Of patient-related factors, age as prognostic factor was evaluated using multivariate analysis with conflicting results. While two Radiation Therapy Oncology Group (RTOG) data base evaluations (Komaki et al. 1998; Werner-Wasik et al. 2000a, b) showed that age > 70 years is an adverse prognostic factor for survival, some studies (Socinski et al. 2004; Basaki et al. 2006) did not find any influence, including one of our own studies which used 60 years as cut-off point. Contrary to this our first analysis (Jeremic et al. 1995) showed that age < 60 years (vs. > 60 years.) carried better prognosis, which was confirmed in our later study (Jeremic et al. 1996). Interestingly, subsequent analysis of RTOG 9410 (Langer et al. 2001) showed better survival for elderly, and subgroup analysis of a large meta-analysis (Auperin et al. 2006) confirmed this finding. In the absence of proven influence on increased treatment-related toxicity and given the current lack of any information suggesting that tumours may have different biological characteristics and clinical course in elderly versus non-elderly, age cannot be considered as important prognosticator in this disease. Investigation of gender as prognostic factor also brought conflicting results. While CALGB (Socinski et al. 2004) and RTOG analyses (Komaki et al. 1998; Werner-Wasik et al. 2000a, b) did not find any influence, another RTOG analysis which focused on patients with supraclavicular node metastasis treated with combined modality therapy showed that female gender carried improved prognosis (Machtay et al. 1999). This was also a consistent finding in our own studies (Jeremic et al. 1995, 1996). Again, this prognostic factor can be considered a weak one, especially in the absence of any explanation why this disease could be easier to cure in females. Performance status seems to be the most frequently investigated prognostic factor. While initial observations of its importance were made more than 30 years ago (Edmonson et al. 1976; Stanley 1980), subsequent studies regardless of the stage of the disease or treatment used, generally confirmed these observations. In particular, in two major RTOG studies using recursive partitioning analysis (RPA), performance status was always the strongest prognosticator which then was used to develop classes (Komaki et al. 1998; Werner-Wasik et al. 2000a, b). In the largest analysis so far, performance status confirmed its leading role in predicting survival of patients treated with RT and/or CHT in 1999 patients enrolled in several studies (Werner-Wasik et al. 2000a, b). Data from CALGB (Socinski et al. 2004) also showed its strong independent influence on survival, as did studies by Firat et al. (2002a, b) and Jeremic et al. (1995, 1996). Compared to other parameters, PS is likely the most important determinant of outcome (Pfister et al. 2004), frequently reflected in the fact that clinical trials designed in the last decades usually use it to stratify patients treated with different approaches, including RT-CHT. Another factor, which has frequently been evaluated, is weight loss. Different cut-off values were used such as 5 or 8 %. While Stanley (1980) identified weight loss as an important prognostic factor in 5000 patients with lung cancer, some contemporary studies failed to observe this correlation (Socinski et al. 2004; Ball et al. 2002). The majority, however, did so (Komaki et al. 1998; Werner-Wasik et al. 2000a, b; Pfister et al. 2004), including our previous studies (Jeremic et al. 1995, 1996).

Among tumour-related factors, stage was frequently investigated. Similarly to our previous findings (Jeremic et al. 1995, 1996) more recent studies showed improved survival for stage IIIA patients. This can be seen as indirect confirmation of earlier observations by Stanley (1980), although a CALGB study (Socinsky et al. 2004) surprisingly showed better outcome for stage IIIB patients. As the authors correctly pointed out, even current staging systems and stage groupings are not perfect, for example regarding the dependence on size and not volume of the tumour and/or nodes in the designation process. In addition, tumour volume may actually be a more important means of investigating the potential influence of T and N components and, therefore stage. It was clearly shown that both total tumour volume or primary tumour volume (Basaki et al. 2006) can be considered as independent prognostic factors for survival. However, in that study other endpoints (local/regional PFS) were not used to prove such relationship in a causal way, as improvement in local/regional control should lead to an improved OS. The most recent staging classification of the International Association for the Study of Lung Cancer (Goldstraw et al. 2007) (Table 2) provides a new framework for further evaluation of the T, N, and M stages as well as stage grouping. While still being imperfect, it represents an ongoing effort of the community of thoracic oncologists to revise stage groupings based on a large body of data. One of the tumour-related factors that have occasionally been investigated upon its independent influence on survival is histology. While one RTOG study (Machtay et al. 1999) showed no influence of histology, similarly to the observation by Basaki et al. (2006), another RTOG study (Werner-Wasik et al. 2000a, b) clearly identified non-LC histology as the one having better prognosis. In our own experience LC carried an unfavourable prognosis while differentiation between SCC and non-SCC histology proved to be important. In a pooled RTOG studies analysis (Cox et al. 1999), histology was related to pattern of failure. ADC had greater risk for distant metastasis than SCC, although in our study (Jeremic et al. 2011) LC fared even worse. Also, it was shown by the RTOG authors that ADC progressed at the primary site more often than either SCC or LC carcinoma, which is only partially in agreement with our findings (Jeremic et al. 2011). Regardless of this, the RTOG study (Cox et al. 1999) clearly emphasized the importance of histology for treatment outcome and suggested separation of SCC from both ADC and LC carcinoma.

