Purpose: Vascular endothelial growth factor (VEGF) is considered as a prime mediator of angiogenesis. Various studies examining the relationship between VEGF protein overexpression with the clinical outcome in patients with head and neck squamous cell carcinoma have yielded inconclusive results.

Experimental Design: We conducted a meta-analysis of 12 studies (n = 1.002 patients) that evaluated the correlation between VEGF (detected by immunohistochemistry) and 2-year overall survival. The relation between VEGF and lymph node involvement (11 studies, n = 722) was also examined. Data were synthesized with random effect and fixed effect risk ratios.

Results: The estimated risk of death in 2 years was 1.88-fold higher in the VEGF-positive patients [95% confidence interval, 1.43-2.45; P < 0.001 random effect calculations]. Between-study heterogeneity was nonsignificant (P = 0.15) but larger studies tended to provide more conservative estimates (P = 0.097). VEGF overexpression was not significantly associated with the presence of lymph node metastasis (risk ratio, 1.20; 95% confidence interval, 0.97-1.49; P = 0.087) and there was significant between-study heterogeneity (P = 0.08).

Conclusions: Although some modest bias cannot be excluded, VEGF positivity seems to be associated with worse overall survival in patients with head and neck squamous cell carcinoma.

Despite advances in diagnosis and treatment, the prognosis of patients with head and neck squamous cell carcinoma (HNSCC) has remained largely unchanged (1), and the most reliable prognostic factor is still the presence of lymph node metastasis at the time of diagnosis (2). There is a need for better markers that can identify patients with poor prognosis in early stages of the disease.

Angiogenesis is considered to be a very important biological process in many neoplasms (3, 4). Angiogenesis is crucial both for the growth of a primary tumor and for the development of distant metastasis. Among the factors causing tumor angiogenesis, vascular endothelial growth factor (VEGF) is a leading candidate. VEGF induces the proliferation, differentiation, and migration of vascular endothelial cells (5), increases capillary permeability (5, 6), and enhances endothelial cell survival by preventing apoptosis (7, 8). Recently, targeted cancer therapy has spurred attention to angiogenesis inhibitors (9, 10). More than 150 antiangiogenic compounds (the majority of which target VEGF and/or its receptors) are being developed (9, 11, 12). It is therefore useful to establish whether VEGF expression is a prognostic factor.

Several studies have evaluated whether VEGF protein immunohistochemical expression may be a prognostic factor for survival in patients with HNSCC (13–23). However, the results of various studies are conflicting or inconclusive. It is unknown whether differences in these investigations have been mostly due to their limited sample size or genuine heterogeneity. Thus, we conducted a meta-analysis of all available studies relating VEGF with the clinical outcome in patients with HNSCC.

Identification and Eligibility of Relevant Studies. Initially, we identified all studies, regardless of publication language, targeting VEGF or any other angiogenesis marker in patients with HNSCC. Sources included MEDLINE and EMBASE (last search update July 2004). The search strategy was based on combinations of “angiogenesis,” “VEGF,” “vasculogenesis,” “growth factor,” “head and neck squamous cell carcinoma,” “oral cancer,” “laryngeal cancer,” and “head and neck cancer.” References of retrieved articles were also screened.

We accepted for the meta-analysis studies measuring VEGF with immunohistochemistry in patients with HNSCC, provided that measurements had been done in the primary tumor. We did not use prespecified quality-related inclusion or exclusion criteria and did not weight each study by a quality score because no such score has received general agreement for meta-analyses of observational studies (24). Whenever reports pertained to overlapping patients, we retained only the largest study to avoid duplication of information. We made an effort to contact by e-mail all primary investigators of studies of VEGF expression in HNSCC to clarify whether they had additional available patient survival data and to standardize their data according to the meta-analysis definitions whenever possible. We also contacted investigators examining other potential angiogenesis markers in HNSCC in order to obtain potentially unpublished data regarding VEGF expression and survival.

Definitions and Standardizations. We used prespecified rules to standardize as much as possible the definition of VEGF positivity. We defined VEGF positivity as positive cell stain in at least 20% of the tumor cells in continuous scales or at least moderate staining in qualitative scales. The above cutoff was used by the majority of the enrolled studies (15–22). When different definitions were used we contacted the primary investigators, and when data with this cutoff were not possible to retrieve we accepted the cutoff that was closest to this 20% cutoff level.

The main outcome of the meta-analysis was overall survival. Overall all-cause mortality avoids biases that may ensue in death attribution. This was standardized to include 24 months of follow-up in all studies to avoid some studies contributing very long-term follow-up data as compared with others. All studies had at least 24 months of follow-up and censoring was very uncommon before this time point. We also performed sensitivity analysis for VEGF expression and HNSCC-specific deaths including the studies with available information.

The secondary outcome of the meta-analysis was the presence of lymph node metastasis at the time of diagnosis. HNSCC preferentially spreads via the lymphatic vessels, and the presence of lymph node metastasis at the time of diagnosis is one of the key reasons of poor outcome (2). Lymph node metastasis was defined as the involvement of at least one lymph node. We also examined whether VEGF expression was associated with tumor location, clinical stage, and histologic differentiation.

