The local Research Ethics Committee approved this study (REC 14/LO/1267). We obtained written informed consent from all participants. Eligible for this study were patients with a diagnosis of DDH who had received a closed or open reduction with or without osteotomy and who were older than 14 years of age at the time of study assessment. An attempt was made to include as many patients as possible from our original study [12] as this would allow for inferences to be made about how patients change over time. One researcher (A.M.) used clinical coding and the database from our previous study [12] to identify eligible patients treated in two tertiary centres from 1995 to 2005. We excluded patients with co-morbidities that exclude the diagnosis of DDH.
Of 311 eligible patients identified, 160 (51%) had evidence of osteonecrosis as per clinical records and radiography (Fig. 1). These included 72 patients studied in 2011 [12] when we had measured their hip function, physical function, and health status.
Of 160 patients, 23 could not be recruited (Fig. 1). In 18 of the remaining patients we could not ascertain the effects of the osteonecrosis on patient-reported outcomes as 16 had undergone hip arthroplasties and 2 had undergone hip arthrodesis. Thus, 117 patients (149 hips) with DDH and osteonecrosis at a mean age of 19.6 ± 3.8 years completed patient reported outcomes to measure the effects of osteonecrosis. These included 54/72 patients (75%) who had taken part in our earlier study [12] and who we could re-examine after a mean period (and standard deviation) of 8.4 ± 0.7 years or at a mean patient age (and standard deviation) of 21.9 ± 2.6 years (of the remaining patients, five had undergone total hip replacements; one patient lived abroad; one was pregnant; one had a mental health condition preventing participation; five could not be contacted; and five declined participation). From 151 patients with DDH but without osteonecrosis, we recruited an age-matched sample of 32 patients (37 hips). In total, we studied 149 patients (186 hips) at a mean age (and standard deviation) of 19.6 ± 3.8 (range, 14 to 26) years (Table I).
Table I. Group differences based on univariate analysis
Variable | DDH with osteonecrosis 149 hips (117 patients) | DDH without osteonecrosis 37 hips (32 patients) | p-value |
Age at study assessment (years)* | 19.3 ± 3.8 | 19.6 ± 3.9 | 0.95 |
Female | 102 (86%) | 30 (97%) | 0.18 |
Left hip affected | 87 (57%) | 16 (46%) | 0.21 |
Centre-edge angle (degrees) * | 18.2 ± 4.4 | 27.3 ± 9.9 | < 0.01 |
Sharp angle (degrees) * | 44.8 ± 6.3 | 41.9 ± 5.5 | 0.04 |
Kellgren-Lawrence grade† | 1 (1, 2) | 0 (0, 1) | < 0.01 |
Number of all operations† | 2 (1, 3) | 1 (1, 2) | 0.09 |
One (%) Two (%) Three or more (%) | 54 (36) 53 (36) 41 (28) | 22 (59) 11 (29) 4 (11) | 0.03 0.2 < 0.01 |
* values given as mean ± standard deviation
† values given as median and inter-quartile range
All participants had a standing antero-posterior radiograph of the pelvis at the time of study assessment. We employed a standard protocol [13] using a digital imaging system (GE Medical Systems Ltd, Buckinghamshire, UK). We graded the presence of osteonecrosis according to Bucholz-Ogden [11]. In grade I the femoral head shows hypoplasia compared to the unaffected side. In grade II, the lateral growth plate is damaged, resulting in a valgus deformity. Global damage of the physis is said to underlay grade III, resulting in marked shortening of the femoral neck and marked trochanteric overgrowth. Damage along the medial aspect of the physis causes the varus alignment of the upper femur seen in grade IV. Grade I we found in 7 patients (10 hips); grade II was seen in 77 patients (93 hips); 18 patients (26 hips) had grade III changes, and 15 patients (20 hips) had changes of grade IV (Fig. 1). We quantified acetabular dysplasia by means of the centre-edge angle of Wiberg [14] and the acetabular angle of Sharp [15]. We evaluated the presence of osteoarthrosis according to Kellgren-Lawrence [16].
