Expression of p16 in non-small cell lung cancer and its prognostic significance: A meta-analysis of published literatures

Lung Cancer 74 (2011) 155–163

Contents lists available at ScienceDirect

Lung Cancer journal homepage: www.elsevier.com/locate/lungcan

Review

Expression of p16 in non-small cell lung cancer and its prognostic significance: A meta-analysis of published literatures Jinlong Tong a,c,∗∗ , Xinchen Sun b,∗ , Hongyan Cheng b , Di Zhao b , Jun Ma b , Qing Zhen c , Yuandong Cao b , Huiping Zhu c , Jianling Bai d a

Medical School of Southeast University, 87 Dingjiaqiao Street, Nanjing, Jiangshu, 210009, PR China Department of Oncology, Affiliated Zhongda Hospital, Southeast University, 87 Dingjiaqiao Street, Nanjing, Jiangshu, 210009, PR China c Department of Oncology, the Second Affiliated Hospital, Southeast University, 1-1 Zhongfu Street, Nanjing, Jiangshu, 210003, PR China d Department of Epidemiology and Biostatistics, School of Public Health, Nanjing Medical University, 140 Hanzhong Street, Nanjing, Jiangshu, 210029, PR China b

a r t i c l e

i n f o

Article history: Received 28 September 2010 Received in revised form 19 February 2011 Accepted 30 April 2011 Keywords: Non-small cell lung cancer p16 Prognosis Meta-analysis Hazard ratio

a b s t r a c t The prognostic value of p16 for survival of patients with non-small cell lung cancer (NSCLC) remains controversial. we performed a meta-analysis of the literatures in order to clarify its impact. Published studies in English were identified using an electronic search in order to aggregate the available survival results. To be eligible, a study had to have dealt with p16 protein assessment in NSCLC patients on the primary site and have reported survival data according to p16 expression. Twenty trials, comprising 1995 patients, provided sufficient information for the meta-analysis. Seventeen assessed any non-small cell lung cancer subtype, three assessed adenocarcinoma only. Eight identified high p16 expression as a favourable prognostic factor and one linked it with poor prognosis, Eleven trials were not significant. The overall combined hazard ratio (HR) calculated using a random-effects model suggested that high p16 expression has a favourable impact on survival in all NSCLC [0.69, 95% CI: 0.59–0.81]; The studies were categorized according to histology, disease stage and laboratory technique. The aggregated survival data showed a poor survival prognosis in squamous cell cancer with lower p16 expression [0.34, 95% CI: 0.13–0.91]. The adenocarcinoma subgroup had an HR of 0.91 [95% CI 0.76–1.10] without statistical significance. In early stage NSCLC (I–II), the aggregated HR was 0.42 [95% CI: 0.28–0.63], showing a worse survival for NSCLC with abnormal p16 expression; Results were significant with the HR of 0.61 [95% CI: 0.45–0.82] for five studies detecting p16 by immunohistochemistry with antibody clone G175–405. In conclusion, our meta-analysis shows that the p16 expression status is an independent prognostic factor in NSCLC, and this tendency is also found in the subgroups of squamous cell lung cancer and early stage NSCLC (I–II), but not in lung adenocarcinoma. © 2011 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Lung cancer remains the most frequent and the most lethal human malignancy, accounting for 1.1 million deaths annually worldwide [1]. Despite recent developments in the therapeutic technology and concept prognosis remains poor, with fewer than 15% achieve a 5-year survival in most countries. Non-small cell lung cancer (NSCLC) comprises more than 80% of lung cancers and nearly

∗ Corresponding author at: Medicine School of Southeast University, 87 Dingjiaqiao Street, Nanjing, Jiangshu, 210009, PR China. Tel.: +86 25 83272000; fax: +86 25 83272011. ∗∗ Corresponding author at: Department of Oncology, the Second Affiliated Hospital, Southeast University, 1-1 Zhongfu Street, Nanjing, Jiangshu, 210003, PR China. Tel.: +86 25 83626483; fax: +86 25 83626060. E-mail addresses: [email protected] (J. Tong), [email protected] (X. Sun). 0169-5002/$ – see front matter © 2011 Elsevier Ireland Ltd. All rights reserved. doi:10.1016/j.lungcan.2011.04.019

75% of NSCLC patients have unresectable advanced disease with lymph nodes and/or visceral metastases at the time of diagnosis. This precludes surgical resection, which remains the most effective treatment for NSCLC. Although TNM stage is the most significant prognostic parameter in lung cancer [2–4], additional effective biomarkers are required for explaining variability of survival. The poor prognosis can be modulated by characteristics related to the patient or the tumor and prognostic factors can be used for a better understanding of the nature of the tumor or better identification of homogeneous patient’s populations with a similar outcome profile. Many types of cancers have an abnormal control of the transition from G1 to S phase of the cell cycle. CDKN2A is regarded as tumor suppressor gene because it is frequently silenced by deletion or inactivating mutation in human cancers [5–7]. The product of CDKN2A, p16 is a potent cyclin-dependent kinase inhibitor and a critical negative G1-specific regulator that halts cell-cycle progression at the G1-S phase boundary. Loss of its function can lead

156

J. Tong et al. / Lung Cancer 74 (2011) 155–163

to uncontrolled cell proliferation [5]. Over the past several decades hundreds of papers have appeared proposing a variety of molecular markers or proteins that may have prognostic significance in non-small cell lung cancer, however, the results from different labs were somewhat controversial. A large number of researches on p16 and NSCLC have not improved our ability to draw precise conclusions relevant to the clinical outcome of this cancer. Several studies have revealed that in particular subgroups of lung cancer patients, lower p16 expression indicate poor outcome. Other studies suggested a closely relationship between p16 and prognosis but fell short of statistical significance on multivariate analysis, and some trials have reported low p16 expression can prolong the survival of NSCLC patients. To clarify whether p16 is indeed a prognostic factor of non-small cell lung cancer, we conducted a meta-analysis of the published literatures in an attempt to gain insights into whether p16 could provide useful guidance in the biological understanding and clinical management of NSCLC.

