Prognostic factors in stage III non-small cell lung cancer: a review of conventional, metabolic and new biological variables

Abstract

Lung cancer is one of the most frequently occurring neoplasms and usually has a poor prognosis because most of the patients present with advanced or metastatic disease at the time of diagnosis. Numerous prognostic factors (PFs) have been studied, but the two most prominent, having both prognostic and operational values, are disease stage and performance status. Even if the literature on PFs in lung cancer is impressive, the number of publications specifically dealing with PFs in stage III non-small cell lung cancer (NSCLC) is limited. We reviewed the literature on this topic and separated the available information into three groups: conventional PFs, metabolic criteria (standardized uptake value [SUV] measured on¹⁸F-FDG-PET) and new biomarkers. Performance status and the distinction between stage IIIA and IIIB confirmed their prognostic value in stage III NSCLC. Other conventional PFs have been suggested such as age, weight loss, response to treatment and some characteristics describing the locoregional extension of the tumour. There is a place for the SUV as a PF for survival in early NSCLC, but its role in stage III NSCLC has to be further assessed. Some new biomarkers involved in cell cycle regulation or in apoptosis have been shown to have potential value. Their role needs to be confirmed in large prospective studies including conventional PFs to determine their independent value as a PF in stage III NSCLC. In conclusion, few PFs have been well evaluated in stage III NSCLC. New studies, taking into account the modifications derived from the 7th international staging system of the UICC, have to be performed.

Introduction

Lung cancer is one of the most frequently occurring neoplasms and is the leading cause of death by cancer worldwide [Boyle and Ferlay, 2005]. Only 10–15% of the patients will ultimately be cured. Non-small cell lung cancer (NSCLC) accounts for 85% of all cases. There is a large heterogeneity among these patients, even within a specific search leading to the need for the identification of prognostic factors (PFs). We defined a PF as a variable, assessed before any treatment, correlated to an evaluation criterion (i.e. survival), that is useful in estimating the patient’s future, independently of the treatment that will be applied. By extension, we also analyse the role of therapeutic variables such as response to treatment.

Literature on PF in lung cancer is exponentially increasing. During the years 1990–2000, the annual number of publications on this topic rose from less than 50/year to 150/year [Brundage et al. 2002]. Disease stage is currently the main PF for survival in patients with NSCLC, having both prognostic and operational values. Other PFs have been commonly reported, mainly performance status, gender, age, histology, haemoglobin level, some primary tumour characteristics such as tumour size or local extension, and mediastinal neoplastic infiltration. Some other variables have been suggested less frequently as PFs in NSCLC. They can be divided into four groups, as follows:

• routinely assessed biological variables (calcaemia, lactate dehydrogenase [LDH] and alkaline phosphatase levels, leukocytosis, neutrophil count);

• patient’s characteristics (weight loss, comorbidity, smoking status, body mass index, ethnicity);

• tumour characteristics (histological grade, number of metastatic sites, cancer-related symptoms, vascular or local invasion, malignant pleural effusion, primary tumour site);

• while not really a PF by definition, treatment parameters (adjuvant chemotherapy to surgery, type of chemotherapy, objective response to treatment).

Stage III NSCLC, for which the interest of a multimodality approach was first recognized, is diagnosed in a heterogeneous group of patients. However, there are large discrepancies in the therapeutics that can be proposed, with some patients receiving surgery as part of the treatment, others are treated with chemoradiotherapy and with a few only receiving palliative chemotherapy. There is obviously interest in finding adequate PFs in stage III NSCLC delineating more homogeneous groups of patients with similar prognosis allowing an appropriate comparison among studies and eventually selecting those patients at higher risk and in need of more intensive therapy.

Our aim was to review and summarize the available evidence about PFs in stage III NSCLC, reviewing the role of ‘conventional’ PF (anatomical factors, clinical and sociodemographic parameters, therapeutic characteristics) that have been reported in unselected NSCLC, metabolic criteria (standardized uptake value [SUV]) and biomarkers.

