Ninety Percent or Greater Tumor Necrosis Is Associated With Survival and Social Determinants of Health in Patients With Osteosarcoma in the National Cancer Database

Abstract

Background

The histologic response of osteosarcoma to chemotherapy is commonly cited as a prognostic factor and typically graded as the percent necrosis of the tumor at the time of surgical resection. Few studies, to our knowledge, have examined the relationship of tumor necrosis relative to other factors. Existing studies are limited by prolonged enrollment periods or analysis of patient subsets without the strongest predictor of mortality: metastasis at diagnosis. Additionally, the definitive threshold value for a good histologic response is commonly set at more than 90% tumor necrosis with little evidence; some authors advocate other values.

Question/purposes

(1) Are there alternative cutoff values for a good response to chemotherapy in a large, national cohort of contemporarily treated patients with osteosarcoma? (2) How does the association of histologic response to survival in osteosarcoma compare with other clinicopathologic factors? (3) What patient and clinical factors are associated with the histologic response?

Methods

We identified 2006 patients with osteosarcoma diagnosed between 2010 and 2015 in the National Cancer Database (NCDB), a registry that includes 70% of all new cancers diagnosed in the United States with 90% follow-up. Patients were excluded for missing documentation of percent tumor necrosis (21% [425 of 2006]) or if definitive resection was not performed (< 1% [1 of 2006]). A total of 1580 patients were included in the analysis, with a mean follow-up duration of 37 ± 22 months. A Kaplan-Meier survival analysis, stratified by the percent tumor necrosis after chemotherapy, was performed for the 5-year period. Other covariates examined were sex, race, socioeconomic score composite, insurance type, Charlson/Deyo score, distance from the hospital, and location (metropolitan, urban, or rural). Clinical and sociodemographic data including patient-identified race from the patient’s medical record is input into the NCDB by certified registrars. The NCDB only allows coding of one primary race for each patient; thus, most of our patients were grouped as White or Black race and the remaining were grouped as Other for our analysis. A multiple Cox regression analysis was performed to evaluate the effect of percent necrosis compared with other demographic, clinicopathologic, and treatment effects on survival. Finally, a multiple logistic regression analysis was performed to assess demographic and clinicopathologic characteristics associated with percent necrosis.

Results

Five-year overall survival for patients with histologic gradings of 90% to 94% necrosis (70% [95% confidence interval (CI) 60.6% to 79.7%) and 95% to 100% necrosis (74% [95% CI 68% to 80.3%) was not different between groups (p = 0.47). A comparison of histologic responses below 90% necrosis found no difference in survival between patients with decreasing histologic response (p > 0.05). Necrosis of less than 90% was associated with worse survival (HR 2.00 [95% CI 1.58 to 2.52]; p < 0.001 compared with more than 90% necrosis), and factors most associated with poor survival were metastasis (HR 2.85 [95% CI 2.27 to 3.59]; p < 0.001) and skip metastasis at the time of diagnosis (HR 2.52 [95% CI 1.64 to 3.88]; p < 0.001). On multivariate analysis, adjusting for demographic, clinicopathologic, and treatment factors, social determinants of health were negatively associated with percent necrosis of 90% or more, including uninsured status (OR 0.46 [95% CI 0.23 to 0.92]; p = 0.02 compared with private insurance) and lower socioeconomic status composite (OR for the lowest first and second quartiles were 0.63 [95% CI 0.44 to 0.90]; p = 0.01 and 0.70 [95% CI 0.50 to 0.96]; p = 0.03, respectively). Race other than White or Black (OR 0.61 [95% CI 0.40 to 0.94]; p = 0.02 compared with White race) was also negatively associated with percent necrosis of more than 90% after controlling for available covariates.

Conclusion

This study suggests that a cutoff of 90% necrosis provides the best prognostic value for patients with osteosarcoma undergoing chemotherapy. Other threshold values did not show different survival benefits. Sociodemographic factors were associated with histologic response less than 90%. These associations must be carefully understood not as cause and effect but likely demonstrating the effects of health disparities and access to care. Although we controlled for multiple variables in our analysis, broad variables such as race may have been associated with histologic response due to unaccounted confounders. Medical providers should be aware of these associations to ensure equitable access and delivery of care because access to care may be responsible for these associations. Future studies should examine potential drivers of this observation, such as a delay in presentation or deviation from standard of care practices.

