, 21 patients; and G3, 8 patients. The outcome for the patients after a mean follow-up of 42.8 months (range, 6-72 months) was 47 tumor-free survivors (58%), 21 deaths (26%) related to cancer, 6 deaths (7%) related to a second cancer that appeared (eg, lung), 6 intercurrent deaths (7%), and 1 lost to follow-up (1%). Of the patients whose death was related to cancer, 19 had a supraglottic tumor and 2 had a glottic tumor; 15 had local recurrence and 6, metastases. The control group consisted of 44 healthy men with the same epidemiological habits as the patient population relative to smoking and alcohol intake (mean age, 57.6 years; range, 37-82 years).
Twenty- to forty-milliliter blood samples were obtained from each patient by venipuncture using a heparinized syringe. Peripheral blood mononuclear cells (PBMCs) were obtained by the Ficoll-Hypaque gradient centrifugation technique (Nycomed Pharma AS, Oslo, Norway). Interphase cells were washed and resuspended in RPMI 1640 (Microbiological Associates Inc, Walkersville, Md) and supplemented with 10% fetal calf serum (Flow Laboratories, Irvine, Scotland). For all patients, the individual cytotoxicity levels were obtained early in the morning on the day before surgery.
Cytotoxicity was quantified by a chromium 51–specific release assay using the K-562 cell line (human erythromyeloid leukemia cells) as target cells and PBMCs as effector cells. Target cells were labeled by incubating 2 to 3 million cells with 3.7 MBq of 51Cr for 90 minutes at 37°C, after which they were washed 3 times and resuspended in RPMI 1640 supplemented with 10% fetal calf serum at a concentration of 50000 cells per milliliter. Then, 5000 target cells (0.1 mL) were mixed with effector cells (0.1 mL) at different effector-to-target ratios (50:1, 25:1, 12:1, 6:1) in triplicate round-bottomed microtiter wells (Nunc, Roskilde, Denmark). Controls included targets incubated with RPMI 1640 (spontaneous release) or with detergent (total release). Plates were incubated for 4 hours at 37°C in a 5% carbon dioxide atmosphere, and 0.1 mL from each well was collected and counted in a gamma counter. The percentages of specific cytotoxicity were calculated as follows: [(A−S)/(T−S)]×100, where A is the mean counts per minute of test samples; S, the mean counts per minute of spontaneous release; and T, the mean counts per minute of total release. Spontaneous release ranged from 5% to 15%.
The data from the groups (patients vs controls and subgroups of patients) were compared by using the Mann-Whitney U test for unpaired samples and a Kruskal-Wallis 1-way analysis of variance. Actuarial survival was estimated according to the Kaplan-Meier method, and survival was compared between subgroups of patients by using the Cox-Mantel test. A logistic regression model (Cox proportional hazards model) was used to select the variables that concurrently, but independently, influenced patient survival. A P value of less than .01 was considered significant. Data are given as mean±SD unless otherwise noted.
We began by studying the NK cell activity of PBMCs from the patients and the control subjects. As indicated in Table 3, PBMCs from patients showed significantly diminished NK cell activity compared with the control group (P<.001) at the 4 dilutions assessed. No significant differences were observed for local tumor location (supraglottic vs glottic, NK cell activity at the 50:1 dilution [NK50 activity], 27.3% vs 27.8%) (P=.74).
Likewise, when the patients were grouped according to disease status, patients who had died of cancer-related causes had significantly lower levels of NK cell activity than did the tumor-free survivors (Table 3) as in local recurrence (NK50 activity, 24.8±9.7%) as in metastases (NK50 activity, 28.2±10.2%).
To evaluate the significance of these results, we devised an actuarial survival rate curve according to the classic prognostic factors: clinical and pathologic findings for tumor class, T4 vs T1, T2, and T3 and pT4 vs pT1, pT2, and pT3; clinical and pathologic nodal involvement, N0 vs N1, N2, and N3 and pN0 vs pN1, pN2, and pN3; degree of tumor differentiation, G1 vs G2 and G3; and NK cell activity, with cutoff points at the median for each dilution, 36%, 28%, 18%, and 10% for the 50:1, 25:1, 12:1, and 6:1 dilutions, respectively.
No significant differences were observed in the 3-year actuarial survival for clinical tumor class (T4 vs T1, T2, and T3: 0.49±0.10 vs 0.61±0.12; P=.10), pathologic tumor class (pT4 vs pT1, pT2, and pT3: 0.49±0.09 vs 0.67±0.13; P=.15), clinical nodal involvement (N0 vs N1, N2, and N3: 0.65±0.13 vs 0.49±0.09; P =0.4), or histopathological differentiation (G1 vs G2 and G3: 0.59±0.09 vs 0.43±0.15; P =.19) (Table 4). However, when we correlated patient survival with histologically assessed nodal involvement, patients with all degrees of regional extension were found to have a significantly diminished survival compared with patients with no nodal involvement (pN0 vs pN1, pN2, and pN3: 0.88±0.07 vs 0.23±0.10; P <.001) (Figure 1).