Methods

We performed a review of the English-, French- and Dutch-language literature assessing PFs in NSCLC using a Medline search completed by the references found in the selected publications and those known by the authors. The studies dealing specifically with or providing enough information concerning stage III NSCLC were selected, if multivariate analyses were performed. We also searched, among phase III randomised clinical trials involving stage III NSCLC, those reporting analyses for potential survival PFs. Concerning the potential prognostic role of biomarkers for survival, we searched meta-analyses published on this topic in NSCLC and studies specifically designed in stage III NSCLC.

To be homogeneous, only studies for which the authors explicitly used the 4th to 7th versions [Goldstraw et al. 2007; Mountain, 1997, 1986] of the International Staging System, or those performed during the period that these versions were used, were considered for this review. The older classification indeed grouped both locoregionally advanced and metastatic tumours into stage III.

Results

When looking at PF studies in lung cancer, we are facing a large heterogeneity. Selected populations are quite different, some studies incorporating all stage III and others dealing with specific subgroups. Diagnostic criteria, work ups and therapeutics are eminently variable. The statistical analyses that were carried out are also variable: there is no consensus about the variables to take into consideration for adjustment when looking at a new potential PF, there is no consensus on how to manage a continuous variable, there might be differences in the definition of the outcome and some studies do not make use of any multivariate techniques. In some studies, that inclusion of posttherapeutic factors such as objective response rate could have an impact on the conclusion about the role of pretherapeutic PFs. In order to reduce this heterogeneity, we separated the publications into groups of similar significance: conventional PFs, metabolic criteria and new biomarkers.

‘Conventional’ prognostic factors

By conventional PFs, we mean anatomical factors, clinical and sociodemographic parameters and some routinely assessed biological variables that are commonly used and are not investigational. We also considered therapeutic characteristics, as in various selected publications, while not exactly dealing with our definition of a PF. The main PFs in NSCLC, all stages considered, were stage, performance status, gender, age, histology and some characteristics related to primary tumour (T) or nodal mediastinal (N) extension. Other variables have been considered as potential PFs for survival. They can be grouped into biological factors (e.g. LDH, calcaemia, alkaline phosphatases, leukocytosis and neutrophil counts), patient’s characteristics (e.g. weight loss and body mass index, comorbidities, smoking behaviour and ethnicity) and tumour features (e.g. histological grade, number of metastatic sites, related symptoms, vascular or local invasion, malignant pleural effusion and primary tumour location). According to the restrictions defined previously, some treatment parameters (adjuvant chemotherapy after surgery, chemotherapy type and objective response to the treatment) have also been suggested.

A list of PFs in advanced NSCLC has been provided in a review by Brundage and colleagues [Brundage et al. 2002]. Among them, disease stage, weight loss and performance status were considered essential PFs. The prognostic role of disease stage was recently confirmed in the very large study performed by the International Association for the Study of Lung Cancer (IASLC) [Sculier et al. 2008]. Outside of the distinct prognosis of stage III from the other stages, the authors showed a statistically significant survival difference between stage IIIA and IIIB patients. Stage was further confirmed as a statistically independent PF in multivariate analysis (p < 0.001), in addition to age, gender, histology and performance status.