Level of Evidence

Level III, therapeutic study.

Introduction

Treatment of osteosarcoma, the most common primary malignancy of bone, includes a multidisciplinary approach of preoperative chemotherapy, surgical resection, and postoperative chemotherapy. A poor histologic response to preoperative chemotherapy is associated with worse event-free survival and overall survival [2, 37]. A threshold of 90% for a good response to chemotherapy is commonly used in place of traditional grading systems, although studies have used or proposed percent necrosis thresholds ranging from 70% to 95% [1-3, 5, 6, 16, 20, 25, 34, 37]. Bielack et al. [5], in their review of 1702 patients, described a stepwise decrease in 5-year survival with increasing percent necrosis, suggesting that histologic response is not simply a binary good or poor response. Metastatic disease at the time of diagnosis is also associated with worse survival [2, 3, 14, 18, 37]. Prior studies have found that patients with the lowest socioeconomic status composite and public insurance are at higher risk of mortality, and patients living in a county with low socioeconomic status have higher odds of presenting with metastatic disease [27, 29]. These studies did not examine whether socioeconomic factors are associated with the treatment response.

Two major limitations of existing studies on histological response in osteosarcoma are relatively small sample sizes and, in many cases, the lack of a contemporary cohort because patients were enrolled over the course of 15 years or more and received differing chemotherapy regimens [1-3, 5, 20]. The EURAMOS-1 trial, a large multinational investigation of contemporary chemotherapeutic strategies using a three-drug combination with methotrexate, doxorubicin, and cisplatin, used 90% necrosis as the cutoff for a “good” response based on this cutoff being used in prior studies [37]. Other response rates were not examined for association with survival, and it is unknown whether alternative cutoffs could be used with contemporary treatment strategies. Large cancer databases such as the National Cancer Database (NCDB) and Surveillance, Epidemiology, and End Results Program (SEER) offer the opportunity to address these questions. Additionally, these databases capture sociodemographic data on patients and allow identification of inequalities in outcomes associated with these factors. Studies to date using these databases have focused on epidemiology, prognostic factors besides histologic response, treatments, and survival in patients with osteosarcoma [10, 14, 15, 18, 29, 30]. The NCDB, but not SEER, collected the histological response of patients, namely during the years 2010 to 2015.

We therefore asked: (1) Are there alternative cutoff values for a good response to chemotherapy in a large, national cohort of contemporarily treated patients with osteosarcoma? (2) How does the association of histologic response to survival in osteosarcoma compare with other clinicopathologic factors? (3) What patient and clinical factors are associated with the histologic response?

Patients and Methods

Study Design and Setting

This was a large-database, retrospective, comparative study drawn from the NCDB, which longitudinally maintains a nationwide registry that includes 70% of all new cancers in the United States with 90% follow-up [7]. We reviewed the NCDB to identify patients with osteosarcoma treated with preoperative chemotherapy from 2010 to 2015. We selected these years because during this period, the NCDB contained documentation of the tumor percent necrosis as a measure of histologic response.

Data Source

The Participant User File is a publicly shared subset of the NCDB that contains deidentified demographic, clinicopathologic, treatment, and survival data items. These data are abstracted from the patient’s medical record by certified registrars.

Inclusion and Exclusion Criteria

We identified 2006 patients with osteosarcoma treated with preoperative chemotherapy from 2010 to 2015 (Appendix 1; http://links.lww.com/CORR/A928). We excluded patients if they were missing documentation of percent tumor necrosis after preoperative chemotherapy (21% [425 of 2006]) or if definitive resection was not performed (< 1% [1 of 2006]) (Fig. 1). All 1580 patients included in this study had a documented vital status at the last contact.

Fig. 1.

This diagram demonstrates the inclusion and exclusion of patients in this study.