The NK cell activity in PBMCs was significantly reduced in nonsurvivors whose death was cancer-related compared with patients with tumor-free survival at dilutions of 50:1 (≥36% vs <36%: 0.83±0.07 vs 0.31±0.11; P<.001) (Figure 2), 25:1 (≥28% vs <28%: 0.84±0.07 vs 0.32±0.11; P<.001), 12:1 (≥18% vs <18%: 0.81±0.10 vs 0.33±0.11; P<.005), and 6:1 (≥10% vs <10%: 0.82±0.06 vs 0.30±0.11; P<.001) (Table 4).
Also, we used the Cox proportional hazards regression model to select the variables that concurrently, but independently, influenced patient survival. The factors that seemed to have a considerable effect on the survival were the histologically determined nodal involvement (relative risk [RR], 12.5; 95% confidence interval [CI], 2.4-59.0; P =.002), the NK cell activity at a dilution of 50:1 (RR, 7.3; 95% CI, 1.8-28.4, P=.004), and the degree of histopathological tumor differentiation (RR, 2.4; 95% CI, 1.01-5.70; P=.06).
Tumor classification is of considerable importance in planning treatment because it provides a prognosis for survival and allows the comparison of results from different series. For laryngeal carcinoma, the present staging system, TNM, based on clinical findings, is imprecise and requires not only refinement, but also the integration of biological determinants of tumor behavior and the host situation.9
Different classifications have been proposed to correlate the tumor type with the biological behavior. They have been based on different tumor measurements,10 on the microscopic features of the lesion,11 or on biological staging.12 The postoperative pathologic classification, pTNM, is an attempt to improve the correlation between staging and prognosis; pathologic examination has demonstrated that 50% to 60% of the tumors are incorrectly classified despite extremely accurate preoperative assessments,13 as observed in our study, but postoperative pathologic classification is no substitute for pretreatment classification. Of the variables commonly used for tumor staging, local and regional tumor extension according to clinical and histopathological findings, only histologically demonstrated nodal involvement has been found to have a meaningful influence on disease outcome.
It seems evident that classifying a tumor is not the same as classifying a patient with cancer, and the clinical classification fails to consider certain important elements, such as host-tumor relationships and concomitant nonneoplastic disease.14
For the host-tumor relationship, the immune status of the patient seems to have a substantial role.15 The cellular arm of the immune response is an important regulator of tumor growth through the different cell subpopulations (T, B, and NK cells and monocytes).16,17 Several clinical reports have indicated a low NK cell activity in PBMCs from patients with a variety of malignant tumors,18- 21 including head and neck tumors.22,23 Others, however, have demonstrated no reduction in NK cell activity in patients with malignant tumors.24,25 The degree of impairment of NK cell activity seems to correlate with disease status, being most marked in patients with advanced cancers.
Cancer of the larynx should be considered a locoregional disease in which joint management of the tumor and the cervical lymph node chains is important. Some studies have correlated regional tumor involvement and low NK cell activity,26,27 having established a range of lytic activity beyond which there seems to be a high probability of regional dissemination of the cancer.28,29
Previous studies have failed to confirm whether the depressed NK cell activity detected in some patients with cancer is secondary to the overt malignant process or whether it precedes the development of the malignant neoplasm,30 although there is convincing evidence to support the suspicion that absent or low NK cell activity in humans or experimental animals is associated with a high frequency of viral disease and cancer.31 Overall, the role of NK cells in the control of the metastatic spread of cancer and in the prevention of metastasis seems to be widely accepted.32
It is probably necessary to determine the relationship between NK cell activity and disease outcome to be able to assess the prognostic value of NK cell activity in patients with malignant tumors, which was the purpose of the present study. There are few studies in this line of research that support the relationship between NK cell activity and prognosis33,34 or find no evidence that it exists.35 In the present study, no significant differences were observed in the disease outcome according to the classic clinical factors of the TNM system; however, there was a correlation with pathologic variables, principally with histologically assessed nodal involvement and, of course, with low NK cell activity. The NK cell activity is the only variable studied with a statistically significant effect on the prognosis that can be determined before any therapeutic decision is reached.
These results support the prognostic value of the pretreatment determination of NK cell activity in PBMCs from patients with laryngeal carcinoma and suggest that assessment of adding such a determination to the current tumor staging system is advisable. Based on the association between deficient NK cell activity and the risk of mortality, the activation of NK cell function by immunomodulators may be of therapeutic value in patients with laryngeal carcinoma.
Accepted for publication March 10, 1998.
This study was financed by grant 92/0560 from Fondo de Investigaciones Sanitarias de la Seguridad Social, Madrid, Spain, and by grant R/1296 from the L.A.I.R. Foundation, Madrid, Spain. Both organizations conducted the study.
We thank Concepción Lorences and Gloria Peraile for their skillful technical assistance and Martha Messman for expert editorial assistance.
Reprints: Francisco M. González, MD, PhD, Service of Otorhinolaryngology, Clínica Puerta de Hierro, C/ San Martín de Porres 4, 28035 Madrid, Spain.