2.2. Data extraction and handling To minimize the bias and to improve the reliability, two reviewers checked all potentially relevant studies independently. Data on the following characteristics were also extracted: journal, country, first author, dates on which the study was published, the disease stage, the types of chemotherapy/radiotherapy and surgery used, the type of measurements used to determine p16INK4a status, cut-off value of positive expression, the antibodies used for IHC, survival curves, the exact data of total and exposed number in case and control groups and hazard ratios and their confidence intervals. Two investigators (Di Zhao and Jun Ma) extracted data from eligible studies independently, discussed discrepancies, and reached at least 95% agreement on all criteria used, the remaining disagreements were resolved through contact with the primary investigators or by reference to a third reviewer (Qing Zhen) for arbitration. 2.3. Statistical method

2. Materials and methods 2.1. Search strategy, inclusion and exclusion criteria This meta-analysis was limited to studies that dealt with the prognostic implications of different p16 protein expression level. To be eligible for our meta-analysis, studies had to be English-language published documents dealing with histopathologically confirmed NSCLN at the time of study inclusion. PubMed and EMBASE database was searched in August 2009 for bibliographic information concerning articles about p16 expression status and survival in lung cancer patients. The search ended in May 2010. The keywords “lung cancer + p16 + prognosis”, “lung cancer + p16 + survival”, “lung cancer + p16” hit 115, 136 and 637 citations, respectively. The references of retrieved articles were hand-searched in order to identify other potentially eligible studies for inclusion in the analysis missed by the initial search. The following criteria for eligibility among studies were set before collecting articles:

(1) Expression of the p16 protein was evaluated in primary lung carcinoma tissues or metastatic tissues as opposed to sera. (2) The methods used to evaluate p16 protein expression included immunohistochemistry (IHC) or semi-quantification of p16 protein (Western blot). (3) The histologic type of the tumors was NSCLC. (4) The authors must offer the size of the sample, hazard ratios (HRs) and their 95% confidence intervals (CIs) or the information that can help infer the survival results in the papers. (5) Articles were published in English in the periodical literature. (6) Finally, when multiple articles were published by the same authors or group, the newest or most informative single article was selected.

Accordingly, the following exclusion criteria were also used:

(1) Reviews, non-original articles, and studies on lung cancer cell lines and animal models were excluded from our review. (2) Studies cannot offer sufficient data to acquire HR and its standard error (SE). (3) The design and the definition of the exposure were obviously different from those of the selected papers. (4) Studies concerning about p16 methylation and survival were excluded from our review.

For the quantitative aggregation of survival results, the impact of p16 expression on survival was measured by hazard ratio (HR) between the two survival distributions. For each trial, this HR was estimated by a method depending on the data provided in the publications. The most accurate method consisted of retrieving the estimated HR and its SE from the reported results or calculating them directly using two of the following parameters: the confidence interval for the HR, the log-rank statistic and its P-value or the O − E statistic (difference between numbers of observed and expected events). If those data were not available, the following were studied: the total number of events, the number of patients at risk in each group and the logrank statistic or its P-value, allowing calculation of an approximation of the HR estimate. Finally, if the only available data were in the form of graphical representations of the survival distributions, survival rates were extracted at specified times in order to reconstruct the HR estimate and its variance, with the assumption that the rate at which patients were censored was constant during the study follow-up [8]. The individual HR estimates were combined into an overall HR using Peto’s method, which consisted of a fixed-effects model assuming homogeneity of the individual true HRs. This assumption was tested by performing Chi-squared Q-tests for heterogeneity. A P-value greater than 0.10 for the Q-test indicates a lack of heterogeneity among studies, so the fixed-effects model was used for meta-analysis. Otherwise, the random-effects model was used [9]. By convention, an observed HR > 1 implied a worse survival for the group with positive p16 expression. This pejorative impact of p16 on survival was considered as statistically significant if the 95% CI for the overall HR did not overlap 1. Analyses were done with Revman version 4.3 review manager software, the overall pooled HR were estimated by calculating the weighted average of the study-specific log HR, with weights proportional to the inverse of the variances of the study-specific log HR estimates. For each study, the survival of the overall patients population was analyzed, when available. If survival was reported separately for particular subgroups, these results were treated in the metaanalysis of the corresponding subgroups. Study estimates, along with pooled estimates, are presented as forest plots. The effect of publication bias on the reported outcomes was assessed graphically using funnel plots and empirically using regression tests according to the method reported by Egger et al. (P < 0.05 was considered representative of statistically significant publication bias) [10]. Survival rates on the graphical representation of the survival curves were read by Engauge Digitizer version 2.5. HRs and their variations were calculated by Review Manager version 4.3. Analyses were done with SPSS version 14.0 and STATA version 9.0.

J. Tong et al. / Lung Cancer 74 (2011) 155–163

157

Fig. 1. Methodological flow chart of the systematic review.

3. Results

in three studies [20,30,31], in other 12 trials, no comments were made upon this aspect.