Extracted Data and Outcomes

Demographic data including age, sex, race, socioeconomic score composite, insurance type, Charlson/Deyo score, distance from the hospital, and location (metro, urban, or rural) were queried. Patient race in the NCDB is self-reported and identified by database registrars from the patient’s medical record. Races other than White or Black were grouped together because of low numbers of patients with self-reported race other than White or Black. The socioeconomic status composite (SES) is inferred by linking patient ZIP Codes to the US Census and US Department of Agriculture Economic Research Service datasets to give information on the median income and urban versus rural residence. Patients were divided into quartiles based on SES composite, with the fourth quartile being patients with highest income and education levels and first quartile having lowest income and education level, as previously described [13, 28, 29]. The modified Charlson/Deyo score is a measure of comorbidities [7]. Clinical characteristics included in the analysis were tumor location and grade, size less than or greater than 8 cm, presence of a skip metastasis, presence of a metastasis at diagnosis, margin status, use of radiation therapy, and type of surgery (limb-sparing resection versus amputation). Percent necrosis was derived from the pathology report at the time of definitive surgical resection after preoperative chemotherapy and ranged from 0% to 100%. The mean follow-up duration in our cohort was 37 ± 22 months.

Patient Demographic and Clinicopathologic Characteristics at Diagnosis

The mean patient age was 23 ± 16 years, 59% (935 of 1580) were male, and 75% (1189 of 1580) were White (Table 1). Eight percent (118 of 1559) of patients with available race did not identify as White or Black and were grouped together for the analysis because of the low number of patients in other groups. Race other than White or Black included the minority groups of Asian/Pacific Islander and American Indian/Alaskan Natives as there were many fewer patients in each of these groups. Most tumors were in the lower extremity (72% [1138 of 1580]) or upper extremity (13% [204 of 1580]) and fewer in the pelvis/sacrum (5% [74 of 1580]) and axial skeleton (9% [134 of 1580]). Most patients had high-grade tumors (99% [1220 of 1236], after excluding missing data) without a metastasis at diagnosis (86% [1365 of 1580]). A total of 35% (550 of 1580) of patients had a histologic grade of 90% or greater necrosis, and 23% (366 of 1580) had a poor histologic response of 0% to 24% necrosis. The median (range) percent necrosis was 70% (0% to 100%).

Table 1.

Demographic and clinicopathologic data (n = 1580 patients)

Parameter	Value
Age in years	23 ± 16
Sex
Male	59 (935)
Racea
White	75 (1189)
Black	16 (252)
Other	8 (118)
Missing	1 (21)
SES composite
1	23 (371)
2	29 (460)
3	28 (447)
4	18 (288)
Missing	1 (14)
Insurance type
Private	62 (976)
Government	33 (519)
Uninsured	4 (60)
Missing	2 (25)
Charlson/Deyo score
0	92 (1456)
1	7 (103)
2 or more	1 (21)
Patient location
Metropolitan	83 (1389)
Urban	13 (212)
Rural	1 (20)
Missing	2 (39)
Tumor location
Lower extremity	72 (1138)
Upper extremity	13 (204)
Pelvis and sacrum	5 (74)
Axial skeleton	9 (134)
Other	2 (30)
Grade
Low (G1)	1 (16)
High (G2-4)	77 (1220)
Missing	22 (344)
Tumor size
0 to 8 cm	35 (553)
> 8 cm	57 (902)
Missing	8 (125)
Discontinuous tumor in primary bone
No	96 (1516)
Yes	3 (43)
Missing	1 (21)
Metastatic disease at diagnosis
No	86 (1365)
Surgery type
Limb-sparing resection	79 (1255)
Amputation	20 (317)
Missing	1 (8)
Margin status
Negative	89 (1407)
Positive	9 (136)
Missing	2 (37)
Radiation
No	96 (1511)
Yes	4 (58)
Missing	1 (11)
Percent necrosis (continuous, %)
Mean ± SD	59.5 ± 35.6
Percent necrosis (categorical)
0% to 24%	23 (366)
25% to 49%	11 (169)
50% to 59%	8 (128)
60% to 69%	6 (96)
70% to 79%	8 (128)
80% to 89%	9 (143)
90% to 94%	10 (160)
95% to 100%	25 (390)
Length of follow-up in months
Mean ± SD	36.7 ± 22.1
Missing	0.01 (1)

Data presented as mean ± SD or % (n).

^aRace was self-reported and identified by database registrars from the patient’s medical record. Socioeconomic status (SES) composite is a measure of community income and education level. Location was defined based on county population: rural (< 2500), urban (25,000 to 50,000), metropolitan (> 50,000).