An orthopaedic resident (A.M.) and an orthopaedic fellow (D.M.) analysed all radiographs electronically (Centricity Enterprise Web V3.0. 2006 GE Medical) and in random order, blinded to patient identifiers and clinical variables.
They first reviewed all radiographic classifications schemes and agreed on definitions and landmarks. They then evaluated all radiographs independently and their inter-rater reliability was established. For Sharp and centre edge angles, the interrater reliability was excellent [17] (intra-class correlation coefficient = 0.86); it was moderate [17] for the Kellgren-Lawrence (κ = 0.62) and Bucholz-Ogden (κ = 0.64) classifications. For the latter two indices, the two observers reviewed all radiographs in consensus to establish final grade. Radiographs also were graded separately by the senior author (A.R.) for the presence of osteonecrosis and we resolved any disagreements in consensus.
The orthopaedic resident (A.M.) examined all patients according to the Children’s Hospital Oakland Hip Evaluation Scale (CHOHES) [18], a valid and reliable hip-specific assessment measure with three domains: pain, hip function, and physical examination. A maximum score of 100 points indicates best hip function. We presented the following patient-reported outcome measures to patients in random order to control for an order effect at the group level of analysis:
Activity Scales for Kids (ASK)
This 30-item valid and reliable tool measures physical function [19]. Its maximum score is 100 points, indicating unlimited physical functioning. The ASK is intended for patients up to 17 years of age [20] and in those 17 years or older, we used the Hip Disability and Osteoarthritis Outcomes Score Physical Function Shortform (HOOS-PS) that also measures physical function [21]. HOOS-PS elicits activity-related symptoms that patients experience due to a hip pathologies. As with the ASK, this measure fits a unidimensional, interval scaled model [22] with a maximum score of 100 indicating unlimited physical functioning.
Health Utilities Index Mark 3 (HUI-3)
HUI-3 is a questionnaire-based method for measuring general health status and health-related quality of life [23]. With its 8 attributes (vision, hearing, speech, ambulation, dexterity, emotion, cognition, pain). HUI-3 scores can be converted into utilities [24]. Utility is defined as the strength of an individual preference for a health state measured under conditions of uncertainty, expressed on a continuous scale from 0 to 1, with 0 representing death and 1 representing perfect health [25]. We used a standard 15-item, English-language version for self-administered, self-assessed two-week health-status assessment.
Statistical Methods
We summarised scores and patient characteristics with means and standard deviations, or medians and inter-quartile range in non-normally distributed data. We determined the relationship between osteonecrosis and hip function, physical function and health status with linear mixed-effects regression models [26] in order to account for within-subject correlation among 37 patients with bilateral osteonecrosis. We decided a priori to adjust all analyses for the total number of operations (with the exception of implant removal) any hip had undergone prior to the study assessment [12]; the degree of acetabular dysplasia at the most recent radiograph [27]; and the age at study assessment [12]. We fitted models for each outcome measure (hip function, physical function, health status) using a backwards stepwise approach [28]. We used Akaike’s Information Criterion (AIC) [29] to assess goodness of fit. We reported least squared means for adjusted outcome scores. We determined the interrater reliability using Cohen’s weighted κ [30] or the intraclass correlation coefficient model 2 [31], respectively. We estimated the sample size according to Cohen [32] based on the primary outcome, hip function. Established CHOHES scores [12] of 88, 88, 80 and 78 for Bucholz grades I-IV respectively (SD = 10) gave effect sizes (Cohen’s d) between 0.2 and 0.8. With α = 0.5 and β = 0.20, we estimated at least 15 patients were needed for each Bucholz-Ogden grade examined. In order to adjust for three variables, at least 105 patients with osteonecrosis were required (15 further patients for each additional variable) [33]. We used the R Language and Environment for Statistical Computing, version 3.1[34] statistical package.