3.1. Study selection and characteristics 3.2. Study results The results of the search strategy for studies are summarized in Fig. 1. At the time of analysis, 22 studies reported the prognostic value for survival of p16 methylation status in lung cancer patients met the exclusion criteria. 36 studies published between 1996 and 2010 were selected [11–46]. One of the articles was excluded because an identical cohort of patients was used in other selected publications (study excluded [25]; study included [26]). Another four articles were excluded because small cell lung cancer, or breast cancer, prostate cancer were enrolled in the studies [35,40–42]. Eight studies could not provide the available survival data to calculate HR [32,34,36–38,44–46], the remaining three studies were given up for continuous variable data [39], detection p16 in blood sample [33] or with FISH [43]. Finally, 16 studies were given up shown in Table 1 [25,31–46], 20 retrospective trials were chosen for meta-analysis in Table 2 [11–24,26–31]. The total number of patients included was 1995, ranging from 38 to 219 patients per study. The main characteristics of the 20 eligible studies for aggregation are shown in Table 2. In total, seventeen dealt with NSCLC of all histological subtypes, three with adenocarcinoma only. Data related to patients treated by surgery (stages I–IV) comprised 19 of the 20 NSCLC trials, 2 of the 20 NSCLC studies were performed in early stage disease (stages I and II) [20,31], while six studies comprising less than 95 patients with distance metastasis deal with any stage of NSCLC (stages I–IV) [11,19,21,23,24,29]. Immunohistochemistry was the mostly common technique used to detect p16 expression with paraffin-embedded specimens (18 trials) and snap-frozen tissue [16], in one trial, authors used Western blot assess p16 protein [23]. Different antibodies were applied to detect p16 expression with immunohistochemistry: five used an anti-p16 mouse monoclonal antibody clone G175–405 (PharMingen, San Diego, CA, USA) [17,18,20,21,30], three used rabbit polyclonal antibody C-20 (Santa Cruz, CA, USA) [12,13,15] and monoclonal antibodies F12 (Santa Cruz Biotechnology, CA, USA), respectively [19,24,28], two used monoclonal antibody clone 16P07 (Neomarkers, Fermont, CA, USA) [27,29] and one used clone SC468 (Santa Cruz, CA, USA) [13]. In 5 of 20 studies, patients received postoperative chemotherapy/radiotherapy [13,16,18,21,26]. Patients received no chemotherapy or radiotherapy after surgical resection

In the selected 20 literatures, a significant association between p16 low expression and poor survival of patients with NSCLC was demonstrated in eight studies (60%), one studies (5%) showed longer survival, and 11 studies linking low p16 expression with poor prognosis were lack of statistical significance. Subgroups analyses related with pathological subtype or clinical stage were performed in seven studies. Multivariate analyses and univariate analyses were conducted in one, nine studies, respectively, both of them were founded in 10 studies (Table 2). Most patients underwent surgery. The included 20 published studies investigated multiple factors of prognosis in non-small cell lung cancer, such as age, sex, histology, clinical stage, p21, p53, ki-67, pRb, cyclin D1, etc. Overall, according to the positivity threshold for p16 expression defined by the authors, p16 protein was low expressed in 48.0% (957/1995) of the evaluable NSCLC patients, in 43.0% (420/977) of lung adenocarcinoma alone and in 60.1% (385/641) of lung squamous carcinoma alone. Statistical difference was not identified between the p16 negative expression rate in squamous carcinoma and adenocarcinoma (P = 0.075). Moreover, p16 negative expression was found in 45.1% (324/719), 47.0% (413/880) and 45.5% (558/1285) of patients presenting with stages I, I–II and I–III, respectively and 45.1% (324/719), 55.3% (89/161) and 42.4% (172/405) of patients presenting with stages I, II and III, respectively. No association between a lack of p16 protein expression and clinical stage was observed (P < 0.05). 3.3. Meta-analysis In 12 of the 20 studies, the useful data for calculation were obtained directly from the original articles, in other eight studies, HRs had to be extrapolated from the graphical representations of the survival distributions. HRs obtained from multivariate analyses were more accurate than HRs obtained from univariate analyses (if both were available for the studies) because intermixed factors were included in the multivariate analyses. All NSCLC included twenty trials comprising 1995 patients, the aggregation produced a statistically significant HR of 0.69 (95% CI:

158

J. Tong et al. / Lung Cancer 74 (2011) 155–163

Table 1 Studies excluded from the present meta-analysis. Reference

Year

No. of patients

Analysis results

Reasons for exclusion

Sakaguchi et al. [32] Sirera et al. [33] Osaki et al. [34] Beasley et al. [35] Nakata et al. [36] Ghazizadeh et al. [37] Mishina et al. [38] Akin et al. [39] Esposito et al. [25] Groeger et al. [40] Igarashi et al. [41] Abdulkader et al. [42] Choi et al. [44]

2004 2008 2001 2003 2006 2005 1999 2002 2004 1999 2004 2005 2009

61 67 104 180 224 38 111 95 68 131 111 205 63

OS:NS OS:NS; TTP:NS 5-year survival:NS OS:NM 5-year survival:NS 5-year survival:NS OS:NM 5-year survival:NS OS:S OS:S OS:NM OS:S OS:NM

Mohamed et al. [43] Burke et al. [45] Vonlanthen et al. [46]

2009 2005 2000

78 106 51

NM OS:NS OS:NM

Absence of eligible data p16 was evaluated in blood sample Absence of eligible data Small cell lung cancer was included Absence of eligible data Absence of eligible data Absence of eligible data Continuous variable data Absence of eligible data Small cell lung cancer was includedContinuous variable data Small cell lung cancer was included Absence of eligible data Large bowel, breast, prostate cancer was included p16 was evaluated with FISH Absence of eligible data Absence of eligible data Absence of eligible data Absence of eligible data

OS: overall survival; TTP: time to progression; S: significant; NM: not mentioned; NS: not significant.