Data Extraction. Two investigators extracted data from eligible studies independently and reached consensus to all items. We extracted data on characteristics of studies and patients, measurements, and results. In particular, in each report we recorded the first author, year of publication, country of origin, number of patients analyzed, tumor-node-metastasis (TNM) staging of HNSCC, tumor location, demographics, antibodies used for immunohistochemistry, definition of VEGF positivity, and blinding of VEGF measurements to the study outcomes. The main outcomes were tabulated in 2 × 2 tables showing the occurrence or not of death during the first 24 months of follow-up and the lymph node status according to VEGF results. We also recorded the number of patients censored alive before 24 months.

Statistical Analyses. Data on the predictive ability of VEGF overexpression for 24-month overall survival were combined across studies using fixed and random effect models for the synthesis of risk ratios (RR; ref. 25). The RR shows the death rate in 2 years in the group with VEGF overexpression divided by the death rate in 2 years in the group without VEGF overexpression. Between-study heterogeneity in the RRs was assessed with the Q statistic (considered significant for P < 0.10; ref. 25). In the same way, we estimated the relation between VEGF overexpression and lymph node metastasis and other clinicopathologic correlates.

Fixed effect models (Mantel-Haenszel) assume differences between the results of various studies are due to chance. Random effect models (DerSimonian and Laird) allow that results may differ genuinely between studies. With between-study heterogeneity, random effect models provide wider confidence intervals (26). In the text we generally present random effect estimates, unless stated otherwise.

Sensitivity analyses examined the effect of limiting the evaluation to studies using the 20% or at least moderate immunohistochemistry cutoff; to studies that clearly stated that VEGF measurements were blinded to outcomes; to HNSCC-specific deaths; to oral cavity tumors; and excluding patients censored alive before 24 months. We examined, using the inverted funnel plots, the Begg-Mazumdar correlation test, and the equivalent regression test, whether the results differed in small versus large studies (claiming significance for P < 0.10; ref. 27). We also evaluated whether results were changing gradually over time with the publication of more recent studies (28, 29). Analyses were conducted in SPSS 10.0 (SPSS, Inc., Chicago, IL) and Meta-Analyst (Joseph Lau, Boston, MA); P values are two tailed.

Eligible Studies. The initial search algorithm retrieved a total of 645 references and we evaluated 154 reports in full text. Overall, we identified 28 reports with VEGF measurements in patients with head and neck squamous cell carcinoma. Also, one investigator coauthoring the meta-analysis provided unpublished data for the correlation between VEGF expression and survival and node status. For all the patients, measurements had been done in the primary tumor, and all specimens had been taken before chemotherapy or radiotherapy, with the exception of 4 patients enrolled in the study of Neuchrist et al. (15). Of the published studies, 12 reports were excluded: one study lacked follow-up data per the response of the corresponding author (30), one overlapped with another study (31), one measured VEGF only with methods other than immunohistochemistry (32), and nine lacked informative clinical data (33–41).

Among studies that did not provide eligible data for the meta-analysis, only one had addressed survival (40), but calculations were done for VEGF levels determined by Western blot. Four studies that did not provide eligible data mentioned in brief analyses correlating VEGF with lymph node status (36, 39, 41) or TNM tumor-node-metastasis stage (38) but without offering usable data to create 2 × 2 tables. Two of them (36, 41), with a total of 92 patients, found a statistically significant association between VEGF overexpression and lymph node metastasis, whereas the other two (38, 39), with a total of 87 patients, found no significant association.

Seventeen studies (n = 1,287 patients) were eligible for the meta-analysis. Overall, for 12 studies (1,002 patients) data on 2-year survival could be obtained from published (13–16, 18) or corresponded (17, 19–23) information. Five studies (n = 285 patients) had data only for lymph node status (42–46). Six of the 12 studies with data on 2-year survival also had data on lymph node status (n = 437 patients). Therefore, 11 studies (n = 722 patients) were synthesized quantitatively for lymph node status.

Characteristics of the 17 eligible studies are listed in Table 1. Eight reports originated from Japan, seven from Europe, and one from the United States. Unpublished data were included from a Brazilian cohort. The reported median age of patients ranged from 58 to 71 years across the eligible studies. Men accounted for 60.9% of the enrolled patients across the 12 studies with gender information. All of the included studies were observational retrospective studies. The VEGF effect on mortality was apparently adjusted for tumor stage and differentiation in six studies, but only two of them provided the final adjusted RR. Laboratory procedures for VEGF determination were reported in sufficient detail in all studies. For eight reports (619 patients) data on the association of VEGF and clinical stage could be obtained from published information, whereas for 10 studies (694 patients) information for the correlation of VEGF with histologic differentiation could be extracted from the published articles. Eight of the eligible studies (all of them having data on 2-year survival) clearly stated that VEGF determinations were blinded to outcomes (13, 14, 16–18, 21) Cancer was located in the oral cavity in 911 patients (70.8%), pharynx in 196 patients (15.2%), and larynx in 180 patients (14%). Information on the prespecified cutoff (20% or at least moderate staining) could be obtained in 10 studies (15–17, 19–23, 44) Different thresholds in the 0% to 25% range were used by six reports (13, 14, 17, 42, 43, 45, 46) whereas one study (46) with data only for lymph node metastasis applied a cutoff of 80%. The most commonly used VEGF antibody was A 20 (Santa Cruz Biotechnology, Santa Cruz, CA).

