Microvessel Density (MVD) in Patients with Osteosarcoma: A Systematic Review and Meta-Analysis

Abstract A meta-analysis was designed and conducted to estimate the effect of tumoral microvessel density (MVD) on the survival of patients with osteosarcoma. There was no difference between high and low MVD regarding the overall (OS) and disease-free (DFS) survival. Low MVD tumors displayed a lower DFS at the third year of follow-up. Although primary metastases did not affect the mean MVD measurements, tumors with a good chemotherapy response had a higher MVD value. Although no significant differences between tumoral MVD, OS and DFS were found, good adjuvant therapy responders had a significant higher vascularization pattern.


Introduction
Osteosarcoma, despite being a rare neoplasm, still, constitutes the most common bone-derived malignancy (1,2).More specifically, the average number of new osteosarcoma cases ranges from 4.6 to 6.5/million per year (1).Furthermore, the incidence rate curve peaks in the adolescents and the elderly, while the male-to-female ratio is estimated at the level of 1.22:1 (2).
Almost 20% of the osteosarcoma cases are diagnosed at an advanced stage, while the remaining 80% is considered to have subclinical metastases (3).Consequently, the treatment modality of osteosarcoma includes the combination of local, tumor burden control and multiagent systemic chemotherapy.The standard protocol of systemic treatment consists of highdose methotrexate and leucovorin, doxorubicin and cisplatin (4).Currently, 5-year survival rates, range from 45% to 65%, for >45 and <45 years old patients, respectively (1).Moreover, longterm survival in localized osteosarcoma (70%) is higher, when compared to metastatic disease (30%) (3).
As a result, attempts to identify possible survival-modifying indicators have been implemented (5-9).Bacci et al. estimated age �14 years, high serum of alkaline phosphatase, tumor volume >200mL, two-drug chemo scheme, inadequate resection margins and poor histologic response to treatment as negative prognostic factors (10).Besides these, various other serological, molecular and genetic markers have been, also, considered (5).
Angiogenesis is a crucial part in tumoral development (11).Oncogenesis and pathobiological research have validated the role of neovascularization for tumoral growth, invasion and metastasis in osteosarcoma patients (12,13).Therefore, angiogenic compounds have been studied in both pre-clinical and clinical setting for a potential prognostic or therapeutic effect (14)(15)(16)(17)(18).
Quantification of tumoral vascularization is available through the estimation of microvessel density (MVD).Weidner et al. proposed the immunohistochemical staining of endothelial cells and the counting of the highlighted microvessels in order to assess the neovascularization potential of a malignancy (19).Currently, various immunohistochemical factors, like von Willebrand factor (vWF), cluster of differentiation (CD)31, CD34 and CD35 have been utilized for the MVD assessment.The prognostic role of tumoral MVD has been confirmed in various malignancies (20)(21)(22)(23).
Several studies have investigated a possible correlation between tumoral MVD measurements and survival endpoints in osteosarcoma patients.Kreuter et al. reported that osteosarcomas with a higher MVD value were associated with a higher 5 and 10 years overall survival rates (24).These findings, however, were not confirmed by Ek et al., who did not identify any significant variation, regarding survival outcomes, between different tumoral vascularization levels (25).Further trials examined the effect of disease endpoints, such as chemotherapy response and presence of primary metastases on MVD values (26)(27)(28).
Taking into consideration the above-mentioned facts, the present systematic literature review and meta-analysis was designed and conducted, in order to estimate the overall effect of tumor vascularity, through MVD assessment, on the survival of patients with osteosarcoma.

Study protocol
The conduction of the present meta-analysis was accomplished on the basis of the Cochrane Handbook for Systematic Reviews of Interventions and the PRISMA guidelines (29).There was no registration in any electronic database.

Primary endpoint
The primary endpoint of this meta-analysis was the pooled hazard ratio (HR) of the overall survival (OS), between high and low MVD measurements, in patients with osteosarcoma.

Secondary endpoints
Secondary endpoints included the HR of the disease-free survival (DFS), the odds ratio (OR) of the OS (1, 3 and 5 years) and the DFS (1 and 3 years) at fixed time points.Moreover, the pooled differences of the MVD measurements between tumors with or without primary metastases and poor or good chemotherapy response, were also recorded.

Eligibility criteria
All prospective or retrospective studies, with a study population of patients diagnosed with osteosarcoma, whose outcomes of interest were reported in English and were retrievable, were considered as eligible.More specifically, the MVD assessment of the primary tumor should be introduced in the study design.
Exclusion criteria for this meta-analysis were studies 1) not written in English, 2) with no outcome of interest, 3) with insufficient data, 4) without human subjects or 5) in the form of editorials, letters, conference abstracts and expert opinion.

