Extension of the thoracic spine contributes to normal elevation of the shoulder[37
], and regardless of age, approximately 15° of thoracic extension is required for full bilateral arm elevation[37
], with unilateral arm elevation requiring approximately 9° of thoracic extension[38
]. Of relevance, a large thoracic kyphosis has been associated with reduced arm elevation in older subjects[37
] and a post-rehabilitation reduction in the kyphosis resulted in increased shoulder elevation range[27
]. Therefore a simple reliable clinical method to measure the thoracic kyphosis may be useful to determine how the kyphosis changes over time as well as the influence of individual or a collection of techniques aimed at reducing the kyphosis. This may enable enhanced understanding of interventions that reduce the kyphosis and if this reduction correlates with an improvement in function and reduction in shoulder pain. The results of this investigation suggest excellent intra-tester reliability for the clinical measurement of the thoracic kyphosis, as used in this study. Although the differences in the ICC(single) and ICC(average) results were small and by themselves do not support recommendations for measuring the kyphosis on one occasion or by calculating the mean of three successive measurements, the SEM findings suggest that there may be less error associated with the clinical measurement using the inclinometer method when the mean of three measurements is employed. This is because a smaller SEM is associated with a more reliable method. In subjects without symptoms, the SEM for ICC(single) for the kyphosis (combined T1/2 and T12/L1) measurements was 2.4° and the SEM for ICC(average) was 1.0°. For subjects with symptoms the SEM for ICC(single) for the kyphosis (combined T1/2 and T12/L1) measurements was 2.5° and the SEM for ICC(average) was 1.7°. This suggests that if an individual clinician wishes to measure the kyphosis using the mean of three sequential measurements, prior to and following an intervention, such as a stretching, strengthening and mobilisation programme aimed at changing the kyphosis, or in a longitudinal study, investigating the change in kyphosis angle over time, in subjects without symptoms, a decrease in the kyphosis of less than 1.0°, and for subjects with symptoms, a decrease in the kyphosis of less than 1.7°, should be considered as measurement error. Real change may be associated with values greater than these. These values appear to be clinically relevant as Wang et al[27
] reported that after a 6 week exercise programme in asymptomatic subjects the upper thoracic curve (C7-T7) became significantly (p < .01) less kyphotic by approximately 3° at rest and during movements of the shoulder. Bullock et al[26
] reported a mean increase in the thoracic kyphosis of 17.9° when subjects adopted a slouched posture. The slouched posture was also associated with a significant decrease in shoulder flexion range of movement. The methods to measure the kyphosis used in both of these studies required considerable set up times and equipment that would generally not be possible in normal clinical practice.
Research into the effect of the thoracic kyphosis on shoulder range of movement has frequently employed designs that place the subject into fixed or relatively fixed positions of thoracic kyphosis[26
]. This potentially would confound an understanding of the role of the kyphosis in shoulder function as this type of comparison produces a relatively unnatural fixed posture. This type of fixed posture would not occur typically in general clinical practice with the exception of conditions such as ankylosing spondylitis or Scheuermann's disease. As such an unconstrained method of reliably measuring the kyphosis would be preferable. The advantage of the method reported in this investigation is that it would allow the patient to adopt their natural posture as part of a cross-sectional or longitudinal assessment of the thoracic kyphosis and not require assessment in extremes of range or in a constrained manner.
The inclinometer used in this investigation is gravity dependent with the weight at the base so that the arrow faces superiorly. A similar method was described by O'Gorman and Jull[40
] where the inclinometer used had a downward facing arrow. This requires a different mathematical calculation for the determination of the thoracic kyphosis angle. However, both methods should ultimately produce a comparable finding. O'Gorman and Jull[40
] reported that three repeated measures in 20 subjects produced F- values below a critical F-value. The description of methods used suggests that only female subjects without symptoms were included in this reliability study and additionally they did not report ICC, 95%CI or SEM results to more fully determine the reliability of the procedure for measuring the thoracic kyphosis[40
]. The method described in the current investigation has been used by others[22
] but the reliability investigation only included 15 subjects with and without symptoms. Lewis et al [22
] reported ICC 2,1 results of .94 (95%CI .83- .98) with an SEM of 2.5°. These results are comparable with the findings of the current study.
A limitation of this study is that there is no certainty that this method is stable over time. It was determined that if the kyphosis was measured on separate days or after one week, then potential natural variations in posture occurring as a result of sporting, vocational or routine activity may artificially confound the reliability results. Another limitation is that there is no certainty that the spinal landmarks T1, T2, T12 and L1 were accurately palpated. Although the method used in this investigation is routinely used clinically in an attempt to identify spinous processes there is uncertainty as to the accuracy of clinical methods to identify bony spinal landmarks [42
]. It would be possible to determine the validity of the palpation technique using radiographs. This could be the focus of future research. Although this method appears to be very reliable, there is no certainty that the procedure is an accurate or valid measurement of the actually anatomical thoracic kyphosis. Future work should compare this method with the Cobb, modified Cobb, computer assisted method for deriving radius of thoracic spine curvature, and thoracic vertebral centroid angles as described by others[6
]. In addition, another limitation of this study is that only the intra-rater reliability was investigated. To be able to generalise this technique an investigation of the inter-rater reliability of the procedure is necessary.