Usage of Cox-Regression Model for Forecasting of Survival Rate in Patients with the Early Stage of Non-Small Cell Lung Cancer

In the past decades a lot of investigations were focused on searching for more accurate markers of lung cancer progression. Researchers indicate that molecular markers may be useful in forecasting of treatment outcome and overall survival rate in patients with non-small cell lung cancer. The aim of our research was to create a forecasting model in order to identify patients with stage I-II of non-small cell lung cancer and dismal prognosis. Our research covered 254 patients with the early stage of non-small cell lung cancer who underwent a cure from June 2008 till December 2012 in the Department of Thoracic Surgery of Zaporizhzhia Regional Clinical Oncologic Dispensary. Surgery was performed for all patients. Adjuvant chemotherapy was performed for 101 patients. In order to carry out multivariate Cox-regression analysis, STATISTICA 6.0 (StatSoft Inc.) program was used. The most significant from 39 variables were selected (tumor size, histological form of tumor, volume of surgical intervention, volume of conducted lymph node dissection, Ki-67 expression, EGFR expression, E-cadherin expression). We propose the computer system which can forecast survival rate in patients with the early stage of non-small cell lung cancer.


Introduction
It is well-known that lung cancer is subdivided into non-small cell lung cancer (NSCLC) which makes approximately 80% -90% from all malignant lung tumors and small cell lung cancer which makes 10% -20% respectively.Fundamental type of treatment for patients with the early-stage of NSCLC is surgery.However, 22% -50% of patients have local recurrences after surgery and 48% -78% of patients have distant metastases [1]- [5].
In the past decades a lot of investigations were focused on searching for more accurate markers of lung cancer progression.Researchers indicate that molecular markers may be useful in forecasting of treatment outcome and overall survival rate in patients with NSCLC [4] [6]- [9].Currently large number of markers are offered for patients with NSCLC but none of them shows its high significance in clinical practice [6].Further research of molecular forecasting markers is needed.By that time their significance in routine practice is not adequate [3].
The aim of our research was to create a forecasting model in order to identify patients with stage I-II of NSCLC and dismal prognosis.

Materials and Methods
Our research covered 254 patients with the early stage of NSCLC who underwent a cure from June 2008 to December 2012 in the Department of Thoracic Surgery of Zaporizhzhia Regional Clinical Oncologic Dispensary.Surgery was performed for all patients.Adjuvant chemotherapy was performed for 101 patients.
In order to carry out multivariate Cox-regression analysis STATISTICA 6.0 (StatSoft Inc.) program was used.39 variables (sex, age, height, weight, body mass index (BMI), BMI (categorized index), localization of the tumor and its clinical form (lung, lobes of the lung, central/peripheral lung cancer), tumor size, tumor size (categorized index), criterion "T", criterion "N", stage of disease, histological and morphological form of tumor differentiation, presence of necrosis in tumor, infiltration of visceral pleura, volume of surgical intervention, volume of conducted lymph node dissection, fact of intraperitoneal delegation of lung blood vessels, conducting of adjuvant radiotherapy and chemotherapy, complaints on weakness, loss of appetite, loss of body weight, cough, shortness of breath, chest pain, hemoptysis, hyperthermia, detection of malignant lung neoplasm during prophylactic examination, Кі-67 expression (categorized index), CD31/CD34 expression, Her2-neu expression, EGFR expression, p53 expression, E-cadherin and pan-cytokeratin expression) were used.The most significant of them were selected.Such selection was performed by means of univariate analysis with retaining of variables that have significance level p < 0.1 [10].

