Results

Patient characteristics

One hundred twenty-one patients were included in the initial database review (Figure 2). Of these, 92 patients met inclusion criteria. Exclusion justification included patients that received more than two neoadjuvant chemotherapy cycles (N=15) or less (N=3). Two patients had surgical resection prior to initiation of chemotherapy and three had no surgical resection performed. An additional six patients received a neoadjuvant regimen other than MAP. In addition to the 29 patients who did not meet inclusion criteria, seven patients did not have adequate medical records, resulting in 85 patients included in the final analysis. Demographic data is shown in Table 1. The median age was 13.9 years. Thirteen (15.3%) had metastatic disease at the time of diagnosis. Fifty-one patients (60.0%) had an osteoblastic subtype, with the second most common subtype being chondroblastic (N=17, 20.0%). Forty-two (49.4%) patients had a femur primary, followed by tibia (N=20, 23.5%). Eighty (94.1%) of patients received the expected four doses of neoadjuvant methotrexate; 3 (3.5%) patients received less than four doses and two (2.4%) patients received more than four doses. Ten (11.8%) patients had positive resection margins at the time of LC. The median interval length was 16 days. Demographic and clinical information was also divided into those with intervals greater than 14 days (51 patients, 60.0%) and less than or equal to 14 days (34 patients, 40.0%). Percentage of patients with metastatic disease was 13.7% and 17.6% respectively.

Tumor Necrosis

There is weak correlation between the interval of last methotrexate dose to surgery and tumor necrosis (rho = -0.18 p = 0.096) (Figure 3) with a trend that longer interval lengths correlate with decreased tumor necrosis. However, there is significant variability in the data, and this relationship is not statistically significant. In multivariable linear regression analysis, each additional day of interval length corresponds to a 1.1% decrease in tumor necrosis among patients with the same age at diagnosis, tumor site, histological subtype, stage, and year of diagnosis, though this estimate does not reach statistical significance with a 95% confidence interval (CI) of a 2.3% decrease to a 0.1% increase (Supplemental Table 1).

Event Free Survival

An interval length of greater than 16 days demonstrated lower 5-year EFS than less than or equal to 16 days (p = 0.019, Figure 4C). Similarly, interval length greater than 17 days corresponded to lower 5-year EFS than less than or equal to 17 days (p = 0.014, Figure 4D). There was no statistically significant difference in 5-year EFS for patients with an interval greater than 14 days versus less than or equal to 14 days or patients with an interval greater than 15 days versus less than or equal to 15 days, although similar trends were observed (Figure 4A and 4B respectively). However, in a multivariable Cox model adjusting for relevant confounders, each additional day of interval length did not correspond to increased risk of an event in a statistically significant manner (HR = 1.04, 95% CI: 0.98 to 1.10; Figure 5A and 5B). Metastatic disease at the time of diagnosis and positive surgical resection margins were associated with an increased risk of having an event (HR = 6.8, p < 0.001; HR = 3.0, p = 0.024). When performing a subgroup analysis on patients with localized disease at the time of diagnosis, each day of increased interval length resulted in a 1.1 times greater hazard rate (95% CI: 1.02 to 1.19, p = 0.016) of having an event in patients with the same age at diagnosis, histological subtype, tumor site, and margin status at surgery (Figure 5C and 5D). This suggests that the cumulative impact of many days of delay could be substantial. Positive surgical resection margins also correlated with an increased hazard of having an event (HR = 3.5, p = 0.020).