What is the clinical detection threshold for lower limb length inequality? In silico study of reproducibility and optimization using a centimeter graduated support.

Abstract

Introduction: The threshold of a Leg Length Discrepancy (LLD) by clinical examination on a sheet or centimeter paper (CP) is not known precisely whether or not it concerns limbs equipped with a hip prosthesis. We therefore conducted a prospective in silico study in order to: (1) determine the reproducibility and sensitivity of the clinical measurement of the LLD in different ideal and "degraded" clinical situations, (2) determine the threshold from which the human eye is capable of detecting a length inequality in clinic, (3) to determine whether the use of a graduated support (centimeter paper) improves the clinical measurement threshold.

Hypothesis: Our hypothesis was that clinical measurement on a centimeter support would improve clinical measurement accuracy.

Material and methods: This was an in silico study, the experiment was conducted on a mannequin. Different inequalities were created on a mannequin and photographed with a total of 30 inequalities from -22 to +22 mm on sheet or centimeter paper (CP). This was a multicenter study, with 40 different readers. We asked the readers to make a second measurement one month later. We evaluated the inter- and intra-observer reproducibility. The error rate at the threshold of 3 mm and 5 mm were calculated versus the gold standard. Finally, we determined at which thresholds respectively 75% and 95% of the measurements were correct.

Results: A total of 4140 measurements were performed and compared to the gold standard. With a threshold of 75% accurate measurement, the LLD detection threshold was 2.8 mm on centimeter paper and 4.5 mm on sheet. With a threshold of 95% accurate measurement, the LLD detection threshold was 3.4 mm on centimeter paper and 5.2 mm on sheet. Interobserver agreement (assessed overall on the 40 observers by Krippendorff's generalized Kappa) was 0.86 (95% confidence interval (CI95%) = 0.79 to 0.92) on CP and 0.71 (CI95% = 0.63 to 0.79) on sheet. Intra-observer agreement assessed by the intraclass correlation coefficient among observers who made 2 measurements had a median value (IQR) of 0.96 (0.94 to 0.99) on CP and 0.90 (0.83 to 0.94) on sheet.

Discussion: The clinical detection threshold on sheet at the patient's bed appears close to 5 mm. A more precise measurement is possible with graduated centimeter paper. A study in daily practice on patients in real situations would confirm our results.

Level of evidence: III; prospective diagnostic comparative in Silico study.