We found, consistent with our main hypothesis, that the prevalence of hip dislocation was higher in the Norwegian counties providing regular care compared with Southern Sweden providing hip surveillance services. In contrast to our second hypothesis, we found that the surveillance programme did not lead to an increase in the total number of hip operations. However, children in the surveillance programme were operated at an earlier age than children offered regular care.
The difference in prevalence of hip dislocation and operations between the two regions is unlikely to be due to chance as suggested by the low p
-values. A strength of the present study is the comparison with contemporary rather than historical data, as has been the case in other studies [6
]. Thus, our results indicate that a general improvement in treatment of spasticity and increased awareness of hip dislocation are unlikely to explain the effect of the surveillance programme. Another strength is that all pelvic radiographs in this study were evaluated by one investigator, since a marked interrater variation in measurements of the MP has been reported [25
]. Moreover, the primary clinical data on CP diagnosis, subtype, GMFCS and associated impairments were collected prospectively. However, in the areas providing regular care, data on hip status and surgery had to be collected retrospectively while this was done prospectively in the surveillance programme. Furthermore, in the surveillance area, more children (84%) had a recent (in the last two years) pelvic x-ray than children in the regular care area (73%); a reasonable effect of the programme (data not shown). Thus, it is theoretically possible that a case of hip dislocation or an operation might have been missed among patients offered regular care. This potential misclassification would, however, lead to an underestimation of hip dislocations and operations in the regular care areas. A potential limitation could be that the reviewer of the radiographs was not blinded to area. However, hip dislocations in the regular care area were originally diagnosed as completely dislocated by local clinicians, and no additional hips were deemed dislocated by the reviewer of the present study. Regarding the Swedish population, the radiographs were evaluated prospectively, and before the present study was planned. Since we used a migration percentage of 100 as a cut-off for hip dislocation, it is most unlikely that a complete dislocation had been overlooked.
Another potential limitation is the lack of complete data on 27% of children with CP in the seven Norwegian counties compared with nearly 100% coverage in Southern Sweden. The primary reason for the missing data in Norway is actually work overload on local doctors who were not able to complete the CP-registration forms. Only one family actively refused to participate in the CP-register, and one further family did not want to participate in the present study. If a more severely affected study population had been selected in Norway, this might have contributed to a higher proportion of hip dislocations. However, we have provided evidence that cases included in the Norwegian CP register are likely to be representative of the total CP population in a previous study [27
]. Moreover, the distribution of children within GMFCS levels III - V among all children with CP was nearly identical in the two areas.
In contrast to the similar distribution of GMFCS levels in the two study populations, a higher proportion of children were classified as dyskinetic in Sweden compared with Norway, whereas more children were classified as bilateral spastic CP in the Norwegian counties. Previous investigators showed a considerable variation in prevalence of the dyskinetic subtype among European countries and attributed that to differences in the classification of CP [20
]. However, dyskinetic CP also have increased risk of hip displacement [5
], and several studies have shown that GMFCS levels are more correlated to hip displacement than the topographic CP subtypes [10
]. Thus, taken together, misclassification or selection bias are unlikely to explain our results.
A potential confounder in this study is that the use of botulinum neurotoxin and baclofen was more common in Norway, since it could be expected to reduce the incidence of hip dislocation [30
]. However, if this was the case, it would have reduced the differences between the groups. On the other hand, one could speculate if this treatment had masked hip displacement by reducing spasticity and alleviating pain [32
We are not aware of other studies using the same approach to study the effectiveness of a hip surveillance programme to prevent hip dislocations. Our results are, however, consistent with the studies showing a reduction in hip dislocation in the total CP population from 8% to nearly 0% in Southern Sweden and elimination of the need for salvage operations in Australia after commencement of their screening programmes [6
Our results suggest that systematic clinical and radiological follow-up from an early age, as provided by the hip surveillance programme, lead to early detection of hip displacement signified by surgical intervention at a younger age. Nonetheless, the number of children operated on and the total number of operations was not higher in the surveillance compared with the regular care areas. This may be somewhat in contrast to the increase in preventive surgery which was reported in Australia after introduction of a similar screening programme [13
In the regular care area 23.5% of the children were treated with solely soft tissue release which might have been insufficient to prevent dislocation in some cases [39
]. In the surveillance area a higher proportion underwent femur osteotomy compared with the regular care areas, and also compared with the Australian surveillance programme. The latter difference is probably due to a longer observation time in Sweden [6
]. In Southern Sweden a standardised protocol is used for hip surgery as outlined in the CP follow-up programme (CPUP) with early soft tissue surgery followed by femur osteotomy in progressive cases [1
], and this most probably explains the variation in types of hip operations in the two study populations.
The difference in prevalence of hip dislocation between the two study populations could theoretically be due to an extraordinary high prevalence in the regular care areas, or an extremely low prevalence in the surveillance area, or both. However, the prevalence of hip dislocation among children with GMFCS III - V in the regular care area corresponds to an estimated prevalence of 5.1% in the total CP-population in this area. This is lower than the 8% reported in Sweden before the surveillance programme was introduced and the 14% reported by Scrutton [9
] from South East Thames. Thus, it is unlikely that the prevalence of hip dislocation is extraordinary high in the seven Norwegian counties compared to other areas without a surveillance programme. In contrast, the prevalence of 0.7% among children with GMFCS III - V in the surveillance area in this study is extremely low, and may not be obtained by other centres adopting the programme. Nonetheless, even a seven-fold higher prevalence (4.7% CI: 2.1 - 9.7) would be considerably lower than in the area providing regular care (15.1%; CI: 9.8 - 22.6). Thus, it is likely that the implementation of the surveillance programme in other populations will result in significant reduction in the proportion of children with hip dislocation.
Our results suggested that plain pelvic x-rays were taken more frequently in the surveillance area compared with the area providing regular care. Although we only had data from two counties in the regular care area, this might be a reasonable finding which could be considered a potential "downside" of the surveillance programme. On the other hand, our limited information suggested that the higher number of x-rays per child in the surveillance area partly was outweighed by an apparent need for a pelvic CT scan in a few patients in the regular care area.