Assessing the role of hip arthrography in the treatment decision making for children with Legg-Calvé-Perthes disease

Abstract

Objective

The aim of this study was to determine the role of hip arthrography in the treatment decision making for children with Legg-Calvé-Perthes disease (LCPD).

Methods

A total of 47 consecutive children with LCPD (42 boys, 5 girls; mean age=7.5 years; range=6–10 years) who underwent operative treatment were included in the study. The patient demographics, physical examination findings (pain and hip range of motion [ROM]), standard anteroposterior and Löwenstein lateral hip radiographs, and hip arthrography data were retrospectively collected. The arthrographies were performed immediately before the surgery under general anesthesia. The patients were staged according to the Catterall and Herring classifications and examined in terms of head-at-risk signs before the study. Four sets of patient files were established based on the aforementioned data, with each child in a randomized and blinded order. Ten consultant pediatric orthopedic surgeons randomly assessed the patient files on 4 separate occasions (Set 1 vs Set 2 and Set 3 vs Set 4), with a minimum time interval of 4 weeks. In the first and second sets, the demographic and clinical information, including the age, gender, hip ROM, and hip radiographs, were presented. In the third and fourth sets, hip arthrography was presented in addition to the data from Set 1 and Set 2. The observers were instructed to choose the best treatment options. The percent agreement (PA) and Gwet’s AC1 statistics were used to establish a relative level of agreement among the observers.

Results

The mean intra-observer reliabilities ranged from fair to moderate after adding the hip arthrography data (Gwet’s AC1 = 0.36 for Set 1 vs Set 2 and 0.42 for Set 3 vs Set 4). The mean PA was 56.6% (range = 29.8% to 78.7%) with a Gwet’s AC1 value of 0.51 (range: 0.21 to 0.77) between Set 1 and Set 3 (moderate intra-observer reliability). The decision for the treatment strategy was changed in 43.4% of the patients. For inter-observer reliability, Gwet’s AC1 was computed as 0.48 (moderate reliability). The correlation between the intra-observer reliability and stage progression was not significant (p>0.05) for any of the subgroups. Thus, there is a negative correlation with the disease progression.

Conclusion

Hip arthrography seems to have a significant role in the treatment decision making for children with LCPD, especially in the advanced stages of the disease.

Level of Evidence

Level IV, Therapeutic study

Introduction

The Legg-Calvé-Perthes Disease (LCPD) is a self-limiting, aseptic, idiopathic, and non-inflammatory avascular necrosis of the femoral head in which a lot of effort has been expended to figure out its ideal treatment algorithm, but it is still full of uncertainties (1–3). The main goals of the treatment are to prevent the deformation of the femoral head, to maintain its containment during the course of the disease, and to avoid subsequent premature degenerative osteoarthritis of the hip (4). In addition to the demographic information (in particular age and gender) and physical examination findings (hip range of motion [ROM], gait pattern, etc.), the radiological findings, including the stage of the disease, amount of necrosis, lateral column height of the head, and head-at-risk signs, are commonly used to guide the treatment in LCPD. The head-at-risk signs were first described by Catterall in 1971 to predict the prognosis of LCPD. These signs include Gage’s sign, calcification on the lateral part of epiphysis, lateral subluxation of the femoral head, and horizontal proximal femoral physis (5). In 1982, Smith et al. added metaphyseal cyst to the original risk signs (6). The anteroposterior (AP) and Löwenstein lateral (also known as frog-leg lateral view) hip radiographs, magnetic resonance imaging (MRI), and arthrography of the hip joint are frequently utilized imaging modalities used to obtain the aforementioned crucial radiological findings.

Hip arthrography is primarily used for the diagnosis of secondary changes such as labral pathologies, femoroacetabular impingement, and lesions like hinge abduction (7). The presence of the hinge abduction, first described by Grossbard in 1981 as a poor prognostic factor, can result in rapid progression to hip osteoarthritis (8, 9). Although previous studies proposed treatment algorithms based on the standard hip radiographs (10, 11), there is not much information on how the hip arthrography modifies the treatment decision. Some authors claimed that hip arthrography could be an important method for decision making (12–15). However, Devilia et al. stated that it is unimportant to perform an arthrography when the clinical findings and radiographs have provided sufficient information; however, in patients with advanced stage, when an osteotomy is planned, arthrography could be important to achieve a more congruent joint (16). Gallagher et al. reported that arthrography is not superior to the radiographs and added that it could be a superior method only for the individual patients for the diagnosis of the epiphyseal protrusion and for prognoses concerning arthritis (17). Thus, the research question of this study was whether hip arthrography changes the treatment modality when combined with the standard hip radiographs in LCPD. We hypothesized that the hip arthrography will change the initial decision based on the standard hip radiographs.