0.59–0.81) (Fig. 2A), the result of the test for heterogeneity was significant (P = 0.02), so it was possible to go further in categorizing the trials. The subgroups analysis were performed according to histology, extent of the disease, most frequently used monoclonal antibody G175–405 for detect p16. Three trials for lung squamous carcinomas were assessable (Fig. 2B). Results were significantly in favour of p16 positivity with an HR of 0.34 (95% CI: 0.13–0.91), however, in the adenocarcinoma subgroup, five subgroup studies comprised together 344 patients, the aggregation with randomeffects model produced a HR of 0.66 (95% CI: 0.36–1.21) which lost statistically significance (Fig. 2C). To treat separately fourteen papers reporting on stages I–III patients, the heterogeneity among the studies provided a chance to aggregate a meaningful HR with random-effects model: HR 0.70 (95% CI: 0.58–0.84) (Fig. 2E). In the early stage, NSCLC (I–II), with six out of twenty trials evaluable, showed a significant HR of 0.42 (95% CI: 0.28–0.63) (Fig. 2D), Once again, the introduction of a random-effects model did not change the tendency. Finally, according to the monoclonal antibodies clones G175–405, HR for the five studies assessing p16 expression with fixed-effects model was 0.61 (95% CI: 0.45–0.82) (Fig. 2F). Publication bias statistics in all NSCLC studies, as determined by using the methods of Egger et al. [10] and Begg and Mazumdar [47] were as follows: bias coefficient 0.24 [−0.57, 1.05] (P = 0.547), which suggest an absence of publication bias in all studies. No evidence of publication bias was observed from either visualization of the funnel plot (Fig. 3) or Egger’s test (P = 0.547).

4. Discussion In this meta-analysis, we enrolled 20 eligible studies comparing the survival of NSCLC according to p16 expression level of the primary tumor, low p16 expression seems to be a poor factor for prognosis in NSCLC. Statistical significance was reached when all NSCLC patients were taken into account. Our data help to clarify the results of individual studies arguing with the hypothesis that high p16 expression is a prognostic factor for non-small cell lung cancer, in addition to improve understanding of lung cancer biology and identification of high-risk subgroup patients for whom specific therapy might be necessary. Moreover, in stratified analysis we also found similar features in lung squamous carcinoma, but did not reveal the same tendency in adenocarcinoma subset with pooled HR 0.66 (95% CI: 0.36–1.21). In NSCLC of early stage (I–II) and I–III stage, 95% confidence intervals of HR which does not overlap 1 were also statistically significant.

The biological mechanism of p16 can explain its prognostic significance in lung cancer. p16 expression is generally high in normal lung tissue and is rarely detected in the lung malignancies. Acting as a cyclin-dependent kinase inhibitor (CDKI) and one number of the INK4 family, p16INK4 blinds to CDK4 (cyclin-dependent kinases 4) and CDK6 and inhibits the activity of cyclin D kinase, arrests the cell cycle in the G1/S phase through a p53-dependent pathway [48,49]. CDK4–6/cyclin D enzyme complex is a requisite for retinoblastoma protein phosphorylation, which results in a conformational change and release of E2 F from the pRb. Increased levels of p16 lead to the formation of inhibitory CDK4–6/p16 complexes and a loss of stimulatory CDK4-6/cyclin D complexes. On the contrary, inactivation of either p16 or pRb function allows the cell to enter the S phase after only a pause at the G1 checkpoint, deletion or loss of function of p16INK4a would be equivalent to amplification of CDK4, which eventually accelerates the cell cycle progression and cell proliferation [50,51]. In addition, the simultaneous loss of p14ARF , which is a biproduct from the same locus of p16INK4a leads to Mdm2 activation, subsequently resulting in acceleration of cell cycle progression through p53 inactivation. It is speculated that the p16INK4a alteration in lung cancer is a rather early event that may accelerate tumor aggressiveness through inactivation of both the RB and p53 pathways [52] and may link to a poor disease outcome [18,31]. The estimated pooled HR (0.69, 95% CI: 0.59–0.81) of 20 trials for NSCLC was statistically significant (P = 0.02), suggesting that low p16INK4a expression is a strong predictor of poor prognosis among patients with NSCLC. This conclusion supports findings from previous studies [11–13,23,24,26,28,31]; however, those studies had limited statistical power, because they were based on relatively small samples. In histology, the immunohistochemical low expression of p16 is as a better marker of progression risk in lung squamous carcinoma than adenocarcinoma. Some reports have pointed out the trend that the lack of p16 expression was significantly more frequent in squamous carcinoma and less common in adenocarcinoma [11,13,33], we could not find such difference in our paper (P = 0.075). Some individual study believed that aberrant p16 promotor methylation would be an early event in lung cancer and may constitute a new biomarker for early detection risk assessment [52], which had some kind of relationship with our findings that p16 low expression also tended to be a significant predictor of poorer prognosis in early stage NSCLC patients. Meta-analysis based on individual data is considered by some authors as a gold standard [53]. However, some kind of potential bias cannot be completely eliminated. This review was totally limited to literatures published in English because other languages,

Table 2 Main characteristics and results of the eligible studies. Country

Year

Histology

Stage

No. of patients

Median follow-up (m)

Cut-off

Antibody

HR estimation

Result (M/U)

Kratzke et al. [11] Taga et al. [12] Volm et al. [13] Kawabuchib et al. [14]

America Japan Germany Japan

1996 1997 1998 1999

NSCLC NSCLC AD AD

I–IV I–IIIB I–III I–IIIA

100 115 58 50

49 29.5 ≥60 52.5

NM >10%e NM NM

R polyclonal C-20 C-20sc-468 R polyclonal

Survival curves Survival curves Survival curves HR + 95% CI

P(U) P(U) P(U) NS(M)

Brambilla et al. [15] France

1999

NSCLC

I–IIIB

168

NM

>10%

C-20

HR + 95% CI

N(M)

