In CPUP, all participating children have their CP diagnosis verified by a neuropaediatrician at four years of age. CP is defined as a non-progressive brain injury which has developed before the age of two years. Motor impairment and specific neurological signs are defined and classified according to the inclusion criteria of the Surveillance of Cerebral palsy in Europe (SCPE) network [18, 19]. Gross motor function is determined by the child’s physiotherapist according to the expanded and revised version of the Gross Motor Function Classification System (GMFCS) [20, 21]. This is a 5-level system for children and adolescents with CP based on their self-initiated movement where level I delineates the highest level of function and level V the lowest.
Children and adolescents aged 6–16 years at GMFCS levels II-V and enrolled in CPUP were recruited from five child rehabilitation units in southern Sweden. The participants and their families were informed about the study by their local physiotherapists and provided with invitation letters with information about the study. Written consent was obtained from the parents/guardians of all participants. Children were recruited consecutively until at least six children at each relevant GMFCS level had accepted. The decision to include six persons in each GMFCS level (except level I) was based on an earlier study evaluating the Posture and Postural Ability Scale in adults with CP [22, 23]. In addition, a reliability study of the scoliometer by Bonagamba and colleagues included 24 participants and that study had enough power to satisfactorily evaluate reliability [24]. Children at GMFCS level I, which constitutes about 40 % of all children with CP, do not have a higher risk of scoliosis compared to the risk of developing idiopathic scoliosis in adolescents [1] and were therefore not included in this study.
The children were examined at one occasion by three examiners, independent of each other, during a period from November 2013 to March 2014. The examinations were performed by two physiotherapists and one paediatric orthopaedic surgeon, all with 10 to 20 years of experience working with children with CP. Each examiner assessed each child individually and noted the grading on separate scoring sheets. They had no information of the children’s medical reports or radiographs, and they had not met the children prior to the examinations. All three examiners had previous experience of clinical examinations of children with CP, but only the orthopaedic surgeon had experience of scoliometer measurements in adolescent idiopathic scoliosis prior to this study. The spine was examined with the child in a sitting upright position, with external support if needed, and then, still in sitting, with the forward bending. The degree of scoliosis was noted according to the CPUP classification [1] and graded as:
-
No scoliosis.
-
Mild scoliosis: discreet curve visible only on thorough examination in forward bending.
-
Moderate scoliosis: obvious curve in both upright and forward bending.
-
Severe scoliosis: pronounced curve preventing upright position without external support.
In sitting position, a scoliometer was placed in forward bending at the top of the thoracic spine, with the 0 (zero) mark over the spinous process, and slowly moved down the spine noting the highest degree of truncal rotation. The degree of scoliosis was recorded separately and independently by the three examiners. A higher degree of truncal rotation indicates worse inclination. The value used to detect moderate scoliosis that should be referred to radiographic examination was set to ≥7°.
Radiographic examinations were performed in a sitting position, in an anteposterior (AP) projection. The magnitude of the curve was determined based on the Cobb angle [16] and moderate or severe scoliosis was defined as Cobb angle ≥20°. The radiographic examinations were performed at three radiology departments using standardized instructions for patient position and x-ray imaging. The Cobb angles were measured by three different radiologists with a special interest in skeletal radiology. The measurements were reevaluated and confirmed by a fourth independent examiner. None of the radiologists had information about the previous results of the clinical examinations or the scoliometer measurements.
Statistical analyses
In the statistical analysis the interrater reliability for the clinical spinal examination and the scoliometer measurement was calculated using weighted kappa scores [25]. The magnitude of weighted kappa was interpreted according to Fleiss 1981 where <0.40 signifies poor agreement, 0.40–0.74 fair to good agreement and ≥0.75 signifies excellent agreement [26]. To calculate 95 % CI for weighted kappa scores all GMFCS levels included were combined and 95 % nonparametric bootstrap confidence intervals were added based on a 1000 re-samples [27, 28].
To evaluate concurrent validity the Cobb angle was used as gold standard. Area under receiver operating characteristic curves (AUC), sensitivity, specificity and predictive values were calculated. The cutoff point chosen for clinical examination was no or mild scoliosis versus moderate or severe scoliosis. We used averaged ratings for analyzing validity of the scoliometer but not for calculation of kappa values.
The AUC is a measure of the capacity of a test to classify a person correctly. In this study the AUC was used as a measure of the capacity to correctly identify scoliosis according to our definition. A value of <0.5 is not better than random, >0.7 is acceptable, >0.8 is excellent, and >0.9 is an extraordinary capability [29].
Likelihood ratio (LR) is a summary of the diagnostic accuracy of a test telling the ratio of the probability of a certain test result in individuals who do have the disease to the probability in individuals who do not. The definition of a positive LR is sensitivity/ 1-specificity. The definition of a negative LR is 1-sensitivity/ specificity. The further a LR is from 1 the greater effect it has on the probability of scoliosis, e.g. a positive LR >10 means that a positive test is good at confirming a scoliosis, while a negative LR <0.1 means that a negative test is good at ruling out a scoliosis [30]. For all statistical computing R software environment version 3.0.0 and STATA version 13.1 were used.
Ethical approval was granted by the Medical Research Ethics Committee at Lund University, Dnr 467/2013.