Ethical Approval

No identifying patient information is included in the NCDB; therefore, this study was exempt from institutional review board approval.

Statistical Analysis, Study Size

Missing variables were addressed before regression analysis through multiple imputation by chained equations using five imputations based on the proportion of missing variables [19]. In the overall dataset, the proportion of missing data was 3% for urban or rural facility location, 22% for tumor grade, 8% for tumor size, and 2% for margin status. All other values needing imputation were missing for less than 2% of patients. There were no missing data for percent necrosis or vital status at the last examination because these data were required to satisfy the inclusion criteria.

Kaplan-Meier survival analysis, stratified by the percent tumor necrosis after chemotherapy, was performed for the 5-year period. Percent necrosis groups were selected to evenly distribute patient numbers across clinically practical thresholds. Thresholds were 0% to 24%, 25% to 49%, 50% to 59%, 60% to 69%, 70% to 79%, 80% to 89%, 90% to 94%, and 95% to 100%. Survival rates at 2 years and 5 years were calculated, and overall survival was compared using the log-rank test and pairwise log-rank test for multiple comparisons.

Next, to adjust for demographic, clinicopathologic, and treatment parameters associated with survival, we performed a multiple Cox regression analysis to evaluate the effect of a percent necrosis less than 90%. Factors included in the multiple Cox regression analysis had p < 0.20 in univariate analysis. We assessed collinearity with a variance inflation factor, which was less than 2 for each variable included. The proportional hazard assumption was validated by examining Schoenfeld residuals to assess independence versus time [38].

Finally, we performed a multiple logistic regression analysis to assess demographic and clinicopathologic characteristics associated with percent necrosis at or above the 90% threshold suggested by the Kaplan-Meier analysis. The logistic regression model was generated by backwards selection using each imputed dataset to optimize the Akaike information criterion, whereby a lower number indicates higher quality of the model. A c-statistic equivalent to the area under the receiver operating characteristic curve was calculated and was determined to be 0.6 or greater for all datasets, indicating a reasonable model [21].

Inferential estimates (ORs, HRs, confidence intervals [CIs], and p values) were pooled from each imputed dataset according to the Rubin rules for inferences, after multiple imputation, as described [9]. Statistical analyses were performed with packages MICE (Multivariate Imputation via Chained Equations), survival, and DescTools in R (R Foundation; https://www.r-project.org/). Data were visualized in R with packages ggplot2 and forestplot. All statistical testing was two-sided, with a p value < 0.05 considered significant.

Results

Histologic Response of 90% Necrosis Is Associated With Increased Survival

Patients with 90% to 94% necrosis had greater 5-year survival (70% [95% CI 60.6% to 79.7%]) than patients with 0% to 24% (59% [95% CI 53.2% to 66.1%]; p = 0.03), 25% to 49% (54% [95% CI 44.4% to 64.3%]; p = 0.03), 50% to 59% (49% [95% CI 39.8% to 61.0%]; p < 0.001), 60% to 69% (51% [95% CI 40.0% to 65.6%]; p = 0.01), 70% to 79% (52% [95% CI 42.9% to 62.8%]; p = 0.001), and 80% to 89% necrosis (57% [95% CI 46.8% to 68.7%]; p = 0.04) (Table 2). Patients with 95% to 100% necrosis had greater 5-year survival (74% [95% CI 68.0% to 80.3%]) than all patient groups with a percent necrosis below 90% (all p values < 0.001). Five-year overall survival for patients with histologic gradings of 90% to 94% necrosis (70% [95% CI 60.6% to 79.7%]) and 95% to 100% necrosis (74% [95% CI 68% to 80.3%]) was not different between groups (p = 0.47). Likewise, there was no difference in 5-year survival between patient groups with a histologic grade below 90% necrosis (p > 0.05) (Fig. 2).

Table 2.

Overall survival by histological grade

Percent necrosis	2-year overall survival (95% CI), %	5-year overall survival (95% CI), %
0-24	79 (74.2-83.1)	59 (53.2-66.1)
25-49	75 (68.4-82.3)	54 (44.4-64.3)
50-59	71 (62.6-79.6)	49 (39.8-61.0)
60-69	74 (65.1-79.6)	51 (40.0-65.6)
70-79	77 (70.1-85.2)	52 (42.9-62.8)
80-89	78 (71.2-85.6)	57 (46.8-68.7)
90-94	90 (84.6-94.9)	70 (60.6-79.7)
95-100	93 (89.8-95.4)	74 (68.0-80.3)

Fig. 2.