Table 1

Characteristics of the eligible studies

Author (year-country)NMedian ageMale (%)Clinical staging I + II (%)LocationStaining for VEGF positivityAntibody usedBlinding of VEGF evaluation
Maeda (1998-Japan) 45 63 62 42 45 oral >5% A20 Yes 
Mineta (2002-Japan) 109 58 72 50 109 tongue >10% M293 Yes 
Kishimoto (2003-Japan) 62 NR 60 61 62 oral >20% R11 Yes 
Smith (2000-USA) 56 NR 82 32 oral, 24 or-ph At least moderate JH121 Yes 
Gallo (2002-Italy) 52 62 79 44 18 oral, 12 or-ph, 22 lar >20% JH121 Yes 
Salven (1997-Finland) 156 60 71 28 44 tongue, 22 mouth floor, 19 lower gum, 12 tonsil, 12 hypo-ph, 47 lar >0% M293 Yes 
Aebersold (2000-Switzerland) 100 NR NR 40 tonsillar fossa, 29 base of tongue, 21 faucial arch, 6 lateral posterior wall, 4 vallecula At least moderate A20 NR 
Neuchrist (1999-Austria) 52 52 98 19* 11 supraglottic lar, 41 glottic lar At least moderate M293 NR 
Uehara (2004-Japan) 63 NR NR 44 28 tongue, 3 palate, 13 lower alveolar and gingiva, 7 upper alveolar and gingiva, 8 buccal mucosa, 4 mouth floor > 20% A20 NR 
Riedel (2000-Germany) 33 58 79 12 6 oral, 10 or-ph, 8 hypo-ph, 9 lar >25% PC37 NR 
Moriyama (1997-Japan) 44 NR NR NR 23 tongue, 12 gingiva, 4 buccal mucosa, 2 mouth floor, 2 soft palate, 1 lip At least moderate A20 NR 
Artese (2001-Italy) 52 71 83 NR 25 tongue, 9 gingiva, 8 mouth floor, 3 retromolar, 2 hard palate >10% A20 NR 
Igarashi (2003-Japan) 58 62 62 NR 58 tongue >10% A20 NR 
Shintani (2004-Japan) 98 66 NR 45 98 oral >80% A20 NR 
Tanigaki (2004-Japan) 73 58 79 49 73 tongue >20% A20 NR 
Kyzas (2004-Greece) 67 65 78 63 18 tongue, 2 mouth floor, 1 palate, 1 tonsil, 9 supraglottic lar, 36 lip >20% JH121 Yes 
Cunha (2004-Brazil) 167 58 83 29 96 oral, 30 hypo-ph, 41 lar >20% A20 Yes 
Author (year-country)NMedian ageMale (%)Clinical staging I + II (%)LocationStaining for VEGF positivityAntibody usedBlinding of VEGF evaluation
Maeda (1998-Japan) 45 63 62 42 45 oral >5% A20 Yes 
Mineta (2002-Japan) 109 58 72 50 109 tongue >10% M293 Yes 
Kishimoto (2003-Japan) 62 NR 60 61 62 oral >20% R11 Yes 
Smith (2000-USA) 56 NR 82 32 oral, 24 or-ph At least moderate JH121 Yes 
Gallo (2002-Italy) 52 62 79 44 18 oral, 12 or-ph, 22 lar >20% JH121 Yes 
Salven (1997-Finland) 156 60 71 28 44 tongue, 22 mouth floor, 19 lower gum, 12 tonsil, 12 hypo-ph, 47 lar >0% M293 Yes 
Aebersold (2000-Switzerland) 100 NR NR 40 tonsillar fossa, 29 base of tongue, 21 faucial arch, 6 lateral posterior wall, 4 vallecula At least moderate A20 NR 
Neuchrist (1999-Austria) 52 52 98 19* 11 supraglottic lar, 41 glottic lar At least moderate M293 NR 
Uehara (2004-Japan) 63 NR NR 44 28 tongue, 3 palate, 13 lower alveolar and gingiva, 7 upper alveolar and gingiva, 8 buccal mucosa, 4 mouth floor > 20% A20 NR 
Riedel (2000-Germany) 33 58 79 12 6 oral, 10 or-ph, 8 hypo-ph, 9 lar >25% PC37 NR 
Moriyama (1997-Japan) 44 NR NR NR 23 tongue, 12 gingiva, 4 buccal mucosa, 2 mouth floor, 2 soft palate, 1 lip At least moderate A20 NR 
Artese (2001-Italy) 52 71 83 NR 25 tongue, 9 gingiva, 8 mouth floor, 3 retromolar, 2 hard palate >10% A20 NR 
Igarashi (2003-Japan) 58 62 62 NR 58 tongue >10% A20 NR 
Shintani (2004-Japan) 98 66 NR 45 98 oral >80% A20 NR 
Tanigaki (2004-Japan) 73 58 79 49 73 tongue >20% A20 NR 
Kyzas (2004-Greece) 67 65 78 63 18 tongue, 2 mouth floor, 1 palate, 1 tonsil, 9 supraglottic lar, 36 lip >20% JH121 Yes 
Cunha (2004-Brazil) 167 58 83 29 96 oral, 30 hypo-ph, 41 lar >20% A20 Yes 

Abbreviations: NR, not reported; or-ph, oropharyngeal; hypo-ph, hypopharyngeal; lar, laryngeal.