The purpose of this study was to evaluate the effect of additional hip arthrography imaging on decision making by using inter- and intra-observer agreements.

Materials and Methods

Forty-seven patients (42 boys, 5 girls; mean age: 7.5 [range: 6 to 10] years) with LCPD, who had completed clinical and radiographic follow-ups and were surgically treated in our institution, were included in this study. The study was conducted with the approval of the Institutional Ethics Committee (Number: 2017-2/13.07) and in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and its later amendments.

The patients’ demographic characteristics (age, gender), physical examination findings (pain and hip ROM), standard hip joint AP, Löwenstein lateral hip radiographs, and hip arthrography data were retrospectively obtained from the hospital’s medical records and picture archiving and communication system, PACS.

All imaging findings were jointly evaluated by three experienced pediatric orthopedic surgeons, and the patients were staged according to the Catterall and Herring classifications before the study. The surgeons also determined whether the patients had the head-at-risk signs. These surgeons did not participate in the study, and this information was not shared with the observers.

The hip radiographs and arthrography data of all patients were taken within a mean time interval of 2.3 (range: 1 to 4) days. The arthrography procedure was performed dynamically immediately before the surgery while the patient was under general anesthesia. The role of hip arthrography was assessed by both the patients’ clinical data and radiological findings. The imaging data were presented in at least two hip positions, particularly in the neutral AP and frog-leg positions.

Four sets of files were prepared in which the order of cases was randomized and blinded and were presented as slides. In the first and second sets, the clinical information, including age, sex, hip ROM, and standard AP and Löwenstein hip radiographs, were presented. In the third and fourth sets, the hip arthrography was presented in addition to the data from Set 1 and Set 2 (Table 1). The examples of the user interface typical of each case are shown in Figure 1.

Table 1.

Contents of the presentations in each set

	Demographic information and clinical findings	Standard hip radiographs	Hip arthrography
Set 1	Yes	Yes	No
Set 2	Yes	Yes	No
Set 3	Yes	Yes	Yes
Set 4	Yes	Yes	Yes

Figure 1.

Presentation of the clinical findings and radiographs of Case no: 1 from Set 1 and Case no: 16 from Set 3

Ten attending consultant pediatric orthopedic surgeons from different institutions, who had a surgical experience of at least five years in pediatric hip disorders, took part in the study. The assessments of the sets were performed in random order by each observer on 4 separate occasions (assessment of Set 1 vs Set 2 and Set 3 vs Set 4), after a minimum time interval of 4 weeks. The observers were instructed to choose the best treatment modality among the options presented in Table 2.

Table 2.

Treatment options presented to the observers

No.	Treatment modalities
1	Conservative treatment (observation, non-weight bearing, bracing, and/or traction)
2	Soft tissue releases (tenotomies)
3	Proximal femoral osteotomies (varus, valgus, derotational osteotomies)
4	Acetabular osteotomies (Salter’s, shelf, and triple osteotomies)
5	Combined femoral and acetabular osteotomies
6	Arthrodiastasis and/or shelf osteotomy
7	Safe dislocation and/or femoral head and neck reconstruction

First, in both the repeated presentations (Set 1 vs Set 2 and Set 3 vs Set 4), the intra-observer reliability was examined to investigate the stability of the choice of treatment in similar patients. Then, the number of times a different treatment modality was chosen for each set was calculated. The inter-observer reliability was subsequently computed by determining whether the response changed based on the radiography. To figure out the intra-observer reliability, Set 1 was compared with Set 3 for each observer separately in accordance to their choice from the 7 treatment options (Figure 2). All calculations were re-performed regarding the classification of the disease (i.e. Catterall, Herring or the presence of head-at-risk signs).

Figure 2.