From our literature review, we found 39 studies meeting our eligibility criteria, published between 1997 and 2009 and presenting results of multivariate analyses for PF identification for overall survival [Albain et al. 2009; Caglar et al. 2009; Garrido et al. 2009, 2007; Kim et al. 2008, 2007; Lee et al. 2008; Nakamura et al. 2008; Sculier et al. 2008, 2004, 1999a, 1999b; Ademuyiwa et al. 2007; Riquet et al. 2007; Van Meerbeeck et al. 2007; Vokes et al. 2007, Basaki et al. 2006; Betticher et al. 2006; Stinchcombe et al. 2006; Tomita et al. 2006; Barlesi et al. 2005; Lorent et al. 2004; Niizeki et al. 2004; Socinski et al. 2004; Tanaka et al. 2004; Brooks et al. 2003; Osaki et al. 2003; Choi et al. 2002; Firat et al. 2002; Kameyama et al. 2002; Andre et al. 2000; Langendijk et al. 2000a; Furuse et al. 1999; Martins and Pereira, 1999; Suzuki et al. 1999a, 1999b, Vansteenkiste et al. 1997; Coen et al. 1995; Jeremic and Shibamoto, 1995]. The number of patients ranged from 42 to 836, with 10 studies including less than 100 patients and only 6 studies including more than 500 patients. Only 20 studies considered all stage III NSCLC [Albain et al. 2009; Caglar et al. 2009; Garrido et al. 2009, 2007; Kim et al. 2008; Nakamura et al. 2008; Ademuyiwa et al. 2007; Vokes et al. 2007; Stinchcombe et al. 2006; Basaki et al. 2006; Barlesi et al. 2005; Sculier et al. 2004, 1999a, 1999b; Socinski et al. 2004; Brooks et al. 2003; Firat et al. 2002; Kameyama et al. 2002; Langendijk et al. 2000a; Furuse et al. 1999; Martins and Pereira, 1999; Suzuki et al. 1999b; Coen et al. 1995; Jeremic and Shibamoto, 1995], others dealing with selected populations. More than 50 different variables were considered in these 39 studies, the details of which are presented in Tables 1 and 2. The most frequently reported PFs were disease stage, performance status, age and characteristics describing locoregional extent (primary tumour and/or mediastinal extension). Prognosis was better in less advanced disease, in patients with good performance status and in younger patients. In surgical series, patients presenting with a TN or a mediastinal downstaging, an objective response after induction chemotherapy or a complete tumoural resection had an improved survival. In these series, the importance of mediastinal nodal neoplastic involvement was underlined, the clinical or pathological positive N2 status being associated with poorer prognosis. Outside of stage, performance status was the principal PF in medical series while age, gender and weight loss were less frequently reported.

Table 1.

Conventional prognostic factors in stage III non-small cell lung cancer: surgical series.

Study	Year	ISS	Stage	N patients	Treatment	Significant factors in multivariate analysis
Vansteenkiste et al. [1997]	1997	4th?	IIIAN2	140	Surgery	T, PS, n N stations, histology, cN2
Suzuki et al. [1999a]	1999	5th	pN2	222	Surgery	cN, T size, n N2 stations, R0
Andre et al. [2000]	2000	5th	III	702	Surgery	pT, N2, induction CT
Choi et al. [2002]	2002	4th	pN2	101	Surgery + other	RO, CT use
Kameyama et al. [2002]	2002	5th	III	152	Surgery	Age, R0, pN
Osaki et al. [2003]	2003	5th	T4N0-2	76	Surgery	pN, R0
Lorent et al. [2004]	2004	?	IIIApN2	131	CT → surgery (or RT)	pT, mediastinal downstaging
Niizeki et al. [2004]	2004	5th	cN2-pN0	45	Surgery	Visceral pleural invasion, CEA level
Tanaka et al. [2004]	2004	5th	pIIIAN2	99	Surgery → CT or RT	cN2, n N2 stations, MCV (CD105), PI
Barlesi et al. [2005]	2005	5th	IIIAN2/IIIB	95/60	CT → surgery	IIIA: mediastinal downstaging, post-operative complication
						IIIB: vascular invasion
Betticher et al. [2006]	2006	5th	IIIApN2	75	CT → surgery	R0, pathological response, mediastinal downstaging
Tomita et al. [2006]	2006	?	pN2	78	Surgery	None
Garrido et al. [2007]	2007	?	IIIAN2/IIIB	136	CT → surgery	R0, response to induction CT, age
Kim et al. [2007]	2007	?	IIIAN2	66	CRT → surgery	ypN
Riquet et al. [2007]	2007	5th	pN2	586	Surgery	Age, pT, one N2 station, adjuvant treatment
Van Meerbeeck et al. [2007]	2007	5th	IIIAN2	332	CT → surgery or RT	Histology
Kim et al. [2008]	2008	5th	IIIAN2/dry IIIB	42	CT → surgery	Downstaging
Lee et al. [2008]	2008	5th	IIIAN2	262	Surgery +/− CRT	Age, adjuvant CT, n N2 stations
Caglar et al. [2009]	2009	?	III	144	CRT +/− surgery	Stage, surgery
Albain et al. [2009]	2009	?	IIIAN2	396	CRT +/− surgery	Weight loss, gender, nN2 stations

ISS, International Staging System; PS, performance status; R0, complete resection; CT, chemotherapy; CRT, radiochemotherapy; CEA, carcinoembryonic antigen; MCV, microvessel density; PI, proliferative index.

