Introduction

Osteonecrosis of the femoral head (ONFH) is one of the intractable diseases that usually affect young people. Most studies have shown that the prognosis of ONFH and the outcomes of surgical treatments are associated with the lesion size [1–8]. Hence, it is very significant to establish whether the lesion size of ONFH can change over time. However, this issue currently remains controversial. Although several reports have demonstrated that it is impossible for the lesion size to change by magnetic resonance imaging (MRI) evaluation during development of the disease without any operative treatments [3, 4, 8, 9], other studies have shown that reductions, and even complete resolution, or increments in the necrotic area can occur over several years [10–15].

The lesion size changes in some of the above-described studies have been evaluated in relatively few coronal or sagittal planes. To our knowledge, MR images of lesions are affected by the hip position during the MRI scan and are also influenced by different parameters such as slice thickness and interslice gap, thereby leading to possible mismatching of the slice planes among image sets and causing inaccuracies in the lesion size changes. Therefore, it is advisable to use volume measurements for evaluating lesion size changes owing to their precision [11].

Severe acute respiratory syndrome (SARS) is a relatively newly described infectious disease caused by the SARS coronavirus (SARS-CoV), which principally damages the cells of the immune system and pulmonary epithelium [16, 17]. Treatment with a steroid is required [18, 19]. For early detection of ONFH and further treatment, 426 SARS patients who had been treated with a corticosteroid between March and May 2003 and were mostly medical staff were screened between July 2003 and January 2004. A total of 87 patients with ONFH were diagnosed by MRI. These patients were detected at an early stage and followed up regularly at similar time intervals to provide the opportunity to study the lesion size changes by MRI.

Materials and methods

Patients

The Ethics Committee of our hospital approved the study, and all patients provided written informed consent. From March to May 2003, 426 medical staff contracted SARS and were treated with a corticosteroid. From September 2003 to January 2004, all the medical staff were screened by MRI and radiography organised by the Beijing Municipal Government, and 87 patients with ONFH were diagnosed by MRI. Subsequently, the patients with ONFH were followed annually and the latest follow-ups were carried out from October to November 2008. All of the patients underwent more than three MRI scans. Patients followed by MRI in both 2006 and 2008 and without surgical intervention were included in this study. A total of 33 patients (56 hips) who subsequently underwent surgery for discomfort during the course of the disease were excluded. Finally, 84 hips in 51 patients without smoking or alcohol abuse were included. The final patients comprised 17 men and 34 women with a mean age of 33.84 ± 9.15 years (range: 20–54 years). All 51 patients with ONFH were at the pre-collapse stage and without joint pain at the first MRI scan. According to the Association Research Circulation Osseous (ARCO) international staging system proposed in 1993 [20], 78 hips were at stage I and six hips were at stage II. The mean time from administration of the corticosteroid to the diagnosis of ONFH was 6.05 ± 1.64 months (range: 2.3–9.0 months). The mean time interval between the baseline and the latest MRI scan was 60.19 ± 1.71 months (range: 57–63.2 months). The mean time interval from the baseline to the scan in 2006 was 31.29 ± 1.85 months (range: 27.1–34.7 months). The mean time interval between the scans in 2006 and 2008 was 28.88 ± 0.67 months (range: 27.7–30.9 months, p = 0.000). The mean sum dosage of the prednisolone-equivalent steroid was 4,124.76 ± 2,277.10 mg (range: 800–11,000 mg) and the mean maximum daily dosage of the steroid was 251.76 ± 175.76 mg (range: 80–820 mg).

All of the patients were treated with hyperbaric oxygen therapy after the first examination. The treatment comprised six daily sessions per week up to a total of 100 sessions. A session involved breathing 100% oxygen at 2–2.4 absolute atmospheres in a multiplace pressure chamber for 90 min using a mask breathing system [21]. Intravenous injection of prostaglandin E1, ligustrazine and Salvia miltiorrhiza was also used for ten daily sessions per month up to a total of 30 sessions. The patients were fully weight-bearing following completion of the treatment.

