We then examined percentage of patients achieving 90% necrosis on histology in each study. Several studies included in the OS analysis did not report percentage necrosis, leaving 29 studies for the necrosis analysis. Of these 29 studies, all 29 included patients with metastatic and nonmetastatic disease,4,6,11,14,15,18,19,21,29,31-36,37,39,40,43-47,49,50,54-57,59 and 13 included only patients with nonmetastatic disease.6,14,15,29,32-35,40,47,54,55,57 Again, because of the known difference in prognosis between patients with metastatic disease and patients with nonmetastatic disease, we performed separate analyses, one for the combined dataset of all 29 studies (Figure 3) and the other for the 13 nonmetastatic studies (Figure 4). Random-effects models showed 90% necrosis for 50% of patients in both analyses: studies that included patients with metastatic and nonmetastatic disease (95% CI, 45%-54%; heterogeneity test, Q = 692.88; P < .001; I2 = 95.5%) and nonmetastatic studies (95% CI, 41%-59%; heterogeneity test, Q = 385.42; P < .001; I2 = 96.9%).
We also performed a meta-regression analysis that included necrosis as a continuous variable for both the overall dataset and the nonmetastatic dataset. Five-year OS was plotted against percentage of patients achieving 90% tumor necrosis for each study. The results are plotted in Figure 5 (combined dataset).
No evidence of publication bias was detected for 5-year OS or percentage necrosis for the analyses of the combined datasets by either Egger test or Begg test. For 5-year OS, Ps were .21 (Egger) and .19 (Begg); for percentage necrosis, Ps were .10 (Egger) and .62 (Begg). In addition, no evidence of publication bias was detected for the analyses of the nonmetastatic studies by either test. For 5-year OS, Ps were .55 (Egger) and .41 (Begg); for percentage necrosis, Ps were .42 (Egger) and .95 (Begg).
Discussion
Five-year OS was 63% (95% CI, 60%-66%) for studies that included patients with metastatic and nonmetastatic osteosarcoma and 71% (95% CI, 67%-76%) for studies that included only patients with nonmetastatic osteosarcoma. These percentages fall within the range found in the literature. Mankin and colleagues37 reviewed 648 cases of patients with osteosarcoma treated at Massachusetts General Hospital in 2004; OS was 68%. In 2011, Sampo and colleagues60 reported 10-year OS of 63% for patients with metastatic and nonmetastatic disease and 73% for patients with local disease at presentation. Five-year OS rates in the literature are consistently about 70%. Ferrari and colleagues61 reported 5-year OS of 73% and 74% for 230 patients treated with 2 different neoadjuvant chemotherapy regimens between 2001 and 2006. The consistency in 5-year OS suggests OS of pediatric patients with osteosarcoma has plateaued, and there has been no significant improvement in survival of patients with osteosarcoma over the past 30 years.
Histologic response to preoperative chemotherapy is strongly associated with survival in pediatric osteosarcoma. Bielack and colleagues31 reported 5-year OS of 75% to 80% for patients who responded well to preoperative chemotherapy (>90% tumor necrosis) and 45% to 55% for patients who responded poorly (<10% necrosis). In our meta-analysis of studies that included patients with nonmetastatic osteosarcoma, 50% achieved necrosis of more than 90%. Percentage of patients achieving necrosis of more than 90% has been about 45%, according to past reports. In 2012, Ferrari and colleagues61 reported that 45% of 230 patients treated with neoadjuvant chemotherapy achieved more than 90% tumor necrosis. Therefore, 5-year OS and percentage of patients achieving 90% necrosis are consistent with previous reports, though this also suggests these numbers have remained constant over the past several decades.
Despite its expansive scale, our study has several important limitations. Data were extracted from published studies, and individual patient data were not available, so we were not able to assess the effects of risk factors (eg, tumor size, location) on 5-year OS. We could not correlate the proportion of patients with 90% necrosis to 5-year OS, as studies did not report OS by necrosis strata. Also, because our numbers were derived from published studies, they may not accurately represent outcomes in the community as a whole. In addition, several successive studies may contain duplicate patient cases. We limited our search to studies published since 2000 to include patients recently diagnosed and treated for osteosarcoma; however, several studies published after 2000 also included patients diagnosed and treated before 2000. Several of these studies are from countries outside the United States and may have a significantly different incidence of osteosarcoma as well as treatment methods and survival rates.
Although this meta-analysis suggests 5-year OS remains about 70% for patients with primary nonmetastatic osteosarcoma, we cannot settle on this conclusion because of the many differences between the studies we included. Therefore, more studies of patients diagnosed and treated within the past 10 years are needed to confirm our beliefs about patient survival.