This figure shows the unadjusted survival analysis stratified by percent tumor necrosis after preoperative chemotherapy. A Kaplan-Meier survival curve is shown with the number at risk for each group, for all histologic types combined; ns denotes no significant difference among groups. ^aStatistical significance between groups less than 90% and 90% or more necrosis. A color image accompanies the online version of this article.

Histologic Response Has Strong Association With Survival Compared With Other Factors

We found that less than 90% necrosis was associated with worse survival compared with at least 90% necrosis (HR 2.00 [95% CI 1.58 to 2.52]; p < 0.001). After adjusting for demographic, clinicopathologic, and treatment factors, those associated with the highest risk of death were metastatic disease at presentation (HR 2.85 [95% CI 2.27 to 3.59]; p < 0.001) and skip metastasis (HR 2.52 [95% CI 1.64 to 3.88]; p < 0.001). Increasing patient age, male sex, race other than White or Black compared with White race, positive margins, tumor location, and treatment with radiation were associated with worse survival but had a smaller risk of death by HR than histologic response. Lower socioeconomic status, uninsured status, and type of surgical treatment were not associated with mortality (Fig. 3).

Fig. 3.

This figure shows the multiple Cox regression analysis of factors associated with death. Unstandardized HRs are displayed on a logarithmic scale with 95% CIs.

Patient and Tumor Factors Are Associated With Histologic Response of ≥ 90% Necrosis

Patient socioeconomic factors had the strongest associations with histologic response. Uninsured status (OR 0.46 [95% CI 0.23 to 0.92]; p = 0.02 compared with private insurance) had the strongest negative association with a histologic response of at least 90% necrosis. Socioeconomic composite in the first quartile (OR 0.63 [95% CI 0.44 to 0.90]; p = 0.01) and second quartile (OR 0.70 [95% CI 0.50 to 0.96]; p = 0.03) were negatively associated with percent necrosis more than 90%. Other race (OR 0.61 [95% CI 0.40 to 0.94]; p = 0.02 compared with White race) was also a negatively associated with percent necrosis 90% or more (Fig. 4).

Fig. 4.

This figure shows the multiple logistic regression analysis of factors associated with percent necrosis at least 90%. Unstandardized ORs are displayed on a logarithmic scale with 95% CIs.

Patients with tumors of the pelvis (OR 0.45 [95% CI 0.23 to 0.86; p = 0.02) and axial skeleton (OR 0.43 [95% CI 0.26 to 0.70]; p < 0.001) were negatively associated with percent necrosis greater than 90% compared with tumors in the lower extremity (Fig. 4). Patients treated with amputation (OR 0.66 [95% CI 0.49 to 0.88]; p = 0.01 compared with limb-sparing procedures) and presence of positive margins (OR 0.63 [95% CI 0.40 to 0.98]; p = 0.04) were negatively associated with percent necrosis more than 90%.

Discussion

The histologic response of osteosarcoma to chemotherapy is an important prognostic factor in the treatment of osteosarcoma and is graded by percent necrosis, although other descriptors exist [2, 4, 12, 36, 37, 39]. The cutoff of 90% necrosis has been studied in contemporary cohorts receiving modern chemotherapeutic regimens and found to be associated with survival, but data on the validity of other cutoff values are lacking. Evaluating other cutoff values of good histological response has previously been limited by small patient sample size and variability in treatment regimens because of long study periods. The study of histological response has also established several clinical factors associated with histological response; however, the relationship with patient demographic factors has not been studied. We aimed to use a contemporary cohort of patients in the NCDB to describe other cutoff values for histologic response as a prognostic factor and the association of patient factors with treatment response. Our study of 1580 patients supports the use threshold of more than 90% necrosis and shows strong association of this threshold with survival, similar to other well-known factors associated with survival such as the presence of metastasis at diagnosis. Importantly, we found that social determinants of health, but not the presence of a metastasis, are associated with histologic response.