Table 2.

Conventional prognostic factors in stage III non-small cell lung cancer: medical series.

Study	Year	ISS	Stage	N patients	Treatment	Significant factors in multivariate analysis
Coen et al. [1995]	1995	4th	III	239	RT+/−CT	RT dose, pleural effusion
Jeremic and Shibamoto [1995]	1995	?	III	169	CRT	Age, stage, gender, PS, weight loss
Furuse et al. [1999]	1999	4th	III	320	CRT	PS, concomitant CRT
Martins and Pereira [1999]	1999	?	III	796	?	PS, weight loss, superior vena cava syndrome, surgery, CT, RT
Sculier et al. [1999b]	1999	4th	III	462	CT versus CRT	PS, age, stage, neutrophils, platelets, response to CT
Sculier et al. [1999a]	1999	4th	III	115	CT → CT or RT	PS
Suzuki et al. [1999b]	1999	5th	pI-IV	836	?	IIIA: T size
						IIIB: pN2-3
Langendijk et al. [2000a]	2000	4th	III	198	RT	PS, weight loss, N, QOL
Firat et al. [2002]	2002	4th	III	112	RT	PS, co-morbidity
Brooks et al. [2003]	2003	?	III	59	CRT	p53, hMSH2
Sculier et al. [2004]	2004	5th	III	328	CRT	PS, stage, haemoglobin, bilirubin
Socinski et al. [2004]	2004	?	Dry III	694	RT, CRT	Haemoglobin, PS, RT vs RT/CT
Stinchcombe et al. [2006]	2006	?	III	102	CRT	Tumour volume, PS, stage
Basaki et al. [2006]	2006	?	III	71	RT (+/− CT)	Tumour volume, TTV
Ademuyiwa et al. [2007]	2007	5th?	III	203	?	Age, gender, ethnicity, BMI, PS, FEV1, smoking, haemoglobin, SUV, stage
Vokes et al. [2007]	2007	5th?	III	366	CRT	PS, age, weight loss
Nakamura et al. [2008]	2008	5th	III	102	RT+/−CT	Tumour site
Sculier et al. [2008]	2008	7th	IIIA	2048	?	Age, gender, histology, PS
Garrido et al. [2009]	2009	5th	III	139	CRT	Concurrent CRT

BMI, body mass index; PS, performance status; FEV1, forced expiratory volume in 1 second; SUV, standardized uptake value; TTV, target treatment volume; RT, radiotherapy; CT, chemotherapy; CRT, radiochemotherapy; N, node; QOL, quality of life; T, tumour.

Metabolic criteria

During the last decade,¹⁸fluorodeoxyglucose positron emission tomography (¹⁸F-FDG-PET) emerged as a cornerstone in the diagnostic procedure for solitary pulmonary nodules and for the staging of lung cancer. SUV is a semiquantitative measurement of the tumour metabolic activity. In lung cancer, small-size studies have assessed the prognostic role of SUV measured by ¹⁸F-FDG-PET on the primary tumour. One meta-analysis [Berghmans et al. 2008a], updated recently [Paesmans et al. 2010], suggested that high SUV might be a poor prognostic factor for survival in patients with NSCLC. Using a random-effects model, the combined hazard ratio (HR) of 21 eligible retrospective studies was 2.08, significantly different from 1, with a 95% confidence interval (CI) ranging from 1.69 to 2.56. A subgroup analysis taking into account the disease extent was further performed. Fourteen studies gave enough information for meta-analysing nonmetastatic patients. The combined HR was 2.18 (95% CI 1.83–2.60).

