LMP1 expression is positively associated with metastasis of nasopharyngeal carcinoma: evidence from a meta-analysis
- 1Department of Pathology, School of Basic Medicine, the Fourth Military Medical University, Xi'an, Shanxi, China
- 2Department of Pathology, Cancer Research Institute, Guangdong Medical College, Dongguan, Guangdong, China
- 3Department of Microbiology and Immunology, School of Basic Medicine, Guangdong Medical College, Dongguan, Guangdong, China
- Correspondence to Dr Xinrong Hu, Department of Pathology, Cancer Research Institute, Guangdong Medical College, Dongguan, Guangdong 523808, China;
Contributors XH initiated and supervised the study and revised the final manuscript. YZ performed the analysis and wrote the paper. YW helped collect the published articles and contributed to the design of the study. SZ helped search the articles. All of the authors have read and approved the final paper.
- Accepted 26 July 2011
- Published Online First 29 October 2011
Objective Latent membrane protein 1 (LMP1) is an oncogene of Epstein–Barr virus (EBV). Although EBV is associated with 90% of nasopharyngeal carcinoma (NPC), the reported detectable expression rates of LMP1 determined by current techniques were from 50% to 80%. This study evaluates the association of LMP1 expression with metastasis in NPC.
Methods Data from published case–control studies on LMP1 expression and metastasis in NPC were collected up to 30 March 2011. The databases of Pubmed, Medline, Embase, Chinese National Knowledge Infrastructure and the Cochrane Library were searched with keywords ‘LMP1 or latent membrane protein 1’, ‘NPC or nasopharyngeal carcinoma’ and ‘metastasis’ to identify articles published in English or Chinese. Review manager V.5.0.24 software was applied for calculating OR and corresponding 95% CI and for statistical analysis. The combined OR and 95% CI for the cumulative metastasis rate in the LMP1 expression cases versus those in the LMP1-negative cases were estimated by using fixed-effects and random-effects models.
Results 718 cases from 12 articles, including 403 cases with LMP1 expression and 315 cases without LMP1 expression, were reviewed. The cumulative metastasis rates were 66.75% (269/403) in cases with LMP1 expression and 46.98% (148/315) in those without LMP1 expression. The combined OR for the cumulative metastasis rates were 1.98 (95% CI 1.38 to 2.837) in the fixed-effects model and 2.27 (95% CI 1.10 to 4.69) in the random-effects model.
Conclusion LMP1 expression is positively associated with metastasis in NPC, thus LMP1 detection in primary NPC might be an effective and feasible means to predict metastasis.
- cancer genetics
- cancer research
- latent membrane protein 1
- nasopharyngeal carcinoma
Nasopharyngeal carcinoma (NPC) is one of the head and neck cancers most commonly seen in southeast Asia, especially in the southern provinces of China. The yearly incidence of NPC is 25–50 per 100 000 people in southern China.1–3 Different from other head and neck cancers, NPC features Epstein–Barr virus (EBV) infection and early invasion of surrounding tissues and metastasis. More than 60% of NPC are first diagnosed with metastasis.4 Approximately 90% of primary NPC can be locally controlled by current clinical therapies such as radiotherapy and chemotherapy.5 6 However, approximately 30–40% of treated NPC still subsequently develop distant metastasis and/or local recurrence, which finally causes treatment failure.4 Therefore, assessment of the metastatic potential of NPC is clinically necessary in the prognosis and guidance of treatment.
Latent membrane protein 1 (LMP1) is a 63 KU transmembrane protein encoded by the LMP1 gene, one of the EBV latent genes.7–9 Although EBV is associated with 90% of NPC, the reported detectable expression rates of LMP1 determined by current techniques were from 50% to 80% in cases.10–13 Some researchers showed that the invasion and metastasis of NPC was promoted by LMP1 expression, and was inhibited when LMP1 expression was knocked down by the use of small hairpin RNA targeting the LMP1 gene in vitro and a mouse model.4 14 Although several clinical case studies with a small sample size provided evidence to support the laboratory-derived opinion that LMP1 is related to NPC invasion and metastasis, there are no large-scale studies or meta-analyses on this issue to substantiate the view so far. Moreover, some clinical studies argued that LMP1 is not associated with NPC invasion and metastasis.11 15 16 Therefore, it is not yet clear whether LMP1 plays a role in NPC invasion and metastasis. Naturally, the true clinical significance of LMP1 in the prognosis of NPC invasion and metastasis is still covered. As a single study with a small sample size is not powerful enough to resolve this issue, we increased the sample size by collecting the data from independent studies on LMP1 expression and metastasis in order to confirm whether LMP1 promotes NPC metastasis. With the cumulative published evidence, we hypothesise that LMP1 expression promotes NPC metastasis, and it is thus one of the negative prognostic factors of NPC.