Hommura et al. [16] Geradts et al. [17] Huang et al. [18]

Japan

1999

NSCLC

I–IIIA

76

51

NM

R polyclonal

HR + 95% CI

N(M)

America Japan

1999 2000

NSCLC NSCLC

I–III I–IIIB

103 171

≥60 8.5

NM ≥10%

G175–405 G175–405

Survival curves HR + 95% CI

NS(U) NS(M)

Chen et al. [19] Jin et al. [20] Gessner et al. [21] Hirabayashi et al. [22] Gonzaálezet al. [23] Cheng et al. [24]

China Japan Germany Japan

2001 2001 2002 2002

NSCLC NSCLC NSCLC NSCLC

I–IV I–II I–IV I–III

107 106 61 108

21.7 NM NM NM

NM NM >10% NM

F-12 G175–405 G175–405 Polyclonal

Survival curves HR + 95% CI Survival curves HR + 95% CI

NS(U) NS(M) NS(U) NS(M)

Spain China

2002 2003

NSCLC NSCLC

I–IV I–IV

98 90

37.6 38.3

>1.1-fold >10%

Antip16 F12

HR + 95% CI HR + 95% CI

P(U) P(M)

Esposito et al. [26]

Italy

2005

NSCLC

I–III

105

25

NM

SC-1661

HR + 95% CI

P(M)

Yoo et al. [27]

Korea

2007

NSCLC

I–IIIA

219

38.9

>80%d

16P07

HR + 95% CI

NS(M)

Mohamedet al. [28] Japan

2007

NSCLC

IIIA

61

42

>10%

F12sc-1661

HR + 95% CI

P(M)

Myong et al. [29] Tanaka et al. [30] Zhu et al. [31]

2008 2004 2009

NSCLC AD NSCLC

I–IV I–III IB

115 57 148

43.8 NM >60

>10% >10% >13.5%b

16P07 G175–405 ZJ11

Survival curves Survival curves HR + 95% CI

NS(U) NS(U) P(U)

Korea Japan China

c

a

Co-factors

Differentiation, lymph node metastases, pleural involvement, vascular invasion Age, stage, histology, Rb, p53, cyclin D1

Differentiation, smoking, sex, age at surgery, nodal status, tumor status pRb,cyclin D1 Age, sex, histology, p27KIP1 , p53, pRb Age, sex, histology, stage, location, p53,TNM status, p16/p53 Clinicalstage, Nstatus, p53, p27 cyclin D1, Histology, stage, p21, p27, cyclin B1, VEGF, age,cyclinE, survivin, bcl-2, Ki-67 p21, p53, ki-67, pRb, cyclin D1

J. Tong et al. / Lung Cancer 74 (2011) 155–163

Reference

M/U: multivariate analysis/univariate analysis; AD: adenocarcinomas; SCC: squamous cell carcinoma; P: positive; N: negative; NS: not significant; NM: not mentioned. R polyclonal: rabbit polyclonal antibody. a Evaluation of p16 expression by Western blot. b Based on receiver operating characteristic curve analysis. c >1.1-fold densitometry of control tissues. d p16 positivity if >80% of tumor cells demonstrated either nuclear or cytoplasmic staining. e Tumours were classified as p16-positive (p16+) when the proportion of stained nuclei was >10% of all nuclei in the tumor.

159

160

J. Tong et al. / Lung Cancer 74 (2011) 155–163

Fig. 2. Forest plots of hazard ratios HRs for p16 in non-small cell lung cancer patients. (A) HR for overall survival (OS) with random effects model in total NSCLC; (B) HR with fixed effects model for lung squamous carcinoma; (C) HR with random effects model for lung adenocarcinoma; (D) HR with fixed effects model for early stage NSCLC (I–II); (E) HR with random effects model for I–III NSCLC; (F) HR with fixed effects model for trials using Ab clone G175–405.

J. Tong et al. / Lung Cancer 74 (2011) 155–163

161

Fig. 2. (Continued )

such as Japanese, Spanish, were not accessible for the investigators, this selection could favour the positive studies, as positive studies are often published in English, while negative studies tend to be reported in native languages or even not submitted for review by the journal, this phenomenon called the “file-drawer problem” [54,55]. The exclusion of negative studies may have resulted in

overestimation of the prognostic significance of p16 high expression in the current meta-analysis. Some studies due to lack of the data necessary for aggregation were excluded from the metaanalysis which also will bring out heterogeneity. Besides we did not look for unpublished trials and abstracts because our methodology required data available in full publications only. Finally another

162

J. Tong et al. / Lung Cancer 74 (2011) 155–163

Fig. 3. Funnel graph for the assessment of potential publication bias in studies of p16 expression in patients with non-small cell lung cancer. The funnel graph plots log HR against the standard error of the log HR. The result of the Egger’s test for publication bias show no statistical significant (P = 0.547).