Nevertheless, only two studies including stage III patients allowed data extraction [Hoang et al. 2008; Eschmann et al. 2006]. In the first study [Eschmann et al. 2006], 137 patients received induction chemoradiotherapy, eventually followed by surgery, for stage IIIA/B NSCLC. It was unclear from the publication whether some stage IV NSCLC were included into this group.¹⁸ F-FDG uptake was quantified by calculating the average standardized uptake value (SUV_avg). The prognostic value of SUV_avg was assessed dichotomizing the patients according to the so-called ‘best’ cut-off method (in fact, a minimal p-value approach without adjustment for multiplicity) found by the authors. In multivariate analysis, stage (p = 0.01), tumour grading (p = 0.04) and SUV_avg (p = 0.05) were statistically independent PFs for survival. The second study included patients with advanced NSCLC (stages IIIA/IIIB/IV) [Hoang et al. 2008]. The maximum standardized uptake value (SUV_max) was quantitatively used to determine¹⁸F-FDG PET activity, dichotomizing the patients according to its median value. Overall, SUV_max was not a PF in this population with advanced NSCLC (p = 0.35). Subgroup analyses were performed, separating the patients with dry stage III from those with wet stage IIIB (now stage IV in the 7th Union Internationale Contre le Cancer (UICC) staging system) and IV. The difference in survival of patients with dry stage III NSCLC stratified by the median SUV_max was of borderline statistical significance (p = 0.067). The HR that we calculated for the studies by Eschmann and colleagues and by Hoang and colleagues were 1.71 (95% CI 1.00–2.93) and 1.34 (95% CI 0.93–1.95), respectively. The combined HR was 1.45 (95% CI 1.07–1.97) (unpublished results). This is suggestive for a prognostic role of SUV in stage III NSCLC. These results as well as the independent value of SUV need to be confirmed in larger studies within a multivariate analysis taking into account other conventional PFs. An individual data meta-analysis is underway and could perhaps solve the question.

Prognostic role for survival of ‘new’ biomarkers in stage III NSCLC

We defined as ‘new’ biomarkers, biological factors that are not used in routine practice and that have not been eventually considered for conventional PF, such as LDH level. The literature on these new biomarkers in lung cancer is impressive. Thousands of articles have been published during the last decade concerning more than 200 different biomarkers. The methodological problems encountered with these studies are the same as with conventional PFs including small-sized, underpowered retrospective studies, lack of multivariate analyses and matchless populations (e.g. different work ups and treatment variability). In addition, we are facing the difficulties in comparing laboratory results due to the lack of standardization. The best example came from immunohistochemistry (IHC) with various laboratory protocols for the detection of the same marker and the absence of an international validated definition of tumour positivity.

We first looked at the most published biological factors for whom meta-analyses have been performed. They are summarized in Table 3. They all concerned NSCLC. For most of them, the abnormal expression of the marker at the protein level assessed by IHC or the evaluation of aberration at the gene level were found to be unfavourable factors for survival, except for TTF-1 and Bcl-2 that were found to be favourable PFs in IHC studies. However, we cannot find from these meta-analyses information specifically dedicated to stage III NSCLC, which does not allow meaningful extrapolation to this population.

Table 3.

Summary of meta-analyses assessing the prognostic role of new biomarkers in non-small cell lung cancer.

