Boston Children Hospital, Harvard Medical School, Boston, USA.
Children's Mercy Hospital, Kansas City, USA.
Eur Spine J. 2019 Mar;28(3):559-566. doi: 10.1007/s00586-018-5821-8. Epub 2018 Nov 16.
This study aims to propose and validate a new unified "Risser+" grade that combines the North American (NA) and European (EU) variants of the classic Risser score. The "Risser+ " grade can effectively combine the North American and European Risser Classifications for skeletal maturity with adequate intra-rater/inter-rater reliability and agreement.
Agreement and reliability were evaluated for 6 raters (3-NA, 3-EU) who assessed 120 pelvic radiographs from the BrAIST trial, all female, average age 13.4 (range 10.1-16.5 years). Blinded raters reviewed x-rays at two time-points. Intra- and inter-rater agreement (RA) were established with Krippendorff's alpha (k-alpha), while intra- and inter-rater reliability (RR) were established with intraclass correlation coefficients (ICC). Acceptable agreement and reliability were set a priori at 0.80.
Inter-RA for the second reading met study requirements (k-alpha = 0.86 [0.81-0.90]) compared to the first reading (0.72 [0.63-0.79]) while combined readings was close to target agreement (0.79 [0.74-0.84]). Removal of 20 readings demonstrating outlier tendencies increased agreement for the first, second, and combined reads (k-alpha = 0.85, 0.89, 0.87, respectively). Intra-RA was sufficient for 4 out of 6 raters (k-alpha > 0.80) and one rater from EU and NA presented subpar intra-RA (k-alpha = 0.64 and 0.74, respectively). Inter-RR met study requirements overall reads (ICC = 0.96 [0.95-0.97]) including the first (0.94 [0.92-0.95]) and second (0.97 [0.97-0.98]) reads, independently.
The Risser+ system showed excellent reliability across multiple reads and raters and demonstrated 79% agreement overall reads and ratings. Agreement increased to over 85% when raters could distinguish Risser 0 + from Risser 5. These slides can be retrieved from electronic supplementary material.
本研究旨在提出并验证一种新的统一的“Risser+”分级,该分级结合了北美(NA)和欧洲(EU)经典 Risser 评分的变体。“Risser+”分级可以有效地将北美和欧洲的骨骼成熟度 Risser 分级结合起来,具有足够的内部和外部评估者之间的可靠性和一致性。
评估了 6 名评估者(3 名北美,3 名欧洲)对来自 BrAIST 试验的 120 张骨盆 X 光片的评估,所有女性,平均年龄 13.4 岁(范围 10.1-16.5 岁)。盲法评估者在两个时间点回顾 X 光片。内部和外部评估者之间的一致性(RA)通过 Krippendorff 的 alpha(k-alpha)来确定,而内部和外部评估者之间的可靠性(RR)通过组内相关系数(ICC)来确定。可接受的一致性和可靠性被设定为 0.80。
与第一次阅读相比,第二次阅读的内部评估者之间的一致性(k-alpha=0.86 [0.81-0.90])达到了研究要求,而合并阅读则接近目标一致性(0.79 [0.74-0.84])。剔除 20 份显示出离群趋势的阅读数据后,第一次、第二次和合并阅读的一致性都有所提高(k-alpha=0.85、0.89、0.87)。有 4 名评估者的内部评估者之间的一致性是足够的(k-alpha>0.80),而来自欧洲和北美的一名评估者的内部评估者之间的一致性较差(k-alpha=0.64 和 0.74)。整体阅读的内部评估者之间的可靠性(ICC=0.96 [0.95-0.97])包括第一次(0.94 [0.92-0.95])和第二次(0.97 [0.97-0.98])阅读,均达到了研究要求。
Risser+系统在多次阅读和评估者中表现出了极好的可靠性,并在整体阅读和评估中达到了 79%的一致性。当评估者能够区分 Risser 0+和 Risser 5 时,一致性增加到 85%以上。这些幻灯片可以从电子补充材料中检索。