Materials and methods
Literature search strategy for identification of studies
Databases of Pubmed, Embase, Medline, Chinese National Knowledge Infrastructure (CNKI, in Chinese) and the Cochrane library (without a language limitation) were searched, covering the period from 1 January 1980 to 30 March 2011. The terms ‘LMP1 or latent membrane protein 1’ and ‘NPC or nasopharyngeal carcinomas’ plus ‘metastasis’ were used as the search keywords, with a result of 50 articles found. Another five qualified articles on this topic were identified by a hand search of the references of retrieved papers. Studies testing the association between LMP1 expression and NPC metastasis were included if the following criteria were met: (1) The paper was published as a randomised case–control study. (2) The study reported the OR and 95% CI, or reported the results about LMP1-positive expression and metastasis that helped to calculate OR and 95% CI. LMP1-positive expression was defined as those that directly expressed a case as positive, or scored a case with LMP1-positive cells greater than 10% of the carcinoma cells.1 17 (3) The selected trials included patients with histology proved to be NPC. (4) The metastasis included local lymph node metastasis and distant metastasis. The exclusion criteria included those with a non-control study. As for overlapping data from the same researchers, we contacted the authors and discussed with them to reach an agreement about what data should be collected and what data should be taken away, or just selected the ones with the maximum number of cases. After the initial search, we reviewed the title, abstract and text of all 55 papers carefully in accordance with the criteria above, and finally selected 12 articles for further analysis.1 11 15 16 18–25
Article evaluation and data extraction
This meta-analysis study was based on the steps of consensus on the quality control requirements of meta-analysis reports that were put forward by Moher et al.26 The selected articles were put into a bibliography by EndNote X2 and were evaluated independently by two reviewers. The following information on each study was extracted and recorded from the articles: the name of the first author; year of publication; country of the first or corresponding author; author address; tumour type; ethnicity; number of observed cases and controls; statistical methods; randomness and reliability of study and publication bias. The ethnicity was categorised as African, Korean, Japanese, Tunisian and Chinese. If there was any disagreement between the two reviewers, a discussion was held to reach an agreement.
Review manager 5.0 program provided by the Cochrane Library was used to perform all the statistical analysis and all the p values were two-sided. Both fixed-effects and random-effects models were utilised to calculated OR and 95% CI for the association between LMP1 expression and metastasis. The association was determined by comparison between the metastasis rates in the LMP1 expression cases and the LMP1-negative cases. The OR and 95% CI were calculated for each study, and the combined OR and 95% CI were calculated for the 12 studies. If the OR was larger than 1, it indicated that the metastasis rate in LMP1 expression cases was higher than that in LMP1-negative cases, and vice versa. It was considered to be statistically significant when 95% CI did not include 1.
A χ2-based Q test and I2 statistics were performed to assess the statistical heterogeneity between studies. If the p value is larger than 0.10, the heterogeneity between studies is insignificant, indicating that the fixed-effects model (the Mantel–Haenszel method) can be used to calculate the combined OR. Otherwise, the random-effects model (the DerSimonian–Laird method) would be preferable. The z test was used to compare the cumulative metastasis rate in LMP1 expression cases with that in LMP1-negative cases. It was considered to be statistically significant when p<0.05. Publication bias was evaluated with Eegger's funnel plot.