Study	Biomarker	N studies	Results	95% CI	Prognostic value
Huncharek et al. [1999]	K-RAS	8	RR 2.35 at 2 years	1.61–3.22	Unfavourable
Huncharek et al. [2000]	p53 gene	8	RR 1.52	1.07–2.16	Unfavourable
Mitsudomi et al. [2000]	p53	43	p < 0.05	–	Unfavourable
Choma et al. [2001]	Aneuploïdy	35	OR 0.67 at 4 years	0.57–0.78	Unfavourable
Steels et al. [2001]	p53	50	HR 1.44	1.20–1.72	Unfavourable
Meert et al. [2002a]	EGF-R	11	HR 1.14	0.94–1.39	Not a prognostic factor (Pejorative in subgroup)
Delmotte et al. [2002]	VEGF	18	HR 1.48	1.27–1.72	Unfavourable
Meert et al. [2002b]	MVD	7–9	HR 1.80–1.99	1.10–2.96	Unfavourable
Meert et al. [2003]	c-ERB2	20	HR 1.55	1.29–1.86	Unfavourable
Martin et al. [2003]	Bcl-2	18	HR 0.72	0.57–0.86	Favourable
Martin et al. [2004]	Ki-67	16	HR 1.56	1.30–1.87	Unfavourable
Mascaux et al. [2005]	K-RAS	28	HR 1.35	1.16–1.56	Unfavourable
Nakamura et al. [2005]	c-ERB2	20	p = 0.0001	–	Unfavourable
Nakamura et al. [2006]	EGF-R	18	HR 1.14	0.97–1.34	Not a prognostic factor
Mascaux et al. [2006]	COX2	10	HR 1.39	0.97–1.99	Not a prognostic factor (Pejorative in subgroup)
Berghmans et al. [2006b]	TTF-1	8	HR 0.64	0.41–1.00	Favourable
Trivella et al. [2007]	MVD	17	HR 1.03	0.97–1.09	Not a prognostic factor (Pejorative in subgroup)
Fan et al. [2008]	Surviving	8	RR 1.88	1.31–2.70	Unfavourable
Zhan et al. [2009]	VEGF		HR 1.46	1.38–1.54	Unfavourable

RR, relative risk; OR, odds ratio; HR, hazard ratio; CI, confidence interval; MVD, microvessel density.

Despite the large number of publications concerning the prognostic role of biomarkers on survival in patients with lung cancer, few are specifically dealing with stage III NSCLC. They are summarized in Table 4. Some series included at least some comparable biomarkers in their analyses. Langendijk and colleagues assessed the role of p53, Bcl-2, topoisomerase 2α and apoptotic index in a series of 161 patients treated by exclusive chest radiotherapy at a dose of 60 Gy [Langendijk et al. 2000b]. Only the apoptotic index was a statistically independent PF in multivariate analysis. Hwang and colleagues assessed the apoptotic index, p53, Bcl-2 and c-Myc in 68 patients receiving chest radiotherapy [Hwang et al. 2001]. In multivariate analysis, Bcl-2, apoptotic index and objective response to radiotherapy were statistically independent PFs for survival. Brooks and colleagues evaluated p53, Bcl-2, EGF-R, c-erbB2 and markers involved in tissue repair and vinorelbine activity, MDR, GST, hMLH and hMSH2 (mismatch repair [MMR] genes) [Brooks et al. 2003]. Fifty nine patients from two therapeutic studies were included. They received a concomitant chemoradiotherapy with vinorelbine or carboplatin plus vinorelbine. In multivariate analysis, high p53 and low hMSH2 expressions were significantly associated with a poorer overall and cancer-specific survival rates. Morero and colleagues searched in a small group of 33 patients treated by induction chemotherapy before surgery or radiotherapy if p53, Bcl-2, p21, p27, Ki67 or apoptotic index have a prognostic role [Morero et al. 2007]. No significant factor was found. Graziano and colleagues reported on immunohistochemistry assessment of p53, c-erbB2, NSE, chromogranin A, Leu-7 and synaptophysin [Graziano et al. 2001]. Patients with unresectable stage III NSCLC, included in three prospective Cancer And Leukemia Group B (CALGB) studies, were treated with a sequential combination of chemotherapy (cisplatin plus vinblastine) followed by chest radiotherapy with or without concomitant carboplatin administration. None of the considered factors was found to be a PF either in univariate or multivariate analyses. We assessed the prognostic role of EGF-R and p53 [Berghmans et al. 2005a, 2005b]. None of the markers was individually associated with survival. Lastly, Grossi and colleagues looked at multiple biomarkers in 87 patients with stage IIIApN2 NSCLC treated with radical surgery [Grossi et al. 2010]. Only cyclin D1 was a statistically independent PF in multivariate analysis.

Table 4.

Summary of studies assessing the prognostic role of biomarkers in stage III non-small cell lung cancer.

