Comparison of automated bone scan indexes in patients with prostate cancer: aBSI and BONENAVI

OBJECTIVE: Bone scan indexes (BSIs) are frequently applied in Japan, Europe and the USA. The BONENAVI (Japan) and aBSI (Europe and USA) software programs were developed based on different training and validation datasets to calculate BSIs. However, whether they are compatible has not been systematically determined. We compared BSIs generated by these programs in a patient cohort to clarify correlations between them. METHODS: Tc-methylene-diphosphonate (MDP) were processed then BSIs were computed by aBSI and BONENAVI. Systematic differences between aBSI and BONENAVI were evaluated using Pearson's correlation coefficients and Bland-Altman analysis. We also categorized BSIs during follow-up as decreased, unchanged, or increased, and evaluated agreement between longitudinal changes. RESULTS: The mean number of bone scans per patient was 10.0 ± 6.8 (1-26), and longitudinal changes in BSIs were evaluated using 305 pairs of follow-up studies. The automatic BSIs from aBSI and BONENAVI were 3.23% ± 3.47% and 2.91% ± 2.84%, respectively. Both strongly correlated (r = 0.947, p < 0.001). Bland-Altman plots revealed that aBSI tended to yield slightly higher BSIs (mean difference, + 0.31% (95% CI, 0.19-0.44, p < 0.001). A subgroup analysis of BSIs > 7% revealed higher automatic BSIs from aBSI than BONENAVI (mean difference, + 1.57%; 95% CI, 0.99-2.15; p < 0.001). The rate of longitudinal changes in BSI during follow-up substantially agreed (78.7% vs. 76.4% with respective thresholds of ± 0.5% and ± 0.3% (κ = 0.63, Cramer's V = 0.65, p < 0.0001 for both). CONCLUSION: The aBSI and BONENAVI BSIs computed by the software programs strongly correlated and longitudinal changes during follow-up studies generally concurred. However, BSIs from aBSI tended to be higher among patients with an increased disease burden compared with BONENAVI. Clinicians and investigators should be aware of these differences when interpreting or comparing results among software programs.