*

The authors do not have staging data for two patients.

Unpublished data.

Data Synthesis: Survival at 24 Months. VEGF overexpression was associated with a worse prognosis regarding the risk of death within 2 years. Mortality was 1.88-fold higher in VEGF-positive patients (P < 0.001; Fig. 1). Between-study heterogeneity was nonsignificant (P = 0.15). When patients censored alive before 2 years were excluded from the calculations, the results were similar (RR, 1.91) and there was again no significant between-study heterogeneity (P = 0.20; Table 2). In studies clearly stating that measurements were blinded to outcomes, the survival difference was clearer [RR, 2.22; Table 2] and between-study heterogeneity was still nonsignificant (P = 0.31). There was no difference in the summary estimate when the unpublished data were excluded from the analysis (RR, 1.84). Analysis for HNSCC-specific mortality also showed a significant effect (RR, 2.29; P = 0.016).

Fig. 1

Meta-analysis of the association between VEGF protein overexpression and the risk of death at 2 years. Each study is shown by the name of the lead author and the RR with 95% CIs. The summary RR and 95% CIs (according to random effect calculations) are also shown (ALL).

Fig. 1

Meta-analysis of the association between VEGF protein overexpression and the risk of death at 2 years. Each study is shown by the name of the lead author and the RR with 95% CIs. The summary RR and 95% CIs (according to random effect calculations) are also shown (ALL).

Close modal
Table 2

RR for association between VEGF and death rate in 24 months

Studies (n)RR estimates (95% CI)
Random effectsQFixed effects
TOTAL 12 (1,002) 1.88 (1.43-2.45) 15.84* 1.97 (1.59-2.43) 
Excluding patients with censoring 12 (942) 1.91 (1.48-2.45) 14.56* 2.03 (1.64-2.50) 
Specific 20% cut off 9 (692) 1.81 (1.27-2.58) 12.76* 1.92 (1.48-2.48) 
Stated blinding 8 (714) 2.22 (1.66-2.98) 8.27* 2.39 (1.84-3.11) 
Excluding unpublished data 11 (935) 1.84 (1.37-2.47) 15.26* 1.92 (1.54-2.39) 
Largest studies 3 (312) 1.56 (1.17-2.09) 1.06* 1.54 (1.15-2.07) 
HNSCC-specific death 6 (484) 2.29 (1.17-4.48) 12.75 2.25 (1.54–3.28) 
Tumors located in the oral cavity 5 (359) 2.23 (1.61–3.08) 5.60* 2.36 (1.71–3.26) 
Studies (n)RR estimates (95% CI)
Random effectsQFixed effects
TOTAL 12 (1,002) 1.88 (1.43-2.45) 15.84* 1.97 (1.59-2.43) 
Excluding patients with censoring 12 (942) 1.91 (1.48-2.45) 14.56* 2.03 (1.64-2.50) 
Specific 20% cut off 9 (692) 1.81 (1.27-2.58) 12.76* 1.92 (1.48-2.48) 
Stated blinding 8 (714) 2.22 (1.66-2.98) 8.27* 2.39 (1.84-3.11) 
Excluding unpublished data 11 (935) 1.84 (1.37-2.47) 15.26* 1.92 (1.54-2.39) 
Largest studies 3 (312) 1.56 (1.17-2.09) 1.06* 1.54 (1.15-2.07) 
HNSCC-specific death 6 (484) 2.29 (1.17-4.48) 12.75 2.25 (1.54–3.28) 
Tumors located in the oral cavity 5 (359) 2.23 (1.61–3.08) 5.60* 2.36 (1.71–3.26) 
*

P > 0.1 for between study heterogeneity.

P < 0.05 for between study heterogeneity

Larger, more precise studies tended to somewhat provide more conservative estimates compared with smaller studies (P = 0.13 for the Begg-Mazumdar test, P = 0.097 for the equivalent regression test, asymmetrical funnel plot, Fig. 2). However, with one exception, all study estimates suggested that VEGF positivity increased the risk of mortality, and the three largest studies still yielded a significant summary RR of 1.56 (P < 0.001). There was no evidence that the observed association had changed significantly over time (RR 1.55 in 1997, 1.98 in 1998, 1.49 in 1999, 1.51 in 2000, 1.78 in 2002, 1.90 in 2003, and 1.88 in 2004).

Fig. 2

Inverted funnel plot showing the relation between RR and 1 / SE. Larger studies (those with larger 1 / SE) show generally smaller RRs.

Fig. 2

Inverted funnel plot showing the relation between RR and 1 / SE. Larger studies (those with larger 1 / SE) show generally smaller RRs.