possible source of confusion is the use of a same cohort of patients in different publications. It might be difficult to avoid the same patients being included more than once in the meta-analysis. To avoid duplicate data, we identified articles that included the same cohort of patients by reviewing inter-study similarities in the country in which the studies were done, investigators in the studies, source of patients, recruitment period, and inclusion criteria, when the same investigators reported results obtained on the same cohort of patients in several publications, only the largest series were included in this analysis. The wide heterogeneity in results among the studies could have been caused by differences in several characteristics of their designs, including population sample size, different tumor stage, race, postoperative therapy, year of publication, length of recruitment period, inclusion criteria, previous treatment, sample storage, primary antibody and dilution, cut-off value, endpoint definition, follow-up period, statistical strategy, and adjustment for cofactors. For example, a variety of antibodies were used for immunodetection of p16 across the studies: mouse monoclonal antibody G175–405 in the majority of the studies [17,18,20,21,30], C20 [12,13,15], F12 [19,24,28], sc-1661 [26,28], 16P07 [27,29] and sc468 [13] in the remaining studies. The antibody concentration is also an important factor having an influence on the staining result because of the intensity is correlated to the antibody concentration used. In the twenty studies, the antibody concentration ranges from 1:400 to 1:20, the definition of p16 protein positivity varied substantially among the studies. Quantitative criteria (i.e. the percentage of stained cells) were evaluated in twelve reports [12,15,18,21–24,27–31] whereas both quantitative and qualitative parameters (i.e. morphological nuclear features) were assessed in seven studies [12,15,18,21,24,27,30], moreover some literatures did not provide the related information at all [13]. As we all know, sometimes the immunohistochemical technique was performed without prior reaction for unmasked epitope on fixed tissue, which will result in different conclusions. Another potential source of bias is related to the method for extrapolating the HR. If the authors did not report the individual HR together with its variance, we calculated it from the survival comparison statistic and its variance whenever possible. If not, we extrapolated it from the survival curves using several time points during follow-up for reading the corresponding survival rates, assuming that censored observations were uniformly distributed. This methodology is described by Parmar et al. [8], reading the survival rates on the graphical representation of the survival

curves was performed independently by three of the authors, but this strategy does not eliminate completely inaccuracy in the extracted survival rates. Consequently, the estimated HR might be less reliable than what obtained directly from published statistics. Furthermore, we determined quite arbitrarily time intervals for reading survival rates on the curves, but we are not aware about any accepted methodology for the choice of the time intervals and their number. However, we compared our estimated HR and its statistical significance with the results published in each individual method and we did not identify any major contradiction between our results and the results available in the papers. We also performed a sensitivity analysis to ascertain the impact of eight estimated HRs on the final result of meta-analysis, the pooled HR was 0.65 (95% CI: 0.51–0.85) (figure not shown), which was identical of the former result (HR: 0.69, 95% CI: 0.59–0.81). For reasons of homogeneity, our meta-analysis was performed only with studies dealing with NSCLC histology. By consequence, the study of Groeger et al. was not included because 22 patients with SCLC were incorporated in the series [40]. To exclude technical biases, we performed subgroup analysis according to the most frequently used methods: IHC with mouse monoclonal antibody clone G175–405, the results were consistent with a favourable survival in the case of p16 high expression (Fig. 2E), making it improbable that the techniques were a source of bias. However, it is difficult to draw any precise conclusions when the studies are not conducted prospectively and when not all relevant data are available. For example, we could not obtain original data regarding the therapies each patient received, thus could not be able to conduct further stratified analysis. In view of our findings, we make the following recommendation to future investigators of this topic: conduct large prospective studies with long-term follow-up, set up a standard of the immunohistochemistry technique particularly concerning the positivity threshold, give a full description of survival events to allow future calculations, conduct multivariate regression analysis taking into account the classical well-defined prognostic factors and assess the prognostic role of p16 expression in the global population and separately in different pathological subtypes and clinical stages.

5. Conclusion To conclude, our meta-analysis suggests that low p16 expression is of poor prognostic significance for survival in patients with non-small cell lung cancer. The results were based on an aggregation of data obtained in 20 retrospective trials. We also found the same tendency in the subgroup of early stage lung cancer (I–II) and lung squamous cancer. The adverse effects of p16 low expression on prognosis, as represented by our study, may necessitate a careful and perhaps more aggressive approach than typical treatment in subgroup patients. To be a useful prognostic factor at the level of molecular biomarker and in the context of individual therapy, these results need to be further confirmed by an adequately designed prospective study with larger populations and determined by an appropriate multivariate analysis. What is more, these findings need to be interpreted with caution when they are introduced into clinical practice. The real role of p16INK4a in the tumorigenesis of lung cancer and its biologic contributions to a poorer prognosis remain unclear, more investigations should be made and the results of meta-analysis should be updated in the near future.

Conflict of interest statement The authors indicated no potential conflicts of interest.

