African-American patients had significantly longer delays to surgical fixation of pathologic fractures related to musculoskeletal cancers and more surgical complications as compared with other patients, a matched-cohort analysis showed.
On average, Blacks waited a day longer for surgery (2.78 vs 1.70, P=0.005) and had 50% more adverse events (44.7% vs 29.8%, P=0.035). In a multivariable analysis, the disparities in wait time (incidence rate ratio 1.57, 95% CI 1.15-2.17, P=0.005) and adverse events (OR 1.86, 95% 1.01-3.42, P=0.047) persisted, despite no significant differences in demographic or clinical characteristics, Micheal Raad, MD, of Johns Hopkins Medicine in Baltimore, and colleagues reported during the Musculoskeletal Tumor Society (MSTS) virtual meeting.
The analysis involved National Surgical Quality Improvement Program data for 828 patients who had surgical fixation of pathologic fractures during 2012-2018. The study population included 94 African-American patients, who differed from the non-African-American patients only with respect to age (64.6 vs 67.0). Propensity scoring produced two cohorts of 94 patients each (one African-American, the other non-African-American), matched for age, sex, BMI, anemia, end-stage renal disease, independent living, congestive heart failure, and pulmonary disease.
“This is a very interesting study and obviously the findings need to be researched further and followed up on, but more importantly, we need to figure out what we can do to make the differences go away,” MSTS program co-chair Kurt Weiss, MD, of the University of Pittsburgh, said during a review of highly rated meeting abstracts.
AI to Predict Mortality in Extremity Metastasis
A clinical algorithm developed from artificial intelligence (AI) demonstrated precision and consistency for predicting mortality in extremity metastasis, according to results of a retrospective external validation study.
The findings came from an analysis of 264 patients who had long-bone metastases from various types of primary tumors, including renal cell (18%), lung (16%), and myeloma (14%). The patients had a 90-day mortality of 19% and a 1-year mortality of 42%. After integration of demographic and clinical characteristics, the algorithm’s predictive performance was associated with an area under the receiver operating curve (AUC) of 0.83 for 90-day mortality (95% CI 0.76-0.88) and 0.84 for 1-year mortality (95% CI 0.79-0.88).
The validation cohort differed substantially from a developmental cohort from the Netherlands with respect to primary tumor histology, previous systemic therapy, and 1-year survival. Nonetheless, the algorithm exhibited good correlation, including AUC, calibration, Brier score, and decision curve analysis, reported Mary K. Skalitzky, of the University of Iowa in Iowa City, and colleagues.
“This was very interesting because they found that even though the validation set included a different collection of malignancies than the development set, the tool still worked pretty good, which leads me to believe that they’re probably on to something,” said Weiss.
The algorithm application is freely available here.
Predicting Mortality After Pathologic Fracture Repair
An algorithm based on seven data points outperformed several existing tools for estimating 30-day postoperative mortality risk after surgical repair of pathologic fracture, results of a validation study showed.
The Pathologic Fracture Mortality Index (PFMI) was used to evaluate 1,219 patients who had fixation surgery during 2012-2018. The study population included 177 patients who did not survive beyond 30 days after surgery. The seven variables that formed the basis for the algorithm were preoperative hypoalbuminemia (<3.5 mg/dL), weight loss ≥10% of body weight in prior 6 months, pulmonary disease, alkaline phosphatase, dependence for daily living, white blood cell count >12,000, and preoperative anemia.
Hypoalbuminemia was weighted at three points, the next two at two points each, and the remaining four variables at one point each. The probability of death within 30 days increased from 4% with a score of 0-2 to 37% for a score ≥5.
The PFMI had an AUC of 0.75 for predicting 30-day mortality, significantly more accurate (P<0.01) than the American Society of Anesthesiologists physical classification index (AUC 0.60) or the modified five-item frailty index (AUC 0.58), Raad and colleagues reported in a poster presentation.
“They think this might be a useful tool for predicting who is going to do really poorly, really quickly,” said Weiss.
Raad and Skalitzky disclosed no relevant relationships with industry.