Close modal

Data Synthesis: Lymph Node Status. VEGF overexpression tended to be associated with the presence of lymph node metastasis, but the effect was modest and not formally statistically significant (RR, 1.20; P = 0.087). There was also statistically significant between-study heterogeneity (P = 0.08; Fig. 3). Formal statistical significance was claimed when analyses were limited to studies using the 20% cutoff or those that clearly stated that measurements were blinded to outcomes; however, between-study heterogeneity remained significant (Table 3).

Fig. 3

Meta-analysis of the association between VEGF protein overexpression and the presence of lymph node metastasis at diagnosis. Each study is shown by the name of the lead author and the RR with 95% CIs. The summary RR and 95% CIs (according to random effect calculations) is also shown (ALL).

Fig. 3

Meta-analysis of the association between VEGF protein overexpression and the presence of lymph node metastasis at diagnosis. Each study is shown by the name of the lead author and the RR with 95% CIs. The summary RR and 95% CIs (according to random effect calculations) is also shown (ALL).

Close modal
Table 3

RR for association between VEGF and lymph node status

Studies (n)RR estimates (95% CI)
Random effectsQFixed effects
Total 11 (722) 1.20 (0.97-1.49) 16.64* 1.30 (1.10-1.53) 
Specific 20% cut off 6 (436) 1.50 (1.04-2.17) 11.80 1.51 (1.22-1.88) 
Stated Blinding 5 (387) 1.47 (1.04-2.08) 9.52† 1.46 (1.19-1.81) 
Excluding unpublished data 10 (554) 1.21 (0.94-1.55) 16.52* 1.31 (1.09-1.58) 
Largest studies 3 (255) 1.15 (0.97–1.37) 1.32 1.17 (0.95–1.44) 
Studies (n)RR estimates (95% CI)
Random effectsQFixed effects
Total 11 (722) 1.20 (0.97-1.49) 16.64* 1.30 (1.10-1.53) 
Specific 20% cut off 6 (436) 1.50 (1.04-2.17) 11.80 1.51 (1.22-1.88) 
Stated Blinding 5 (387) 1.47 (1.04-2.08) 9.52† 1.46 (1.19-1.81) 
Excluding unpublished data 10 (554) 1.21 (0.94-1.55) 16.52* 1.31 (1.09-1.58) 
Largest studies 3 (255) 1.15 (0.97–1.37) 1.32 1.17 (0.95–1.44) 
*

0.05 < P < 0.1 for between study heterogeneity.

P < 0.05 for between study heterogeneity.

P > 0.1 for between study heterogeneity.

VEGF Association with Stage, Differentiation, and Location. We observed a trend towards a correlation of VEGF positivity with higher (III + IV) clinical stage [8 studies (n = 619); RR, 1.21; 95% confidence interval (95% CI), 1.00-1.46; P = 0.05] and poor histologic differentiation [10 studies (n = 694); RR, 1.37; 95% CI, 0.95-1.97; P = 0.095]. Tumors located in the oral cavity were not more likely to show VEGF positivity than other tumors [5 studies (n = 359); RR, 1.02; 95% CI, 0.88-1.19; P = 0.77].

This meta-analysis showed that VEGF protein overexpression, as detected with immunohistochemistry, is associated with worse overall survival in patients with HNSCC. All studies, with one exception, had results in the same direction; a prognostic effect on mortality was seen also in the three largest studies; and VEGF expression had also a significant prognostic effect on HNSCC-specific mortality. However, results should be interpreted very cautiously. VEGF might be a potential prognostic marker in HNSCC, but associations with traditional prognostic factors such as clinical stage or differentiation may explain some of its prognostic effect. Moreover, the strength of the association was significantly smaller in larger studies. This may signify publication bias against the dissemination of “negative” results (47).

Some studies have examined VEGF in HNSCC using methods other than mmunohistochemistry (reverse transcription-PCR, ELISA, or Western blot; refs. 32–34, 40). However, only one of these studies (39) had evaluated the correlation between VEGF and survival, and this study used immunohistochemistry measurements as well. Mineta et al. (40) found that high VEGF expression as determined by Western Blot in 60 patients was an independent prognostic factor for 2-year survival. There was also a significant correlation between the levels of VEGF determined by Western Blot and those determined by immunohistochemistry (P < 0.001). Although results obtained with different methods are not interchangeable, these findings are consistent with our meta-analysis.

The potential association between VEGF and lymph node status is far from certain for many reasons. The observed effect was modest and not formally significant. Between-study heterogeneity was significant. Moreover, considerable reporting bias probably exists. Also, considerable variability in definitions for VEGF positivity was documented across studies. Other members of the VEGF family such as VEGF-C and VEGF-D, but not VEGF, are considered responsible for tumor lymphangiogenesis and lymph node metastasis (48, 49). However, the capability of VEGF to increase vascular permeability both in blood and lymphatic vessels, and helping cancer cells to enter lymphatics and get established in lymph nodes, may offer a possible explanation for the observed modest association (6).