J. Tong et al. / Lung Cancer 74 (2011) 155–163

Acknowledgements This work was supported by the grants from the National Natural Science Foundation of China (grant no. 30970792, to M.D. Xinchen Sun). References [1] Jemal A, Siegel R, Ward E, Hao YP, Xu JQ, et al. Cancer statistics, 2009. CA Cancer J Clin 2009;59:225–49. [2] Naruke T, Tsuchiya R, Kondo H, Asamura H, Nakayama H. Implications of staging in lung cancer. Chest 1997;112:242S–8S. [3] Goya T, Asamura H, Yoshimura H, Kato H, Shimokata K, Tsuchiya R, et al. Prognosis of 6644 resected non-small cell lung cancers in Japan: a Japanese lung cancer registry study. Lung Cancer 2005;50:227–34. [4] Naruke T, Tsuchiya R, Kondo H, Asamura H. Prognosis and survival after resection for bronchogenic carcinoma based on the 1997 TNM-staging classification: the Japanese experience. Ann Thorac Surg 2001;71:1759–64. [5] Serrano M, Hannon GJ, Beach D. A new regulatory motif in cell-cycle control causing specific inhibition of cyclin D/CDK4. Nature 1993;366(6456):704–7. [6] Kamb A, Gruis NA, Weaver-Feldhaus J, Liu Q, Harshman K, Tavtigian SV. A cell cycle regulator potentially involved in genesis of many tumor types. Science 1994;264:436–40. [7] Nobori T, Miura K, Wu DJ, Lois A, Takabayashi K, Carson DA. Deletions of the cyclin-dependent kinase-4 inhibitor gene in multiple human cancers. Nature 1994;368:753–6. [8] Parmar MK, Torri V, Stewart L. Extracting summary statistics to perform meta-analyses of the published literature for survival endpoints. Stat Med 1998;17:2815–34. [9] Lau J, Ioannidis JP, Schmid CH. Quantitative synthesis in systematic reviews. Ann Intern Med 1997;127:820–6. [10] Egger M, Smith GD, Schneider M, Minder C. Bias in meta analysis detected by a simple, graphical test. BMJ 1977;315:629–34. [11] Kratzke RA, Greatens TM, Rubins JB, Maddaus MA, Niewoehner DE, Niehans GA, et al. Rb and p16INK4a expression in resected non-small cell lung tumors. Cancer Res 1996;56:3415–20. [12] Taga S, Osaki T, Ohgami A, Imoto H, Yoshimatsu T, Yoshino I, et al. Prognostic value of the immunohistochemical detection of p16INK4 expression in non-small cell lung carcinoma. Cancer 1997;80:389–95. [13] Volm M, Koomägi R, Mattern J. Prognostic value of p16INK4A expression in lung adenocarcinoma. Anticancer Res 1998;18(4A):2309–12. [14] Kawabuchib B, Moriyamas S, Hironaka M, Fujii M, Koike M, Moriyama H, et al. p16 inactivation in small-sized lung adenocarcinoma: its association with poor prognosis. Int J Cancer 1999;84:49–53. [15] Brambilla E, Moro D, Gazzeri S, Brambilla C. Alterations of expression of Rb, p16INK4A and cyclin D1 in non-small cell lung carcinoma and their clinical significance. J Pathol 1999;188:351–60. [16] Hommura F, Dosaka-Akita H, Kinoshita I, Mishina T, Hiroumi H, Ogura S, et al. Predictive value of expression of p16INK4A, retinoblastoma and p53 proteins for the prognosis of non-small-cell lung cancers. Br J Cancer 1999;81:696–701. [17] Geradts J, Fong KM, Zimmerman PV, Maynard R, Minna JD. Correlation of abnormal RB, p16ink4a, and p53 expression with 3p loss of heterozygosity, other genetic abnormalities, and clinical features in 103 primary non-small cell lung cancers. Clin Cancer Res 1999;5:791–800. [18] Huang CI, Taki T, Higashiyama M, Kohno N. Miyake M.p16 protein expression is associated with a poor prognosis in squamous cell carcinoma of the lung. Br J Cancer 2000;82:374–80. [19] Chen JT, Chen YC, Chen CY, Wang YC. Loss of p16 and/or pRb protein expression in NSCLC. An immunohistochemical and prognostic study. Lung Cancer 2001;31:163–70. [20] Jin ML, Inoue S, Umemura T, Moriya J, Arakawa M, Nagashima K, et al. Cyclin D1, p16 and retinoblastoma gene product expression as a predictor for prognosis in non-small cell lung cancer at stages I and II. Lung Cancer 2001;34:207–18. [21] Gessner C, Liebers U, Kuhn H, Stiehl P, Witt C, Schauer J, et al. BAX and p16INK4A are independent positive prognostic markers for advanced tumour stage of nonsmall cell lung cancer. Eur Respir J 2002;19:134–40. [22] Hirabayashi H, Ohta M, Tanaka H, Sakagushi M, Fujii Y, Miyoshi S, et al. Prognostic significance of p27KIP1 expression in resected non-small cell lung cancers: analysis in combination with expressions of p16INK4A ,pRB, and p53. J Surg Oncol 2002;81:177–84. [23] González-Quevedo R, Iniesta P, Morán A, de Juan C, Sánchez-Pernaute A, Fernández C, et al. Cooperative role of telomerase activity and p16 expression in the prognosis of non-small-cell lung cancer. J Clin Oncol 2002;20:254–62. [24] Cheng YL, Lee SC, Harn HJ, Chen CJ, Chang YC, Chen JC, et al. Prognostic prediction of the immunohistochemical expression of p53 and p16 in resected non-small cell lung cancer. Eur J Cardiothorac Surg 2003;23:221–8. [25] Esposito V, Baldi A, Tonini G, Vincenzi B, Santini M, Ambrogi V, et al. Analysis of cell cycle regulator proteins in non-small cell lung cancer. J Clin Pathol 2004;57:58–63. [26] Esposito V, Baldi A, De Luca A, Tonini G, Vincenzi B, Santini D, et al. Cell cycle related proteins as prognostic parameters in radically resected non-small cell lung cancer. J Clin Pathol 2005;58:734–9. [27] Yoo JY, Jung JH, Lee MA, Seo KJ, Shim BY, Kim SH, et al. Immunohistochemical analysis of non-small cell lung cancer: correlation with clinical parameters and prognosis. J Korean Med Sci 2007;22:318–25.