Flow chart of the study to evaluate the mean intra-observer reliability

Statistical analysis

In recent literature, some statistical methods have been widely used to demonstrate the extent of agreement between two or more observers on nominally scaled data. Herein, the most important challenge faced is the paradoxes of kappa (18). As it is less affected by the prevalence and marginal probability and is a more stable agreement coefficient than kappa, Gwet’s AC1 statistic was chosen for this study (18–20). The percent agreement (PA) and Gwet’s AC1 statistics were used to establish a relative level of agreement among the observers for the assessments. The interpretation of the data was performed according to the Landis and Koch kappa benchmarks (<0: poor agreement; 0–0.20: slight agreement; 0.21–0.40: fair agreement; 0.41–0.60: moderate agreement; 0.61–0.80: substantial agreement; and 0.81–1.00: almost perfect agreement) (21).

The mean intra-observer reliability for each of the 10 observers showed normal distribution for each classification system and presence of head-at-risk signs (Shapiro-Wilk test; p>0.05). The correlation between the intra-observer reliability and stage progression was examined using the Pearson’s Correlation Coefficient.

Statistical analyses were performed using the R software package v.3.4.2 (The R Foundation, Vienna, Austria). The statistical significance level was set at p<0.05.

Results

The involvement was on the right side in 27 patients and on the left side in 20 patients. Of the patients, 6 (12.8%) were Catterall Stage II, 27 (57.4%) were Catterall Stage III, and 14 (29.8%) were Catterall Stage IV. According to the Herring’s classification, 15 (31.9%) patients were in Group B and 32 (68.1%) were in Group C. Twenty-three patients (48.9%) had the signs of head-at-risk.

In evaluating the repeated presentations, based on the sets that included hip joint radiographies, Gwet’s AC1 was found as 0.36 (fair) for the mean intra-observer reliability of the treatment choices. Conversely, according to the repeated presentations created by the addition of hip arthrography, the mean intra-observer reliability was calculated as 0.42 (moderate) (Figure 2).

The mean PA was found as 56.6% (range: 29.8% to 78.7%) with a Gwet’s AC1 value of 0.51 (range: 0.21 to 0.77) (moderate intra-observer reliability) (Table 3). It was observed that the treatment strategy was changed in 43.4% of the patients. For inter-observer reliability, Gwet’s AC1 was computed as 0.48 (moderate reliability). When the disease staging systems were taken into consideration, the agreement of the treatment choices was also reanalyzed to understand if there was a change according to the stage of the disease. The Gwet’s AC1 values for inter-observer reliability, mean intra-observer reliability, and PA are given in Table 4. The correlation between the intra-observer reliability and stage progression was not statistically significant (p>0.05) for each of the subgroups (Table 4).

Table 3.

Results of the intra-observer statistical analysis

	Observers										Mean
	1	2	3	4	5	6	7	8	9	10
Gwet’s AC1	0.26	0.74	0.37	0.75	0.55	0.52	0.77	0.65	0.29	0.21	0.51
PA (%)	36.2	76.6	44.7	76.6	59.6	57.5	78.7	68.1	38.3	29.8	56.6
PD (%)	63.8	23.4	55.3	23.4	40.4	42.6	21.3	31.9	61.7	70.2	43.4

PA: percent agreement; PD: percent disagreement

Table 4.

The results of inter- and intra-observer statistical analyses according to the classification of the disease. The values except inter-observer reliability were obtained by calculating the average intra-observer reliability of each of the 10 observers

	Catterall classification			Herring classification		Head-at-risk sign
	Stage II (n=6)	Stage III (n=27)	Stage IV (n=14)	Group B (n=15)	Group C (n=32)	Absent (n=24)	Present (n=23)
Inter-observer Gwet’s AC1	0.58	0.46	0.48	0.46	0.49	0.52	0.44
Mean intra-observer Gwet’s AC1	0.71	0.46	0.53	0.59	0.48	0.52	0.51
Mean PA (%)	73.3	52.2	57.9	63.3	53.4	56.7	56.5
Mean PD (%)	26.7	47.8	42.1	36.7	46.6	43.3	43.5
Pearson’s Correlation Coefficient	r:−0.303 p:0.104			r:−0.250 p:0.287		r:−0.004 p:0.985

PA: percent agreement; PD: percent disagreement

Discussion

The treatment of LCPD shows wide variations among the orthopedic surgeons, with numerous contra versions and no standard treatment protocol or algorithm being available (22). Besides the patient’s age and gender, hip ROM and findings obtained from the radiographs, MRI, and arthrography are used to determine the optimal treatment modality for each individual case (23). In fact, the hip arthrography is not a routine imaging method used in LCPD. Because it is an invasive technique, it is used only for high-risk patients, the cases in which poor prognosis is expected, and in those where treatment cannot be decided clearly. However, as per our knowledge, it is currently not known whether hip arthrography modifies the treatment choice or not.