Some limitations of this meta-analysis should be discussed. First of all, as reported above, potential publication bias is a concern. We tried to identify all relevant data and retrieve additional information that was not available from the published reports, but it is unavoidable that some data could still be missing. Missing information may reflect “negative” or more conservative association of VEGF with survival (27) that could reduce the significance of VEGF expression as a predictor of mortality. Therefore, our results should be interpreted cautiously. Second, in prognostic factors meta-analyses, variability in definitions, outcomes, measurements, and experimental procedures may contribute to between-study heterogeneity (50). In the current meta-analysis, despite the fact that we tried to optimize standardization, some remaining variability in definitions was unavoidable. Although the final estimates of the synthesis of studies using the standardized cutoff did not differ significantly from the overall result in the total study population, conclusions need to be drawn cautiously (23, 50). Third, the estimates that we obtained were not adjusted for other variables such as tumor size, histologic differentiation, and clinical staging. However, we found trends for modest correlations of VEGF positivity with higher clinical stage and poor differentiation. Individual-level data would be required to assess the independent prognostic effect of each of these variables (23). Even then, it might be difficult to arrive at robust conclusions, given the correlation pattern of these prognostic factors. Finally, the available data do not allow examining whether VEGF may affect the response to specific therapeutic regimens.

VEGF is the leading candidate but not the only proposed angiogenesis factor. The prognostic value of VEGF in patients with HNSCC should be examined in the context of other proposed molecular markers such as angiogenin, interleukins 10 and 8, platelet-derived endothelial growth factor, fibroblast growth factor, angiopoetins, and thrombospodin (51). Some of the studies included in the meta-analysis had already addressed a significant association of VEGF with other key biomarkers, such as cyclooxygenase-2 (17),TP53(14, 18), and the nuclear proliferating marker ki-67 (18, 21). Also, angiogenesis may not be the only function of this protein. Recent publications show the existence of VEGF receptors in cancer cells in HNSCC (21, 52), in non–small cell lung carcinoma (53), and in cancer cell lines (54), suggesting an autocrinic role for VEGF. Considering these caveats about pathophysiology, and despite the limitations reported above, we conclude that VEGF expression might have prognostic significance for patients with HNSCC.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Gallo (Department of Oto-Neuro-Opthalmologic Surgery, University of Florence Medical School, Florence, Italy), Dr. Kishimoto (Department of Oral and Maxillofacial Surgery, Biopathological Science, Graduate School of Medicine and Dentistry, Okayama University, Japan), Dr. Aebersold (Department of Radiation Oncology, University of Bern, Inselspital, Bern, Switzerland), Dr. Uehara (Division of Oral and Maxillofacial Surgical Reconstruction and Fuctional Restoration, Department of Development and Reconstructive Medicine, Graduate School of Biomedical Sciences, Nagasaki University, Sakamoto, Nagasaki, Japan), Dr. Tanigaki (Department of Biology and Function in the Head and Neck, Yokohama City University, Graduate School of Medicine, Kanazawa-ku, Yokohama, Japan), Dr. Soares (Medical and Research Center Hospital do Cancer, Sao Paulo, Brazil), and Dr. Sauter (The Wistar Institute, Philadelphia, PA) for providing us with additional data or information from their studies.