Literature search
A systematic literature search in the electronic scholar databases (Medline, Web of Science, Scopus and CENTRAL) and Grey literature repositories (OpenGrey.euand medRxiv) was performed, to identify the eligible studies.The last search date was December 2022.
The literature screening was performed through the following search algorithm: � osteosarcoma and (MVD or microvascular density or microvessel density)

Study selection and data collection
After the removal of the duplicate entries, the titles and abstracts of the studies were screened based on the eligibility criteria.The next step included the full text review of the remaining articles.All electronic database search, study selection, data extraction and methodological assessment of the trials were performed blindly and in duplicate by two independent investigators (P.K., N.P.).In order to resolve any disagreement that occurred, mutual revision and discussion was applied.If a consensus was not reached, the opinion of a third researcher was considered (D.K.) Newcastle-Ottawa Scale (NOS) (30) was utilized for the methodological and quality assessment of the included studies.This evaluation tool assesses non-RCT trials in certain endpoints such as selection and comparability of the study groups and confirmation of the exposure.All eligible studies were rated with a score ranging from 0 to 9. Cohen's k statistic was also calculated.

Statistical analysis
The Cochrane Collaboration RevMan version 5.3 and the IBM SPSS version 25 were utilized for the data analysis and the statistical computations.The results of the analyses were provided with the corresponding 95% Confidence Intervals (95%CI).
In case that the included trials did not provide the HR or the OR directly, then these endpoints were estimated based on the algorithm provided by Parmar et al. (31) and Tierney et al. (32).More specifically, the required study results were estimated through reconstruction from the provided Kaplan-Meier (KM) curves (33).In order to increase the accuracy of the extracted coordinates, a digitizing software (Digitizelt) was used (34).Furthermore, if the trials did not provide the mean and the standard deviation (SD) of the continuous variables, then they were estimated from the median and the respective median, based on the formula by Hozo et al. (35).
The applied statistical method was the Mantel-Haenszel (MH) and the inverse variance (IV) for OR and HR, respectively.Both fixed effects (FE) and random effects (RE) models were calculated.The model that was finally used was based on the Cochran Q test.In case of a statistically significant heterogeneity (Q test P < 0.1), then the RE model was estimated.Heterogeneity was also quantified through the calculation of I 2 .Statistical significance was considered at the level of p < 0.05.

Risk of bias across studies
The funnel plot of the primary outcome was visually inspected for possible outliers.An Egger's test was also performed for the determination of the possible presence of publication bias.

Study selection
Electronic databases screening resulted in the identification of 224 records (Figure 1).After the removal of 116 duplicates, 108 titles and abstracts were submitted to the first phase of the screening.During this step, 64 studies (2 reviews or meta-analyses, 17 non-human studies and 45 irrelevant records) were excluded.In the second phase of the literature search, 44 remaining documents were submitted to full-text review.In total, 37 trials were removed due to inconsistency with the eligibility criteria (13 studies with no survival data, 1 non-Human trial, 4 reviews or meta-analyses and 19 irrelevant records).As a result, 7 studies (24)(25)(26)(27)(28)36,37) were included in the qualitative and quantitative synthesis of the present meta-analysis.

Study characteristics
In Table 2 (Supplementary Material), the characteristics of the tumors are displayed in detail.The most frequent tumor site was the femur, followed by the tibia.Furthermore, most of the malignancies were primary tumors at the time of the diagnosis.Moreover, osteoblastic, and fibroblastic, were the most common histological diagnoses.Primary metastases, response to the administered chemotherapy and the Enneking stage classification, are also reported in Supplementary Material Table 2.

Risk of bias within studies
The results of the quality and methodological evaluation of the eligible trials, based on the NOS scale, are displayed in Table 4 (Supplementary Material).All studies, except the trial of Ek et al. (25), achieved a 5-star grade.A significant interrater agreement was documented (Cohen's k statistic: 85.7%, p < 0.001).