The same physician can make different decisions at different times when evaluating the patients with similar demographic, clinical, and radiographic characteristics. In this respect, the present study was designed to evaluate the intra-observer reliability in repeated presentations. The mean intra-observer reliability obtained in the last two presentations (including hip arthrography) was higher than that obtained from the first two sets (not including arthrography). Although adding the arthrography data changed the clinical decision for 43.4% of the patients, it was seen that the physicians were more consistent in decision making (fair to moderate).

In this study, not only the inter-observer but also the intra-observer reliabilities were found to be moderate, with Gwet’s AC1 values of 0.48 and 0.51, respectively. The observers changed their treatment decisions in approximately 4 of 10 patients (PD: 43.4%). The second noteworthy finding of this study was that the agreement was worse in the advanced stages of the disease. When the analysis was performed after grouping the patients according to different clinical classification methods of LCPD, it was more likely that the low extent of involvement of the femoral head was substantial in deciding the type of treatment for a more stable condition (PA was 73.3% in Catterall Stage II, 63.3% in Herring Group B). It was interesting that we obtained closer values from the analysis regarding the head-at-risk signs. Among the subgroups, although there were negative correlations between the strength of the relationship and the progression of the disease stage, no statistical significance was detected because of the r-value being close to 0.00 and the p-value of >0.05 (Table 4).

In the literature, the role of hip arthrography in LCPD is controversial. Whereas some authors have asserted the importance of hip arthrography in the treatment selection, others have reported that it did not make a significant contribution to decision making (12–17, 24, 25). Axer and Schiller have emphasized that the arthrography examination provides more useful information, especially in protrusion cases, than the standard radiographs (25). Gallagher et al. reported that arthrography had no superiority over the radiographs (17). Devalia et al. reported that arthrography was a useful method in supporting their decisions, but because it did not give any additional information, it did not cause any change in the treatment decision in any LCPD patient (16). However, the results of our study clearly demonstrated that the hip arthrography is influential in decision making in LCPD, and it may change the planned treatment modality.

In this study, the gold standard for treatment in LCPD has not been questioned; rather, our study was focused on whether the hip arthrography changed the treatment modality in the light of the responses given by the observers in both the presentations. The arthrography images caused changes in 43.4% of the treatment strategies among all observers. Moreover, with the addition of arthrography imaging to the evaluation of patients, it was seen that the stability of the surgeon’s decision increased.

One of the limitations of this study was that most of our patients were in the advanced stage of LCPD, and complex patients were referred from other hospitals because our institution is a referral center for the pediatric hip disorders. Therefore, the decision-making process may be more complicated than usual, although all our observers were pediatric orthopedic surgeons. Second, only two static hip arthrography images were presented to the observers. However, hip arthrography is a dynamic examination under fluoroscopy, together with the assessment of hip ROM. Third limitation may be the possible variation in the surgical experience of the observers. The observers may have unintentionally chosen the treatment they were more experienced in, although they were instructed to choose the best treatment modality. Finally, owing to the exclusion of the patients in Catterall Stage I and Herring Group A and B/C border, the study has missing data. Thus, the statistical data obtained from the analyses in accordance with the classification systems may have caused questionable results. The lack of comparison of the use of arthrography and MRI in making the decision is an important limitation of our study.

In conclusion, the hip arthrography examination in addition to standard radiographs caused alteration in the treatment options, especially in the advanced stages of the LCPD. Therefore, we recommend hip arthrography whenever necessary, particularly for complex cases and patients in the advanced stage of the disease. Deciding whether hip arthrography is a useful method in decision making in LCPD is a dilemma. The comparison of the treatment outcomes provided with or without hip arthrography will demonstrate the benefits of this imaging method more clearly in future.

HIGHLIGHTS.

• Treatment of Legg-Calvé-Perthes disease is still controversial.

• Patients are generally treated according to the standard hip radiographic findings.

• Hip arthrography is an imaging method help in making the decision to treat complex cases and advanced stages of the disease.