Characteristics of studies
According to the inclusion and exclusion criteria, 12 articles were available for analysis. All 12 articles provided clear data for LMP1 expression and metastasis in NPC patients. Four articles were published in Chinese.18 19 23 25 Although these Chinese articles cannot be searched in full text at present in English databases such as Medline, they were included because their data were completed and the data represented the largest population that frequently had NPC in the world. Several experimental methods were used by the studies to detect LMP1 expression, including immunohistochemistry, western blotting and/or PCR. Three articles reported additional information on distant organ metastasis, in which 75–100% of the patients with LMP1 expression had distant organ metastasis.16 19 23
The characteristics of the cases are summarised in table 1. There were 718 cases, among which 417 cases had metastasis and 301 cases were without metastasis, or 403 cases were with LMP1 expression and 315 cases were without LMP1 expression. The cumulative metastasis rate and LMP1 expression rate of NPC were 58.08% (417/718) and 56.13% (403/718), respectively. The LMP1 expression cases had a cumulative metastasis rate of 66.75% (269/403) that was higher than 46.98% (148/315) in LMP1-negative cases.
Among the 718 cases, 44 were African Tunisian, 59 were Korean, 80 were Japanese and 535 were Chinese. The Chinese patients accounted for 74.51% of the total cases. The age distribution for the patients at diagnosis was from 10 to 82 years, and the mean age was 52 years.
OR for association of LMP1 expression with metastasis in NPC
The OR and 95% CI for the association of LMP1 expression with metastasis of NPC calculated in this study are provided in figure 1 (with the fixed-effects model) and figure 2 (with the random-effects model). In the individual study, nine of the 12 articles gave an OR greater than 1, and the remaining three articles gave an OR less than 1. In this analysis, if the OR was greater than 1, there was a greater tendency for metastasis to occur in NPC with LMP1 expression; whereas when the OR was less than 1, there was a lower tendency for metastasis to occur in NPC with LMP1 expression. Six of the nine OR greater than 1 gave 95% CI that did not include 1 (statistically significant), while all three OR less than 1 gave 95% CI that included 1 (statistically insignificant).
The combined OR (95% CI) for the association of LMP1 expression with metastasis in NPC were 1.98 (1.38 to 2.83) calculated by the fixed-effects model and 2.27 (1.10 to 4.69) calculated by the random-effects model, suggesting that NPC with LMP1 expression had a 1.98 or 2.27 times greater odds of metastasis occurring than NPC without LMP1 expression. The two combined OR were statistically significant because their 95% CI did not include 1.
Test of heterogeneity
The heterogeneity accompanying the fixed-effects model between studies was tested. As seen in figure 3, the χ2 test was 34.05 with 11 d.f., p<0.10 and I2=68%. The test demonstrated a significant heterogeneity that accompanied the fixed-effects model between studies. So the random-effects model would be preferable to account for the extra between-study variation by giving a relatively larger weight to the less precise studies.
Sensitivity and stability analysis
One study was omitted to investigate the influence of a single study on the overall analysis. The omission of any study made no significant difference, indicating that the result was statistically reliable. Moreover, the result from the fixed-effects model was statistically similar to the result from the random-effects model, suggesting the stability and sensitivity of this analysis.
Test of publication bias
Eegger's funnel plots were used to assess the possible publication biases. Figure 4 displayed a slightly asymmetrical funnel plot, suggesting that there was slight publication bias in this meta-analysis.
As a result of the sample size limitation for the individual studies, broad agreement on the association between LMP1 expression and metastasis in NPC has not yet been reached. Some studies reported that LMP1 expression promotes metastasis,4 14 and others reported that LMP1 does not affect metastasis.15 16 To overcome the sample size limitation, we performed a meta-analysis of the association of LMP1 expression with the risk of NPC metastasis based on a systematic review of 12 case–control studies including 718 cases in this study. Both the fixed-effects and random-effects models provided evidence that the expression of LMP1 is positively associated with an increased risk of NPC metastasis. The cumulative OR were 1.98 on the fixed-effects model and 2.27 on the random-effects model, with statistically significant 95% CI of 1.38 to 2.83 and 1.10 to 4.69, respectively. As the heterogeneity between studies was statistically significant (p<0.10), the result from the random-effects model would be preferable to account for the extra between-studies variation by giving a relatively larger weight to the less precise studies, which indicated that the probability of metastasis in LMP1 expression NPC is 2.27 times larger than that in LMP1-negative NPC. This finding clearly suggests that LMP1 expression promotes NPC metastasis. To the best of our knowledge, this is the first meta-analysis so far on the association between LMP1 expression and metastasis in NPC with a large sample size.