Limitations

This study has several limitations that are common to large-database studies. First, the data input in the NCDB relies on multiple institutions, and there is potential for variations in treatment protocols. Our focus on patients treated between 2010 to 2015 limits the variation in treatment and histological grading among institutions more than prior studies that had longer enrollment periods. The NCDB relies on trained registrars to gather data from the medical record and uses automated systems to detect missing and inconsistent data [7]. Second, the NCDB data come from hospitals with American College of Surgeons Commission on Cancer accreditation, which applies to approximately 30% of hospitals in the United States. This may induce selection bias for patients with access to these facilities. However, we believe these findings are generalizable as the NCDB captures 70% of new cancer diagnoses [31]. Additionally, the strict follow-up requirements for the NCDB make this database ideal for survival analysis. Third, we are limited by the sample size of patients with osteosarcoma diagnosed between 2010 and 2015. Our study may be underpowered to show a difference in survival among groups with less than 90% necrosis. Fourth, we did not include histologic subtype in our analysis as most cases (68% [1069 of 1580]) had unspecified histology. Prior studies have suggested some subtypes may be more resistant to chemotherapy than others [1, 3, 5]. These prior studies have suggested that subtypes with worse histological response have worse survival, consistent with our findings. Finally, we report the association of sociodemographic factors (insurance, socioeconomic status, and race) with the biological outcome of tumor response to chemotherapy. Reporting of race in the NCDB is dependent on patient-reported race extrapolated from the medical record and has a number of shortfalls. The NCDB only allows coding of one primary race for each patient. Most (91%) of our patients were grouped as White or Black race and the remaining were grouped as Other. This “other” group was composed of multiple minority groups and had a statistically significant decrease in OR for good tumor response after controlling for available variables. These associations must be carefully understood, not as cause and effect, but as likely demonstrating the effects of health disparities and potential unaccounted covariates. Patients with limited access to care may be diagnosed later or receive inadequate care. Consequently, we believe it is important to highlight these social factors while understanding the limitations of reporting sociodemographic factors influence of biologic outcomes.

Histologic Response of 90% Necrosis Is Associated With Increased Survival

We found that the 5-year overall survival for patients with histologic gradings of 90% to 94% necrosis and 95% to 100% necrosis were greater than that of patients with less than 90% necrosis, thus endorsing a threshold value of higher than 90%. Interestingly, we did not see a stepwise decrease in survival for patients in groups with less than 90% necrosis as suggested by Bielack et al. [5], who saw a stepwise decrease in survival with increasing grade of histologic response based on the Salzer-Kuntschik grading system and reported 5-year overall survival of 77.8% ± 2% in patients treated between 1980 and 1998. Our study expands on prior investigations of the association of histologic grade to survival in patients with osteosarcoma, including those with metastatic disease at diagnosis and tumors of all locations, treated recently in the United States. Increased survival after a good histologic response was first reported in early studies using modern chemotherapeutic regimens [11, 26, 35]. Most recent studies have used 90% tumor necrosis as the threshold for a good response to chemotherapy and have shown this is associated with event-free and overall survival [1-3, 5, 16, 17, 33, 37, 40]. Other studies have shown that survival is associated with a threshold of 95% [6, 34] or even less than 90% [20, 25, 39]. Two large studies conducted over approximately a 20-year period in the 1980s to 2000s examining patients with nonmetastatic osteosarcoma of the extremities found higher overall 5-year survival in patients with 90% or more necrosis than in those with less 90% (73% to 78.4% versus 48% to 63.7%) [2, 40] but did not examine other cutoff values. Our results reinforce the binary cutoff of 90% for a “good” response, and providers should be wary of interpreting histological response as a spectrum of better prognosis as response approaches 90% to 100%. Histologic response is commonly used as a decision point in therapeutic trials, and future studies may consider values of histologic response of less than 90% as carrying similar mortality. Future studies should examine whether patients with less than 90% necrosis have different response to treatment intensification to assess these findings.