Results
All obtained after univariate analysis variables and their characteristics are present in Table 1.
Research on collinearity (independence) was performed after ANOVA.The variables which dependent on each other were excluded.Only the most significant markers were left.Reverse selection of variables was used.First model was calculated with all possible predictive variables and then re-calculated after removing of variability with the least significant interconnection, with response variable (survival rate).The process lasted until the model contained only statistically significant variables (Table 2).
Survival rate based on these prognostic factors can be calculated for any patient with the early stage of NSCLC.
Let Clinical examination data: Considered complaints on pain in the right side of the chest, shortness of breath and cough since February 2008.At clinic were performed such examination as: abdominal ultrasound: ultrasound signs of urine acid diathesis; CT-scan of lung: in right lung superior lobe near the root of lung, peribronchially, relatively to proximal bronchus, 49 × 29 mm neoplasm was marked.Enlarged lymph nodes: paratracheal up to 10 mm, bifurcation up to 13 mm, bronchial (right side) up to 13 mm, bronchopulmonary (right side) up to 13 mm; Fiber-optic bronchoscopy: proximal bronchus with peribronchial tumor growth; Cytological conclusion No7153: moderately differentiated squamous cell carcinoma; Spirography: 1st degree respiratory failure.After examination surgery was performed: right-sided pneumonectomy, lymph node dissection D1.Histological conclusion: No 33959-70: non-keratinizing squamous cell lung carcinoma with foci of necrosis.In bronchus and lymph nodes at the operating region the malignant growth elements were not found (pT2N0M0 G2).
Figure 1 shows forecasting of survival rate in patient G.As it may be seen, survival rate median is not reached, patient's chance to five-year survival rate is more than 60%.Control period of observation-June 01, 2013, patient is alive.Period of observation is 64 months.
Example 2. Patient M., 66-year-old man.Received institutional treatment at Department of Thoracic Surgery of Zaporizhzhia Regional Clinical Oncologic Dispensary between July 14, 2010 and August 02, 2010.
Diagnosis: Peripheral cancer of right lung superior lobe.pT2N0M0.Stage IB.Clinical examination data: Considered complaints on shortness of breath and cough near 2 months ago.At clinic were performed such examinations as: X-ray chest examination: in rear segment of superior lobe of right lung marked 6 × 6, 5 cm extensive tuberous neoplasm with path to the root of lung.Also, subpleurally in the right root of the lung determined enlarged bronchial-pulmonic lymph nodes; CT-scan: in S2 of right lung viewed 40 × 40 mm peripheral neoplasm with luminous contours.Also, in S10 of lower lobe marked 21 × 26 mm neoplasm.Mediastinal lymph nodes were not enlarged; Abdominal ultrasound: abdomen pathology is not found; Fiber-optic bronchoscopy: without intraluminal pathology; After examination surgery was performed: right-side upper lobectomy with lymph node dissection D1.Histological conclusion No 22729-34: non-small cell lung carcinoma with necrosis.In bronchus and lymph nodes from operating region, the malignant growth elements were not found (pT2N0M0 G3).
Figure 2 shows forecasting of survival rate in patient M. As it may be seen, survival rate median equals 10 -11 months; patient's chance to five-year survival rate is almost equals 0%.Indeed patient's survival rate was 11 months and he died from the progression of disease.
It is noted that the most significant factors are the follows: overall clinical status of the patients, tumor size, vascular invasion, poor differentiation and presence of necrosis in the tumor, distant extrapulmonary metastases, type of conducted chemotherapy, number of white blood cells, level of CYFRA 21-1, Ca 125, Ca 15-3, and Ca 72-4 [2] [4] [7] [8] [10]- [13].In our research 39 prognostic markers of 254 patients were examined.At the same time independent prognostic factors selected by Cox-regression analysis were: size and histological type of tumor, type of surgical intervention, volume of conducted lymph node dissection.Among the molecular and genetic markers, such factors were prognostically significant: p53, factor 8, erb-b2, Cd-44, retinoblastoma recessive gene, cyclin E, cyclin D1, methylation DAPK, IL-10 expression, р21.Also genetic prognostic markers have been identified-genes Wnt3a, Erbb3, LSK, Rnd3 [2] [5] [6].We have demonstrated the prognostic role of Ki-67, EGFR and E-cadherin expression.Other factors that were analyzed showed no effect on survival rate in patients with stage I-II of NSCLC.
Several studies of foreign authors state that the usage of single marker for prediction is less accurate than the use of combination of factors [14]- [16].We have also observed that prediction through multivariate analysis was more accurate than the use of prognostic factor.It is necessary to use forecasting regression model in order to predict survival rate in patients with the early stage of NSCLC.It allows selecting patients with dismal prognosis.
The future researches of conducting adjuvant polychemotherapy in this category of patients are necessary.
Cox-regression model with tumor size, histological type of tumor, volume of surgical intervention, volume of lymph node dissection, Ki-67 expression, EGFR expression and E-cadherin expression, could be useful for forecasting of survival rate in patients with stage I-II of NSCLC.

Figure 1 .
Figure 1.Patient G. forecasting of survival rate.

Figure 2 .
Figure 2. Patient M. forecasting of survival rate.
Patient G., 71-year-old man.Received institutional treatment at Department of Thoracic Surgery of Zaporizhzhia Regional Clinical Oncologic Dispensary between August 26, 2008 and September 17, 2008.Diagnosis: Central cancer of right lung superior lobe.pT2N0M0.Stage IB.Concomitant diseases: Coronary heart disease: atherosclerotic cardiosclerosis.Circulatory inefficiency stage II.Obliterating atherosclerosis of the lower extremities.
us consider some examples: Example 1.

Table 1 .
Variables and their characteristics after univariate analysis.

Table 2 .
Statistically significant variables of forecasting model.