Concerning LMP1-positive expression, there were several scored methods in the quoted articles. LMP1 expression was expressed as positive or negative by Liu et al15 and Chen et al20 as average percentage of LMP1-positive tumour cells by Horikawa et al21 and as 0 (0–10%), 1 (10–50%) and 2 (>50%) by Tsuji et al.1 As LMP1-positive cells greater than 10% of the carcinoma cells is the generally accepted criterion to score a LMP1-positive expression case,1 17 we defined LMP1-positive expression cases as those that were directly expressed as positive, or those with LMP1-positive cells greater than 10% of the carcinoma cells.1 17
Many studies show that EBV infection is closely associated with NPC.27 28 LMP1, the best-known oncogene among EBV latent genes, has been suspected to play an important role in the occurrence and progression of NPC, especially in invasion and metastasis.4 14 In laboratory research, LMP1 was shown to take part in the expression regulation of tumour metastasis-related genes such as E-cadherin, MMP, c-Met, VEGF, EGFR and COX-2, by enhancing several cell survival and proliferation signal pathways such as natural factor kappa B, activator protein 1, ets-1, mitogen-activated protein kinase, JAK/STAT, phosphatidylinositol 3 kinase/Akt. LMP1 could thus be involved in NPC migration, invasion and metastasis.4 14 29 In clinical studies, LMP1 was detected in distant tissues and local lymph nodes with NPC metastasis by immunohistochemistry.4 14 16 19 20 23 30 Some clinical studies reported that approximately 75–100% of NPC with LMP1 expression occurs with metastasis in the late course of the disease.16 19 23 30 However, other clinical studies have shown that there was no clear correlation between the expression of LMP1 and NPC invasion, metastasis and long-term prognosis,11 15 16 which confused the theory about the association between LMP1 expression and metastasis in NPC. Therefore, further multicentre studies with large samples have long been expected in this field. The results in this meta-analysis with the largest sample size so far clearly support that LMP1 expression is positively associated with NPC metastasis. It surely provides a strong clue for laboratory research to study LMP1 further as a solid oncogene of NPC and for the establishment of a clinically aware system of NPC invasion and metastasis. It might be worth starting a clinical trial in which NPC patients with LMP1 expression will be given a stronger therapy than the routine treatment, such as the combination of radiotherapy and chemotherapy, expansion of the radio exposure scope and expansion of the local lymph node dissection etc. After the treatment, a more careful and longer follow-up of the patients with LMP1 expression should also be considered.
The omission of any single study made no significant difference in the overall meta-analysis, indicating that the result was statistically reliable. However, there may be a small risk of publication bias according to the funnel plot assessment, in which the slight asymmetry of the funnel plot was displayed. The publication bias may come from papers published not in English and not in Chinese. In this study, papers published in English or Chinese were all included, but papers published in other language were not included. Chinese patients with NPC included those living in and outside China. As Chinese patients accounted for 74.51% (535/718) of the total case numbers reported in English or Chinese, the case numbers from non-English and non-Chinese papers at a lower risk of NPC would be small, which may create a slight publication bias at the most. Statistically, the slight publication bias can be considered as the heterogeneity between studies. The impact of the heterogeneity can be counteracted by using the random-effects model to perform the study. As the conclusion of this study is mainly based on the random-effects model analysis, the slight publication bias as shown by the Eegger's funnel plot assessment could be accepted statistically.
This meta-analysis indicated that LMP1 expression is positively associated with a significantly increased risk of metastasis in NPC. The detection of LMP1 expression at the diagnosis of NPC may be helpful to determine the tumour invasion and metastasis trend, and thus might be an effective and feasible means to provide prognosis in NPC.
Funding This work was supported in part by grant 30670806 from the National Natural Science Foundation of China.
Competing interests None to declare.
Provenance and peer review Not commissioned; externally peer reviewed.