Of treatment-related factors, interfraction interval in hyperfractionated (Hfx) RT regimens was shown to be an independent prognosticator of treatment outcome (OS and LPFS) (Jeremic and Shibamoto 1996, Jeremic et al.2004). Contrary to our repeated findings, RTOG (Werner-Wasik et al. 1999) did not confirm any influence of interfraction interval on treatment outcome. However, this study suffered from serious shortcomings which we discussed earlier (Jeremic et al. 2004). First, definition of interfraction interval (mean of all daily intervals) was not optimal because a patient could be assigned an interval of 6 h when half of the treatments were given with 4 h and half with 8 h. Second, a number of CHT regimens were used, including concurrent and induction administration. Third, patients with stage II were included as well, and no separate analysis was provided for them. Fourth, no information on the influence of interval on endpoints other than oesophageal acute and late toxicity was provided. Fifth, only 3 pretreatment prognostic factors were correlated with 2 treatment-related factors (interval and CHT). Finally and most importantly, no information was provided on the influence of interval on local control, a vital part of this analysis, since interval might influence overall survival by influencing local control first. Concurrent CHT was proven to be an independent prognosticator of survival in several analyses. This is in contrast with findings by Ball et al. (2002) and Basaki et al. (2006), but is in agreement with the available data from RTOG and CALGB (Werner-Wasik et al. 1999; Socinsky et al. 2004) and findings from our previous studies (Jeremic et al. 1995, 1996). This has also been recently confirmed by the outcome of a meta-analysis that showed a significant advantage of concurrent RT-CHT over RT alone (Auperin et al. 2006). Only rarely published analyses included not only OS but two additional endpoints, namely LPFS and DMFS. Results of our most recent multivariate analysis (Jeremic et al. 2011) using LPFS as an endpoint mimicked those of OS, and confirmed our previous findings (Jeremic et al. 1995).

In patients not suitable for radical radiotherapy who were included in a randomised study of different palliative RT regimens, PS, weight loss and appetite loss were the most important prognostic factors for survival (Sundstrom et al. 2006). For example, 2-year survival was 22 % (no appetite loss) versus 3 % (appetite loss) in patients irradiated to 42 or 50 Gy.

Although controversial, induction therapy followed by surgery in locally advanced NSCLC has been evaluated in many studies. The investigators aimed to define the most appropriate group for this approach, i.e. patients benefitting in terms of survival. Among the patients entering a course of induction CHT or RT-CHT, not all can complete the prescribed or planned procedure. Cerfolio et al. (2008) evaluated 402 patients with biopsy-proven, non-bulky N2 disease who underwent neoadjuvant RT-CHT and surgery. Among these patients 81 % completed their neoadjuvant therapy, 50 % returned for definitive pathologic restaging and 37 % underwent thoracotomy for attempted resection. Multivariate analysis found that only age younger than 70 years, more than one lymph node involvement and response to neoadjuvant therapy remained significant predictors of moving forward to the surgical arena after neoadjuvant therapy. The 5-year survival was 8 % for the 253 patients who did not return for restaging but was 47 % for the 149 patients who underwent thoracotomy (p < 0.001). The 5-year survival for selected subgroups of patients who underwent complete resection was 42 % for the 14 patients who had unsuspected residual N2 disease, 49 % for the 65 patients who had a partial response (PR), and 53 % for the 34 patients who had a complete response (CR).