Study	N patients	Treatment	Investigated biomarkers	Independent biomarkers PF in multivariate analysis
Langendijk et al. [2000b]	161	RT	p53, Bcl-2, topoisomerase 2α, apoptotic index	apoptotic index
Hwang et al. [2001]	68	RT	apoptotic index, p53, Bcl-2, c-myc	apoptotic index, Bcl-2
Graziano et al. [2001]	132	CT → RT	p53, c-erbB2, NSE, chromogranin A, Leu-7, synaptophysine	None
Brooks et al. [2003]	59	CRT	p53, Bcl-2, EGF-R, c-erbB2 MDR, GST, hMLH1, hMSH2	p53, hMSH2
Berghmans et al. [2005a]	93	Various	p53	None
Morero et al. [2007]	33	CT → RT or surgery	p53, Bcl-2, p21, p27, Ki67, apoptotic index	None
Berghmans et al. [2005b]	99	Various	EGF-R	None
Grossi et al. [2010]	87	Surgery	EGF-R, c-erbB2, c-kit, COX2, surviving, Bcl-2, cyclin D1, cyclin B1, MMP-2, MMP-9	Cyclin D1
Uramoto et al. [1999]	91	Surgery	Fas	Fas (univariate)
Broermann et al. [2002]	28	CT → surgery	K-ras mutation	None
Spierings et al. [2003]	87	RT +/− CT	rhTRAIL + DR4-5	DR5
Chang et al. [2005]	102	CRT	Blood lymphocyte mutagen sensitivity assay	Mutagen test
Berghmans et al. [2006a]	108	Various	TTF-1	None
Berghmans et al. [2008b]	84	Various	Mdm-2	None
Kreuter et al. [2009]	142	Various	MVD	None

RT, radiotherapy; CT, chemotherapy; PF, prognostic factor; CRT, radiochemotherapy; MDR, multidrug resistance; GST, glutathione-s-transferase; hMLH1, human MutL homolog 1; hMSH2, human MutS homolog 2; COX2, cyclooxygenase 2; MMP, metalloproteinase; rhTRAIL, recombinant human tumour necrosis factor-related apoptosis-inducing ligand; MVD, microvessel density; PF, prognostic factor.

Other various biomarkers have been the subject of small retrospective studies. Uramoto and colleagues looked at the Fas expression among 220 resected tumours [Uramoto et al. 1999]. In a subgroup analysis of 91 patients, Fas expression was associated with prolonged survival in stage III NSCLC in univariate analysis. Broermann and colleagues assessed the prognostic role of K-ras mutation (codon 12) after induction chemotherapy on surgical samples [Broermann et al. 2002]. Forty patients were initially included of whom 28 went to surgery. A trend to reduced survival was observed in the presence of a mutation, without reaching statistical significance. Spierings and colleagues studied the role of the recombinant human tumour necrosis factor-related apoptosis-inducing ligand (rhTRAIL) and its receptors DR4 and DR5 among 87 patients treated in a randomized trial comparing radiotherapy with or without carboplatin chemotherapy [Spierings et al. 2003]. In their analysis including conventional variables (histology, N status, stage, follow-up duration, gender), the authors found an increase in the risk of death with DR5 expression. Chang and colleagues investigated the relationship between the sensitivity of peripheral blood lymphocytes to a mutagen bleomycin sensitivity assay and further survival of 102 patients treated by a combination of radiotherapy and chemotherapy [Chang et al. 2005]. High bleomycin sensitivity correlated with poor overall survival and disease-specific survival. Kreuter and colleagues studied the microvessel density (MVD) from mediastinal lymph node biopsies of 142 stage IIIA/B NSCLCs [Kreuter et al. 2009]. They found no association with survival in the whole group while high MVD was a statistically independent favourable PF for survival in the subgroup of patients with stage IIIA having a complete surgical resection of the tumour (p = 0.04). We also assessed the prognostic role of TTF-1 and Mdm2 [Berghmans et al. 2006a, 2008b] which displayed no prognostic significance.