Histologic Reponse Has Strong Association With Survival Compared With Other Factors

We found patients with less than 90% necrosis had a higher risk of death according to HR than all patient and clinical factors except the presence of skip metastasis or metastatic disease at diagnosis. Factors with the largest HR for mortality in our study were metastatic disease at presentation and skip metastasis, with more than a twofold increased risk of mortality. Skip metastasis and presence of metastasis are well-known risk factors for mortality in osteosarcoma [5, 18, 20, 22-24, 29, 30, 37], and we show that a histologic response less than 90% has a comparable risk of death according to HR in a contemporary group of patients. Prior studies have reported that a poor histologic response increases the risk of mortality by greater than twofold, similar to our findings. A multinational analysis of patients with osteosarcoma treated over a 16-year period had similar findings regarding the HR for mortality in patients with less than 90% necrosis, which was second only to metastasis in increasing risk [33]. A more recent analysis of patients younger than 40 years of age without metastatic disease also reported a greater than twofold risk of mortality for patients with less than 90% necrosis, but that study could not compare this risk with that of metastasis because the authors analyzed patients with nonmetastatic disease [37]. That analysis found that poor histologic response was second only to intralesional resection regarding mortality risk. A 20-year multinational study of patients 40 years and older with metastatic disease reported that a histologic response less than 90% was the strongest risk of mortality but did not include metastatic disease or quality of resection [40].

Patient and Tumor Factors Are Associated With Histologic Response of ≥ 90% Necrosis

Here, we demonstrate the association of several social determinants of health with the biological outcome of histologic response to chemotherapy. Patient social factors of uninsured status, lower socioeconomic status, and race were negatively associated with percent necrosis of 90% or more. In our study, the socioeconomic factors found to be negatively associated with histologic response may represent difficulties in accessing care by these patients and thus a delay in care, which has been shown to be a risk factor for poor histologic response [5]. Medical providers should be aware of patient factors that may lead to difficulties with access to care, and future studies should examine whether the barrier of lower socioeconomic status leads to incomplete treatment or disrupted chemotherapy schedules. We found that race other than White or Black had an association with worse histological response compared with White patients. This observed association with a poorly defined variable such as patient-identified race should be carefully understood as it is likely due to unaccounted variables that the NCDB is unable to gather. These differences may be another reflection of lower access to care among minority groups; however, future studies should continue to examine for inequalities in outcomes for these groups. Socioeconomic factors have been suggested to be associated with survival in patients with osteosarcoma [32]. One study reviewed socioeconomic measures of 3503 patients in the NCDB with osteosarcoma who were younger than 40 years and found that patients in the lowest quartile of composite socioeconomic status and those with Medicaid insurance had decreased survival at 5 years [29]. We did not find an association between survival and socioeconomic status. However, our study was limited to the years that the NCDB collected histological response data. Our analysis included many fewer patients than this prior study and may be underpowered to repeat this finding.

Our nationwide study of patients with osteosarcoma of all locations and stage found that disease factors of location, surgical treatment, and margin status were associated with percent necrosis. Tumors of the pelvis were more likely to have percent necrosis less than 90% than tumors in the lower extremity. We find this to be consistent with the observation that patients with axial and pelvic osteosarcomas are known to have worse prognosis and more severe disease at presentation [5, 8, 37]. Patients treated with amputation were more likely to have percent necrosis less than 90% than patients who underwent limb salvage procedures. Patients suspected of having a poor prognosis or those who cannot tolerate traditional chemotherapy may be offered amputation. Finally, the presence of positive margins was associated with percent necrosis less than 90%. Microscopic disease at the surgical margin may not be fully addressed by preoperative treatment if the chemotherapeutic response is poor. Prior studies have reported histologic subtype, the chemotherapeutic regimen, presence of metastasis, delay in presentation, and male sex to be associated with poor histologic response [1, 3, 5]. These studies primarily contained younger patients with nonmetastatic extremity osteosarcomas who were treated in the 1980s through early 2000s. Interestingly, metastatic disease at presentation and skip lesions were not associated with histologic response but were strongly associated with poor survival. This suggests that patients who present with advanced disease may still have good histologic responses to chemotherapy.

Conclusion

We found that 90% or more necrosis in the histologic grading of osteosarcoma response to chemotherapy is an appropriate cutoff for good response to treatment of osteosarcoma, and sociodemographic factors are associated with histologic response in patients recently treated in the United States. The multidisciplinary care of osteosarcoma patients should include indentification of barriers to care to ensure equitable care is received. Future studies should examine whether delay in presentation or deviation from standard of care treatments drive the observed sociodemographic associations with histologic response and how care can be made more equitable for all patients.