163

[28] Mohamed S, Yasufuku K, Hiroshima K, Nakajima T, Yoshida S, Suzuki M, et al. Prognostic implications of cell cycle-related proteins in primary resectable pathologic N2 nonsmall cell lung cancer. Cancer 2007;109:2506–14. [29] Myong NH. Cyclin D1 overexpression, p16 Loss, and pRb inactivation play a key role in pulmonary carcinogenesis and have a prognostic implication for the long-term survival in non-small cell Lung carcinoma patients. Cancer Res Treat 2008;40:45–52. [30] Tanaka R, Wang D, Morishita Y, Inadome Y, Minami Y, Iijima T, et al. Loss of function of p16 gene and prognosis of pulmonary adenocarcinoma. Cancer 2005;103:608–15. [31] Zhu ZH, Sun BY, Ma Y, Shao JY, Long H, Zhang X, et al. Three immunomarker support vector machines-based prognostic classifiers for stage IB non-smallcell lung cancer. J Clin Oncol 2009;27:1091–9. [32] Sakaguchi M, Fujii Y, Hirabayashi H, Yoon HE, Komoto Y, Oue T, et al. Inverse correlated expression of p16 and Rb protein in non-small cell lung cancers: an immunohistochemical study. Int J Cancer 1996;65:442–5. [33] Sirera R, Gil M, Blasco A, Cabrera A, Safont MJ, Iranzo V, et al. Retrospective analysis of the prognostic role of p16 protein inactivation in plasma in patients with locally advanced non-small cell lung cancer. Lung Cancer 2008;61:104–8. [34] Osaki T, Oyama T, Inoue M, Gu CD, Kodate M, Aikawa M, et al. Molecular biological markers and micrometastasis in resected non-small-cell lung cancer prognostic implications. Jpn J Thorac Cardiovasc Surg 2001;49:545–51. [35] Beasley MB, Lantuejoul S, Abbondanzo S, Chu WS, Hasleton PS, Travis WD, et al. The P16/cyclin D1/Rb pathway in neuroendocrine tumors of the lung. Hum Pathol 2003;32:136–42. [36] Nakata S, Sugio K, Uramoto H, Oyama T, Hanagiri T, Morita M, et al. The methylation status and protein expression of CDH1, p16(INK4A), and fragile histidine triad in nonsmall cell lung carcinoma: epigenetic silencing, clinical features, and prognostic significance. Cancer 2006;106:2190–9. [37] Ghazizadeh M, Jin E, Shimizu H, Fujiwara M, Arai S, Ohaki Y, et al. Role of cdk4, p16INK4, and Rb expression in the prognosis of bronchioloalveolar carcinomas. Respiration 2005;72:68–73. [38] Mishina T, Dosaka-Akita H, Kinoshita I, Hommura F, Morikawa T, Katoh H, et al. Cyclin D1 expression in non-small-cell lung cancers: its association with altered p53 expression, cell proliferation and clinical outcome. Br J Cancer 1999;80:1289–95. [39] Akin H, Yilmazbayhan D, Kilic¸aslan Z, Dilege S, Do˘gan O, Toker A, et al. Clinical significance of P16INK4A and retinoblastoma proteins in non-small cell lung carcinoma. Lung Cancer 2002;38:253–60. [40] Groeger AM, Caputi M, Esposito V, De Luca A, Bagella L, Pacilio C, et al. Independent prognostic role of p16 expression in lung cancer. J Thorac Cardiovasc Surg 1999;118:529–35. [41] Igarashi T, Jiang SX, Kameya T, Asamura H, Sato Y, Nagai K, et al. Divergent cyclin B1 expression and Rb/p16/cyclin D1 pathway aberrations among pulmonary neuroendocrine tumors. Mod Pathol 2004;17:1259–67. [42] Abdulkader I, Sánchez L, Cameselle-Teijeiro J, Gude F, Chávez JE, López-López R, et al. Cell-cycle-associated markers and clinical outcome in human epithelial cancers: a tissue microarray study. Oncol Rep 2005;14:1527–31. [43] Mohamed S, Yasufuku K, Nakajima T, Hiroshima K, Chiyo M, Yoshida S, et al. Nuclear survivin in pN2 nonsmall cell lung cancer: prognostic and clinical implications. Eur Respir J 2009;33:127–33. [44] Choi CM, Seo KW, Jang SJ, Oh YM, Shim TS, Kim WS, et al. Chromosomal instability is a risk factor for poor prognosis of adenocarcinoma of the lung: fluorescence in situ hybridization analysis of paraffin-embedded tissue from Korean patients. Lung Cancer 2009;64:66–70. [45] Burke L, Flieder DB, Guinee DG, Brambilla E, Freedman AN, Bennett WP, et al. Prognostic implications of molecular and immunohistochemical profiles of the Rb and p53 cell cycle regulatory pathways in primary non-small cell lung carcinoma. Clin Cancer Res 2005;11:232–41. [46] Vonlanthen S, Heighway J, Kappeler A, Altermatt HJ, Borner MM, Betticher DC. p21 is associated with cyclin D1, p16INK4a and pRb expression in resectable non-small cell lung cancer. Int J Oncol 2000;16:951–7. [47] Begg C, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088–101. [48] Cordon-Cardo C. Mutations of cell cycle regulators. Biological and clinical implications for human neoplasia. Am J Pathol 1995;147:545–60. [49] Jang SJ. Cell cycle regulators as prognostic predictor of colorectal cancers. Korean J Gastroenterol 2004;44:346–9. [50] Maelandsmo GM, Berner JM, Florenes VA, et al. Homozygous deletion frequency and expression levels of the CDKN2 gene in human sarcomas—relationship to amplification and mRNA levels of CDK4 and CCND1. Br J Cancer 1995;72: 393–8. [51] Wei G, Lonardo F, Ueda T, et al. CDK4 gene amplification in osteosarcoma: reciprocal relationship with INK4A gene alterations and mapping of 12q13 amplicons. Int J Cancer 1998;80:199–204. [52] Belinsky SA, Nikula KJ, Palmisano WA, Michels R, Saccomanno G, Gabrielson E, et al. Aberrant methylation of p16(INK4a) is an early event in lung cancer and a potential biomarker for early diagnosis. Proc Natl Acad Sci USA 1998;95:11891–6. [53] Stewart LA, Parmar MK. Meta-analysis of the literature or of individual patient data: is there a difference? Lancet 1993;341:418–22. [54] Egger M, Zellweger-Zahner T, Schneider M, Junker C, Lengeler C, Antes G. Language bias in randomized controlled trials published in English and German. Lancet 1997;350:326–9. [55] Earleywine M. The file drawer problem in the meta analysis of the subjective responses to alcohol. Am J Psychiatry 1993;150:1435–6.