Primary endpoint
In total, data concerning the hazard ratio of the overall survival were extracted from five studies (Figure 2).Meta-analysis of these data did not show a statistically significant (p ¼ 0.94) hazard ratio for OS between high and low MVD groups (HR: 1.68, 95%CI: 0.49-5.73).Heterogeneity levels were high (Q test P < 0.00001, I 2 ¼ 87%) and as a result, a RE model was applied.
Due to the high heterogeneity levels, further analyses were performed.Sensitivity analysis identified that the study of Kreuter et al. significantly impacted the levels of heterogeneity (Supplementary Material).Subgroup analysis (Supplementary Material Tables) for the type of study (p ¼ 0.007), center (p ¼ 0.007), maximum magnification (p ¼ 0.0007) and spots examined (p ¼ 0.0003) resulted to significant differences.The use of the anti-CD31 antibody did not alter the overall heterogeneity (Q test P < 0.001, I 2 ¼88%) or the pooled results (HR: 1.41, 95%CI: 0.26 − 7.69).Meta-regression (Supplementary Material) did not identify any statistically significant factor.
In total, three studies reported mean MVD measurements for tumors with primary metastases (Supplementary Material).Pooled comparisons did not confirm any significant difference between the tumors with and without metastases (WMD: −5.51, 95%CI: −28.68 − 17.65, p ¼ 0.64).
Finally, data extraction regarding chemotherapy response (Supplementary Material) was available for 109 patients.A statistically significant higher mean MVD value was recorded for tumors with a good response to adjuvant therapy (WMD: 16.74, 95%CI: 5.83 − 27.65, p ¼ 0.003).These results displayed a low level of heterogeneity (Q test P ¼ 0.88, I 2 ¼0%).

Risk of bias across studies
Visual inspection of the funnel plot of the primary endpoint (Supplementary Material),  revealed a symmetrical distribution of the trials on both sides of the combined effect size line.Egger's test (p ¼ 0.341) confirmed the absence of a statistically significant publication bias.