Outside of the individual prognostic role of the biomarkers, some authors suggested that a combination of some factors could be helpful in defining groups of patients with distinct survival within stage III NSCLC. Based on a hierarchical clustering analysis of their immunostaining data, Grossi and colleagues identified two subsets of patients with better (clusters 1 and 2) and worse (clusters 3, 4 and 5) prognoses (median survival times of 51 and 10 months, p < 0.0001) [Grossi et al. 2010]. In univariate analyses, we found that patients whose tumour was positive for EGF-R and negative for TTF-1 had longer survival than the others (median survival times 450 and 346 days, p = 0.11) [Berghmans et al. 2008b]. The difference reached statistical significance in the subgroup of squamous cell carcinoma (p = 0.02). Also, the EGF-R/TTF-1 status was of statistical significance for cancer-specific survival (p = 0.04), while it was even more pronounced in squamous cell carcinomas (p = 0.004). In multivariate analysis, the EGF-R/TTF-1 status was only of borderline statistical significance (p = 0.06) for cancer-specific survival.

Discussion

Following this review, we observed that the subgroup of stage, performance status, age, gender, weight loss and some characteristics describing the tumoural locoregional extension are the most prominent prognostic factors in stage III NSCLC. We suggest that there is a potential prognostic role for new metabolic (SUV) and biological factors (biomarkers) such as p53, Bcl-2, the apoptotic index, hMSH2, cyclin D1 or the DR5 receptor of the rhTRAIL.

While the number of publications on PFs in lung cancer is impressive, studies involving specifically stage III NSCLC are, in comparison, relatively limited. The difficulties in comparing the studies are due to the heterogeneity of this literature. Studies are frequently of limited size and therefore underpowered to identify PFs with a moderate to small effect. The variables taken into consideration are fairly different from one study to another, leading to final multivariate models with few overlaps in the selected covariates preventing us to assess a true independent value of a new PF. Measures of goodness of fit or of the prognostic ability of the models are rarely provided. There are large discrepancies among the patients recruited according to the disease extent, the diagnostic and extension work up or the administered treatment that could influence further survival. To reduce this variability, we separated studies of conventional PFs into surgical and medical series. In these two groups, stage and performance status confirmed their prognostic value.

The same problems are present in the biomarkers literature. Only small-size retrospective studies are available in which the statistical power was limited. They suggested that some factors involved in the cell cycle regulation or the apoptosis could be of prognostic value. There are some explanations to the divergent results found in these publications. While multivariate analyses were performed, few studies included conventional PFs, clinical and biological variables in their model, and those included were also different among the publications [Berghmans et al. 2008b, 2006a, 2005a, 2005b; Chang et al. 2005; Spierings et al. 2003; Broermann et al. 2002; Hwang et al. 2001; Langendijk et al. 2000b]. There was no standardization in the immunostaining techniques as for the use of different antibodies for the same marker. Positivity thresholds were variable. For example, two distinct scores [Brooks et al. 2003; Graziano et al. 2001], a separation into three groups [Langendijk et al. 2000b] or a threshold at 10% positive cells were used for p53 [Morero et al. 2007; Berghmans et al. 2005a]. The distribution of histological subtypes was not homogeneous, the different proportion of adenocarcinoma potentially influencing the prevalence of markers as EGF-R or TTF-1. Treatments proposed to the patients were difficult to compare. Some groups have systematically given a cisplatin-based chemotherapy [Morero et al. 2007; Graziano et al. 2001] while exclusive radiotherapy [Hwang et al. 2001; Langendijk et al. 2000b] or non conventional chemotherapy regimen were administered by others [Brooks et al. 2003]. Lastly, never the role of the biomarkers was confirmed in a validation series, so that the independent role of these potential PF remains to be validated in further large prospective studies.

In conclusion, disease stage and performance status remain the most prominent prognostic factors in stage III NSCLC while other clinical and therapeutic variables have some prognostic value. New biological (biomarkers) and metabolic (SUV) variables are promising prognostic factors. Their definitive role needs to be confirmed in large, well-designed, prospective studies in stage III NSCLC taking into consideration the modification provided in the new international staging classification.

Funding

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Conflict of interest statement

The authors declare no conflicts of interest in preparing this article.