1
Pisani P, Parkin DM, Bray F, Ferlay J. Estimates of the worldwide mortality from 25 cancers in 1990.
Int J Cancer
1999
;
83
:
8
–29.
2
Kowalski LP, Medina JE. Nodal metastases: predictive factors.
Otolaryngol Clin North Am
1998
;
31
:
621
–37.
3
Folkman J.
What is the evidence that tumors are angiogenesis dependent? J Natl Cancer Inst
1990
;
82
:
4
–6.
4
Folkman J. Angiogenesis in cancer, vascular, rheumatoid and other disease.
Nat Med
1995
;
1
:
27
–31.
5
Ferrara N, Davis-Smith T. The biology of vascular endothelial growth factor.
Endocr Rev
1997
;
18
:
4
–25.
6
Breslin JW, Pappas PJ, Cerveira JJ, Hobson RW, Duran WN. VEGF increases endothelial permeability by separate signaling pathways involving ERK-1/2 and nitric oxide.
Am J Physiol Heart Circ Physiol
2003
;
284
:
92
–100.
7
Gupta K, Kshirsagar S, Li W, et al. VEGF prevents apoptosis of human microvascular endothelial cells via opposing effects on MAPK/ERK and SAPK/JNK signalling.
Exp Cell Res
1999
;
247
:
495
–504.
8
Reinmuth N, Liu W, Jung YD, et al. Induction of VEGF in perivascular cells defines a potential paracrine mechanism for endothelial cell survival.
FASEB J
2001
;
15
:
1239
–41.
9
Underiner TL, Ruggeri B, Gingrich DE. Development of vascular endothelial growth factor receptor (VEGFR) kinase inhibitors as anti-angiogenic agents in cancer therapy.
Curr Med Chem
2004
;
11
:
731
–45.
10
Wachsberger P, Burd R, Dicker AP. Tumor response to ionizing radiation combined with antiangiogenesis or vascular targeting agents: exploring mechanisms of interaction.
Clin Cancer Res
2003
;
9
:
1957
–71.
11
Herbst RS, Hammond LA, Carbone DP, et al. A phase I/IIA trial of continuous five-day infusion of squalamine lactate (MSI-1256F) plus carboplatin and paclitaxel in patients with advanced non–small cell lung cancer.
Clin Cancer Res
2003
;
9
:
4108
–15.
12
Lara PN Jr, Quinn DI, Margolin K, et al. SU5416 plus interferon α in advanced renal cell carcinoma: a phase II California Cancer Consortium Study with biological and imaging correlates of angiogenesis inhibition.
Clin Cancer Res
2003
;
9
:
4772
–81.
13
Salven P, Heikkila P, Anttonen A, Kajanti M, Joesuu H. Vascular endothelial growth factor in squamous cell head and neck carcinoma: expression and prognostic significance.
Mod Pathol
1997
;
10
:
1128
–33.
14
Maeda T, Matsumura S, Hiramura H, et al. Expression of vascular endothelial growth factor in human oral squamous cell carcinoma: its association with tumour progression and p53 gene status.
J Clin Pathol
1998
;
51
:
771
–5.
15
Neuchrist C, Quint C, Pammer A, Burian M. Vascular endothelial growth factor (VEGF) and microvessel density in squamous cell carcinomas of the larynx: An Immunohistochemical study.
Acta Otolaryngol (Stockh)
1999
;
119
:
732
–8.
16
Smith BD, Smith GL, Carter D, Sasaki CT, Hafty BG. Prognostic significance of vascular endothelial growth factor protein levels in oral and oropharyngeal squamous cell carcinoma.
J Clin Oncol
2000
;
18
:
2046
–52.
17
Gallo O, Franchi A, Magnelli L, et al. Cyclooxygenase-2 pathway correlates with VEGF expression in head and neck cancer: implications for tumor angiogenesis and metastasis.
Neoplasia
2001
;
3
:
53
–61.
18
Mineta H, Miura K, Ogino T, Satoru T, Misawa K, Ueda Y. Vascular endothelial growth factor (VEGF) expression correlates with p53 and ki-67 expressions in tongue squamous cell carcinoma.
Anticancer Res
2002
;
22
:
1039
–44.
19
Kishimoto K, Sasaki A, Yoshihama Y, Mese H, Tsukamoto G, Matsumura T. Expression of vascular endothelial growth factor-C predicts regional lymph node metastasis in early oral squamous cell carcinoma.
Oral Oncol
2003
;
39
:
391
–6.
20
Aebersold DM, Beer KT, Laissue J, et al. Intratumoral microvessel density predicts local treatment failure of radically irradiated squamous cell cancer of the oropharynx.
Int J Radiat Oncol Biol Phys
2000
;
48
:
17
–25.
21
Kyzas PA, Stefanou D, Batistatou A, Agnantis NJ. Potential autocrine function of vascular endothelial growth factor in head and neck cancer via vascular endothelial growth factor receptor-2. Mod Pathol 2004 Oct 08 Epub ahead of print.
22
Uehara M, Sano K, Ikeda H, et al. Expression of vascular endothelial growth factor and prognosis of oral squamous cell carcinoma.
Oral Oncol
2004
;
40
:
321
–5.
23
Tanigaki Y, Nagashima Y, Kitamura Y, Matsuda H, Mikami Y, Tsukuda M. The expression of vascular endothelial growth factor-A and -C, and receptors 1 and 3: correlation with lymph node metastasis and prognosis in tongue squamous cell carcinoma.
Int J Mol Med
2004
;
14
:
389
–95.
24
Altman DG. Systematic reviews of evaluations of prognostic variables.
BMJ
2001
;
323
:
224
–8.
25
Pettiti DB. Meta-analysis, decision analysis and cost-effectiveness analysis. 2nd ed. New York, Oxford: University Press; 1999.
26
Lau J, Ioannidis JP, Schmid CH. Quantitative synthesis in systematic reviews.
Ann Intern Med
1997
;
127
:
820
–6.
27
Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias.
Biometrics
1994
;
50
:
1088
–101.
28
Ioannidis JP, Contopoulos-Ioannidis DG, Lau J. Recursive cumulative meta-analysis: a diagnostic for the evolution of total randomized evidence from group and individual patient data.