Discussion
Osteosarcoma is defined as a mesenchymal malignancy with tumoral cells that excrete osteoid and bone matrix (3).The fact that a significant proportion of the tumors present with primary metastases underlined the metastatic potential of this tumor.Therefore, systematic chemotherapeutic approach alongside surgical resection were proposed for the treatment of osteosarcoma (3,10,38).However, despite the fact that various first and second-line protocols were implemented, the survival rates have stagnated at the levels of 60% (39).Besides the lack of improvement in the effectiveness of the chemotherapy, the development of drug resistance, has been considered as another etiologic factor for the survival rate plateau (39).Chemotherapy resistance could be attributed to mechanisms such as decreased intracellular drug accumulation, drug inactivation, enhanced DNA repair, differentiations in signal transduction pathways, altered gene expression, cancer stem cells and autophagy (40).In order to overcome these obstacles, current research on novel potential agents focuses on areas such as tumor immunotherapy, targeted molecular therapy, gene therapy and antitumor angiogenesis therapy (39).
The formation of novel vessels is of utmost importance for the growth and dissemination of solid tumors (41).Neovascularization, is a multistep process and it is stimulated upon demand for nutrients and oxygen (41).More specifically, the reduction in the concentration of these elements in marginally ischemic tissues (e.g.malignant tumors) promotes the development and remodeling of blood vessels, through the upregulation of the angiogenic switch, that is the pro and antiangiogenic factors balance (11,41,42).Although the role of angiogenesis in carcinomas has been confirmed (43,44), the exact effect in sarcomas has not yet been clarified, since the two malignancies are characterized by different microvascularization patterns (16,45).
Angiogenesis and several of the related mediators have been studied in osteosarcoma (24,26).In vivo and in vitro experiments identified the role and the intersecting pathways of plateletderived growth factor (PDGF) and transforming growth factor-b (TGF-b) in osteosarcoma vasculogenesis (46).Moreover, reduction of apurinic/ apyrimidinic endonuclease-1 (APE1) expression in tissue samples, resulted in downgrading the TGF-b cascade (47).WW domain-containing oxidoreductase (WWOX), a discrete tumor suppressor, was found to impede osteosarcoma invasion and neovascularization (48).Jian et al. reported that, in xenograft osteosarcomas, overexpression of thrombospondin-1 (TSP1), a cellto-cell and cell-to-matrix mediator, resulted in reduced angiogenesis (49).Furthermore, connective tissue growth factor (CTGF), through upregulation of angiopoietin-2 in cell lines, was found to induce the formation of novel vasculature in osteosarcoma (50).
Subsequent trials have attempted to validate a correlation between angiogenesis and clinical outcomes in osteosarcoma.The intratumoral expression of the vascular endothelial growth factor (VEGF) has been found to significantly impact the OS and DFS rates of osteosarcoma patients (51)(52)(53).Overexpression of epidermal growth factor-like domain containing protein 7 (EGFL7), was positively associated with advanced stage tumors (47).In a meta-analysis by Ouyang et al. (54), high levels of hypoxia-inducible factor-1 (HIF-1), an important factor of tumorigenesis, angiogenesis and metastasis, were associated with a worse prognosis (HR: 3.67).
The association of MVD with the above-mentioned angiogenesis mediators and the prognostic role of tumoral vascularity on the clinical outcomes has been extensively researched.High vascularized tumors have been estimated to have a 84% 5-and 10-year overall survival rate, compared to 49% in low MVD tumors (24).The improved accessibility of the chemotherapeutic agents to the chemo-sensitive cancer cells, due to an increased microvascular patterns has been proposed for these results (24).Contrary results, however, have been published by other researchers (25,55).More specifically, Kunz et al, in a multicenter trial, utilized a computer-assisted whole slide analysis combined with a enzymatic epitope retrieval technique and concluded that low vascularization of osteosarcoma provided a significant overall and relapse-free survival rate superiority (37).Kubo et al., found that MVD was closely associated to patient age and chemotherapy response (26).Furthermore, a pairwise correlation between APE1, TGFb and MVD and an inverse relation of TGFb and MVD and patient prognosis was validated (36).Higher MVD values were also correlated with high EGFL7 expression (28).
The pooled comparisons between high and low vascularized tumors showed that MVD can not be used as a prognostic factor of survival in osteosarcoma, since a significant correlation between MVD and HR of OS and DFS was not validated.These results are in accordance with the outcomes from various trials (25,27).Literature reports suggest that the prognostic value of MVD in osteosarcoma is not the same as in carcinomas (25).As it has been already stated, the different vessel patterns have been proposed as the main reason for these results (25).More specifically, in carcinomas, the microvessels are clustered in bursts in contrast with sarcomas, where a more homogeneous distribution can be noted (56).Moreover, a possible reason for the diversity in the reported results in the literature, could possibly be the immunochemical stains applied (25).Although both CD31 and CD34 have been extensively used, there are inherent differences between the two stains, since CD31 binds to all endothelial cells despite the activation status, whereas CD34 is characterized by a low specificity and an affinity to lymphatics and perivascular stromal cells (57).
Although the metastatic status did not alter the mean tumoral MVD values, malignancies with a good chemotherapy response displayed a denser vascular pattern.Previous studies have, also, concluded to these results (24)(25)(26).It is widely acknowledged that osteosarcoma is a chemo-sensitive tumor and that response to chemotherapy is an important prognostic factor (3,14,39,55).A potential explanation for the above-mentioned findings could be the denser microcirculation of the tumor that allows an improved delivery of chemotherapy agents to the proliferating malignant cells (26).Consequently, the correlation between tumor vascularization and chemotherapy response could be of value in clinical settings and more specifically as a prognostic factor and as a pivot for the development of novel treatment modules (13,14,40,55).
Before appraising the results from this review, certain limitations should be acknowledged.First, heterogeneity levels in the primary and in several secondary endpoints were above the normal levels.Although sub-analyses were performed to identify possibly bias introducing factors, the significant heterogeneity combined with the small sample size reduces the validity of our results.Another possible source of bias could be the reported variations in the MVD assessment technique.Bias in the pooled results could also be introduced through the lack of stratification in confounding factors such as the tumor site, the histological diagnosis, and the Enneking stage.Moreover, all the included studies had a retrospective study design, thus lowering the grade of the overall evidence.Finally, since the estimation of the pooled HR and OR required the extraction and reconstruction of the approximate raw data from the provided KM curves, a certain level of bias should be anticipated through this procedure.
The present meta-analysis is the first study that estimates the pooled effect of MVD on the survival of patients with osteosarcoma.Higher tumoral MVD values were not associated with significant changes in the OS and the DFS.Moreover, the presence of primary metastases at the time of the diagnosis did not impact the mean MVD value.However, denser neovascularization of the tumor was correlated with higher 3year DFS, and good chemotherapy responders were characterized by higher MVD measurements.Since the exact relationship between MVD and osteosarcoma clinical outcomes is still not clear, and due to the above-mentioned limitations, we conclude that further prospective trials, with an adequate sample size and a higher methodological and quality level, are necessary.

Figure 2 .
Figure 2. Hazard ratio of overall survival.

Figure 3 .
Figure 3. Hazard ratio of disease-free survival.

Figure 4 .
Figure 4. Odds ratio of overall survival.