Mediastinal downstaging after induction therapy is one of the most important and evaluated factors for long term survival (Albain et al. 1995; Betticher et al. 2006; Bueno et al. 2000; De Waele et al. 2006). Patients with persisting mediastinal involvement after induction therapy have a poor prognosis and will usually not benefit from surgical resection. Therefore accurate staging of the mediastinum is of utmost importance to determine the appropriate subgroup of patients for surgical resection after induction. Either minimally invasive techniques (transthoracic fine-needle aspiration biopsy—FNAB, transbronchial needle aspiration—TBNA, endobronchial ultrasound with FNAB—EBUS, endoscopic esophageal ultrasound with FNAB—EUS) or invasive techniques (re-mediastinoscopy, video assisted surgery—VATS) can be used for this approach and have a sensitivity between 70 and 80 %. All of these approaches require experience within the field. The non-invasive restaging modalities with CT and PET have a rather low accuracy of 64 and 72 % (Annema et al. 2003; Herth et al. 2008; Kunst et al. 2007; De Waele et al. 2008, 2011; Marra et al. 2008; de Leyn et al. 2007). Betticher et al. (2003) for the Swiss Group for Clinical Cancer Research (SAKK) evaluated 90 potentially operable patients with stage IIIApN2 receiving 3 cycles of cisplatin plus docetaxel followed by surgical resection. Postoperative RT to 60 Gy was administered for positive resection margins or involvement of the uppermost mediastinal lymph node. Seventy-five patients (83 %) underwent resection. The overall response rate was 66 %, with 19 % pathological complete response (pCR). The median survival was 27.6 months and the 3-year survival was 33 %. Mediastinal nodal clearance and complete surgical resection were strong independent predictors of increased survival in a multivariate analysis.

A French group aimed to evaluate whether pCR in early stage NSCLC after platinum-based neoadjuvant CHT resulted in improved outcome, where pCR was defined by the absence of viable cancer cells in the resected surgical specimen. Among the 492 patients analyzed, 8.3 % achieved pCR and in the pCR group, 5-year overall survival was 80 % compared with 55.8 % in the non-pCR group (p = 0.0007). In multivariate analyses, pCR, SCC, weight loss less than or equal to 5 %, and stage-IB disease was found to be favorable prognostic factors of overall survival. Five-year DFS was 80.1 % in the pCR group compared to 44.8 % in the non-pCR group (p < 0.0001). SCC was the only independent predictor of pCR in the study population (Moulliet et al. 2012). All patients with pCR were clinically considered to be responders, either CR (n = 5) or PR (n = 36). These results confirm the previously reported findings, showing that cCR was predictive of pCR, and that CT assessments likely underestimate pCR rate (Milleron et al. 2005).

In a very recent retrospective analysis by Lococo et al. (2013) the long term results of the patients with pCR after induction therapy followed by surgery for locally advanced NSCLC were evaluated. Among the 195 patients treated with induction therapy, 137 were operated with radical intent. Among these, 27 % showed a pCR. Within this pCR group the overall 3- and 5-year long-term survival and disease-free survival rates were 67 and 64 %, and 68 and 71 %, respectively. Initial single N2 station involvement (p = 0.010), resection to a lesser extent than pneumonectomy (p = 0.005) and adjuvant therapy (p = 0.005) were found to be predictors of increased 5-year survival. Most of the recurrences in this cohort were distant.

One of the most important factors for the survival outcomes after surgery preceded by induction therapy is postoperative mortality. Several studies reported increased mortality after surgery with induction treatment, especially with concurrent RT-CHT, and with pneumonectomy. The Memorial Sloan Kettering Cancer Centre group updated their operative mortality and the factors associated with this from their surgical database of 549 patients. All received CHT, and 17 % also had radiation. The most common procedures were lobectomy (71 %) and pneumonectomy (13 %). Hospital mortality was 1.8 %, with only one death after right pneumonectomy. Multivariate analysis showed that predicted postoperative pulmonary function tests (predictive postoperative product, predicted postoperative diffusion capacity, and preoperative percentage of predicted postoperative diffusion capacity) were important indicators for postoperative morbidity and mortality (Barnett et al. 2011).

Perioperative mortality was assessed in a meta-analysis after neoadjuvant therapy and pneumonectomy for NSCLC. Based on 27 studies, 30- and 90-day perioperative mortalities were 7 and 12 % in the entire cohort. Cumulative mortalities were 11 and 5 % for right and left pneumonectomies. Both 30- and 90-day mortality remained greater in right than left pneumonectomy (p = 0.02). Among 11 studies providing both 30- and 90-day mortalities, mortality difference was 5 % (p < 0.0001) (Kim et al. 2012).

The predictive effect of nutritional parameters was investigated in a retrospective European study. Fifty-one patients with locally advanced NSCLC undergoing concurrent RT-CHT followed by surgery were evaluated. Postoperative complications occurred in 49 %. Weight loss ≥5 % during induction period was associated with shorter OS (p = 0.03), and especially overweight patients experiencing weight loss ≥5 % during the induction period had shorter OS and also PFS (van der Meij et al. 2011).