J Clin Epidemiol
1999
;
52
:
281
–91.
29
Ioannidis J, Lau J. Evolution of treatment effects over time: empirical insight from recursive cumulative meta-analyses.
Proc Natl Acad Sci U S A
2001
;
98
:
831
–6.
30
Sauter ER, Nesbit M, Watson JC, Klein-Szanto A, Litwin S, Herlyn M. Vascular endothelial growth factor is a marker of tumor invasion and metastasis in squamous cell carcinoma of the head and neck.
Clin Cancer Res
1999
;
5
:
775
–82.
31
Burian M, Quint CH, Neuchrist C.
Angiogenic factors in laryngeal carcinomas: Do they have prognostic relevance? Acta Otolaryngol (Stockh)
1999
;
119
:
289
–92.
32
O-charoenrat P, Rhys-Evans P, Eccles SA. Expression of vascular endothelial growth factor family members in head and neck squamous cell carcinoma correlates with lymph node metastasis.
Cancer
2003
;
92
:
556
–68.
33
Eisma RJ, Spiro JD, Kreutzer DL. Vascular endothelial growth factor expression in head and neck squamous cell carcinoma.
Am J Surg
1997
;
174
:
513
–7.
34
Eisma RJ, Spiro JD, Kreutzer DL. Role of angiogenic factors: coexpression of interleukin-8 and vascular endothelial growth factor in patients with head and neck squamous cell carcinoma.
Laryngoscope
1999
;
109
:
687
–93.
35
Okamoto M, Nishimine M, Kishi M, et al. Prediction of delayed neck metastasis in patients with stage I/II squamous cell carcinoma of the tongue.
J Oral Pathol Med
2002
;
31
:
227
–33.
36
Tae K, El-Naggar A, Yoo E, et al. Expression of vascular endothelial growth factor and microvessel density in head and neck tumorigenesis.
Clin Cancer Res
2000
;
6
:
2821
–8.
37
Carlile J, Harada K, Baille R, et al. Vascular endothelial growth factor (VEGF) expression in oral tissues: possible relevance to angiogenesis, tumor progression and field cancerisation.
J Oral Pathol Med
2001
;
80
:
449
–57.
38
Sawatsubashi M, Yamada T, Fukushima N, Mizokami H, Tokunaga O, Shin T. Association of vascular endothelial growth factor and mast cells with angiogenesis in laryngeal squamous cell carcinoma.
Virchows Arch
2000
;
436
:
243
–8.
39
Homer JJ, Prentice MG, Cawkwell L, Birchall M, Greemman J, Stafford ND. Angiogenesis and expression of vascular endothelial growth factors A and C in squamous cell carcinoma of the piriform fossa.
Arch Otolaryngol Head Neck Surg
2003
;
129
:
1110
–4.
40
Mineta H, Miura K, Ogino T, et al. Prognostic value of vascular endothelial growth factor (VEGF) in head and neck squamous cell carcinomas.
Br J Cancer
2000
;
83
:
775
–81.
41
Gallo O, Massini E, Bianchi B, Bruschini L, Paglierani M, Franchi A. Prognostic significance of cyclooxygenase-2 pathway and angiogenesis in head and neck squamous cell carcinoma.
Hum Pathol
2002
;
33
:
708
–14.
42
Riedel F, Gotte K, Schwalab J, Schafer C, Hormann K. Vascular endothelial growth factor correlates with p53 mutations and angiogenesis in squamous cell carcinoma of the head and neck.
Acta Otolaryngol
2000
;
120
:
105
–11.
43
Artese L, Rubini C, Ferrero G, Fiorini M, Santinelli A, Piatelli A. Microvessel density (MVD) and vascular endothelial growth factor expression (VEGF) in human oral squamous cell carcinoma.
Anticancer Res
2001
;
21
:
689
–96.
44
Moriyama M, Kumagai S, Kawashiri S, Kojima K, Kakihara K, Yamamoto E. Immunohistochemical study of tumor angiogenesis in oral squamous cell carcinoma.
Oral Oncol
1997
;
33
:
369
–74.
45
Igarashi T, Kumamoto H, Ooya K, Echigo S. Immunohistochemical analysis of vascular endothelial growth factor (VEGF) expression and tumor angiogenesis in tongue squamous carcinoma.
Oral Med Pathol
2003
;
8
:
13
–9.
46
Shintani S, Li C, Ishikawa T, Mihara M, Nakashiro K, Hamakawa H. Expression of vascular endothelial growth factor A, B, C, and D in oral squamous cell carcinoma.
Oral Oncol
2004
;
40
:
13
–20.
47
Easterbrook PJ, Berlin JA, Gopalan R, Matthews DR. Publication bias in clinical research.
Lancet
1991
;
337
:
867
–72.
48
Joukov V, Pajusola K, Kaipainen A, et al. A novel vascular endothelial growth factor, VEGF-C, is a ligand for the Flt4 (VEGFR-3) and KDR (VEGFR-2) receptor tyrosine kinases.
EMBO J
1996
;
15
:
290
–8.
49
Kyzas PA, Stefanou D, Agnantis NJ. COX-2 expression correlates with VEGF-C and lymph node metastasis in patients with head and neck squamous cell carcinoma. Mod Pathol 2004 Jul 23 Epub ahead of print.
50
Simon R, Altman DG. Statistical aspects of prognostic factor studies in oncology.
Br J Cancer
2001
;
69
:
979
–85.
51
Reinmuth N, Parikh AA, Ahmad SA, et al. Biology of angiogenesis in tumors of the gastrointestinal tract.
Microsc Res Tech
2003
;
60
:
199
–207.
52
Lalla RV, Boisoneau DS, Spiro JD, Kreutzer DL. Expression of vascular endothelial growth factor receptors on tumor cells in head and neck squamous cell carcinoma.
Arch Otolaryngol Head Neck Surg
2003
;
129
:
882
–8.
53
Galligioni E, Ferro A. Angiogenesis and antiangiogenic agents in non–small cell lung cancer.
Lung Cancer
2001
;
34
:
3
–7.
54
Masood R, Cai J, Zheng T, Smith DL, Hinton DR, Gill PS. Vascular endothelial growth factor (VEGF) is an autocrine growth factor for VEGF receptor-positive human tumors.
Blood
2001
;
98
:
1904
–13.