This investigation supports using accelerometers as a possible replacement for existing clinical measures of sagittal, but not lateral, spine mobility in patients with radiographic-AxSpA. Our dataset also showed, for the first time, that lateral clinical tape measures of spine mobility have excellent criterion-concurrent validity.
The OST and its derivatives (i.e. MST and MMST) are the current clinical measures for diagnosing and monitoring sagittal plane spine mobility in patients with AxSpA. Consistent with previously published work, this study shows weak to moderate correlations between these measures of sagittal plane spine mobility and the radiographic gold standard. One possible explanation for these poor correlations is that the measures of sagittal plane spine mobility (OST, MST and MMST) are affected by stretching of the skin. For example, large systematic differences at end ranges of spinal flexion have been reported when using the current clinical tape measures.  At larger ranges of flexion the skin begins to slide across the underlying tissues rather than continuing to stretch, thereby causing disproportional changes to the Schober’s measurement.  Accelerometers present an interesting alternative for measuring sagittal plane spine mobility partly because their measurement is less influenced by skin stretching. Correlation between measures of sagittal plane spine mobility derived using accelerometers and the radiographic gold standard, reported herein, is an improvement over current clinical measures; however, the relationship using accelerometer-derived measures of sagittal plane spine mobility only explained 35% of the variance in spine mobility. Further, accelerometer-derived measures of frontal plane spine mobility systematically underestimated lateral bend angles from the radiographs and demonstrated a weaker correlation to the radiographic gold standard than either the LSFT or DT. Modeling techniques could potentially be employed to correct for these systematic differences that may improve the correlation. Future studies could include a larger dataset to develop and test such models.
Despite the fact that clinical measures of frontal plane spine mobility (LSFT and DT) are commonly used to assess radiographic-AxSpA patients,  the criterion-concurrent validity for these measures had not been previously established. Strong correlations observed between each of the clinical measures to the radiographic gold standard in this study suggest strong criterion-concurrent validity for the LSFT and DT. This is likely due to a combination of reduced impact of skin stretching and reduced variability in the measure since an average of measurements from left and right lateral bending is used to represent the clinical measurement. Taking the average from repeated measurements of spine mobility during forward bending could be considered as a potential way to improve the reliability and criterion-concurrent validity of the sagittal tape measures. Future work could examine this in more detail. This study’s findings support using lateral measures of spine mobility (not sagittal) for monitoring disease progression.
There are several limitations to this study. Although we used a broad recruitment strategy, this study was limited to a 3-month collection phase resulting in a small sample size. While the small sample limited the statistical power of the comparison between accelerometer and radiographs, it was great enough to determine the validity of the Domjan test, and to determine that it would be worth pursuing future research to develop accelerometers as tools to assist with the diagnosis and monitoring of spinal mobility impairments. A second limitation is that participants included in this study had advanced stages of the disease. There is a chance that correlations between measures might have been different in a population with greater range of spinal mobility. Finally, the underestimation of lateral bend spine angles as identified by the Bland Altman Analysis suggests a systematic difference in the measure obtained from the accelerometers, which points to a limitation of the instrumentation. As previously mentioned, it is possible that employing modelling techniques to correct for these differences could improve correlations. This is a possible direction for future research.