MRI protocol

MRI was performed with two machines using at least two protocols, i.e. coronal T1-weighted imaging and coronal short τ inversion recovery (STIR). One machine was a Signa Excite 1.5 T Imager (GE Medical Systems, Milwaukee, WI, USA). Coronal STIR images were obtained using repetition time (TR) 2,560 ms and echo time (TE) 108 ms, while coronal T1-weighted sequences (TR 400/TE 8.6/Ef) were obtained using a pelvic phased array coil. Images of 4-mm thickness with 1-mm gaps and a 34 × 34-cm field of view were obtained using a 256 × 192 matrix and four NEX (number of excitations). The other machine was a Signa 0.5 T Imager (GE Medical Systems, Milwaukee, WI, USA). Coronal STIR images (TR 2,500/TE 80) and coronal T1-weighted sequences (TR 340/TE 15) of 5-mm thickness with 2-mm gaps were obtained.

MR image evaluation

ONFH was diagnosed using MRI by experienced radiologists and orthopaedic surgeons. MRI was diagnostic of osteonecrosis if crescentic subchondral areas demarcated by a band or ring of decreased signal intensity on T1-weighted images and increased signal intensity on the corresponding STIR images were observed [10].

Measurement of the volume of osteonecrosis

Measurement of the volume of osteonecrosis was carried out on coronal T1-weighted images. The area of osteonecrosis was considered to be the sector demarcated by the serpiginous line corresponding to the band-like hypointense margin. The outer border of this low-intensity region was assumed to represent the edge of the necrotic area. For each coronal slice, the area of osteonecrosis was outlined using an image analysis program (QWin, Leica, Heidelberg, Germany), and the volume of osteonecrosis was calculated by multiplying the area of osteonecrosis by the thickness (thickness of the slice plus the gap). The total volume of necrotic bone was the sum of the individual volumes of all slices [22].

The percentage changes in the lesion size from the baseline were calculated by dividing the differences in lesion volume between the baseline and the follow-up scans by the baseline lesion volume and multiplying by 100%. ONFH was arbitrarily estimated to be improved or worsened if there was a >15% change in the necrotic volume, while changes of ≤15% were considered to be “no change” [12].

Statistical analysis

The volume difference between each scan was examined using one-way analysis of variance (ANOVA). When the Levene test for homogeneity of variances was significant, the Mann-Whitney U test was used to further check the results. Pearson’s chi-square test or Fisher’s exact test was performed to analyse the correlations of the lesion size change with time, age, ARCO stage and lesion size. For all tests, values of p < 0.05 were considered to be significant. All statistical analyses were carried out with the Statistical Package for the Social Sciences (SPSS) software (version 13.0, SPSS Inc., Chicago, IL, USA).

Results

At the first MRI scan, 78 hips were at stage I and six hips were at stage II. Between April and July 2006, 19 hips were at stage I and 65 hips were at stage II. At the latest follow-up, six hips showed complete regression, ten hips were at stage I, 66 hips were at stage II and two hips had progressed to stage III.

At the first MRI scan, the mean volume of necrosis was 10.12 ± 8.05 cm3 (range: 0.39–41.62 cm3). In 2006, the mean volume of necrosis was 7.82 ± 7.59 cm3 (range: 0.11–39.65 cm3). In 2008, six hips showed complete regression, and the mean volume of necrosis was 5.67 ± 6.58 cm3 (range: 0.00–31.47 cm3). The volume changes between the first MRI scan (Vbaseline) and V2006 and between Vbaseline and V2008 were significant (p = 0.046 and p = 0.000, respectively; Fig. 1). The volume change between V2006 and V2008 was not significant (p = 0.062; Fig. 1).

Fig. 1
figure 1

Bar chart showing the mean lesion volumes at the three MRI scans (*p < 0.05)

Full size image

According to the arbitrarily defined criteria, four hips did not show a significant reduction in 2008, and their mean lesion size at the first scan of 20.37 ± 7.41 cm3 was larger than that of hips that showed a reduction (p = 0.008). The average volume of the six hips that showed complete regression was 2.04 ± 1.57 cm3 (range: 0.39–4.69 cm3) at the first scan and relatively smaller (p = 0.000; Fig. 2).

Fig. 2
figure 2

Complete regression of a necrotic lesion. a Coronal T1-weighted MR image at the first MRI scan. b Corresponding STIR image at the first MRI scan. c Corresponding coronal T1-weighted MR image to that shown in a performed in April 2006, showing a reduction of the lesion size. d Corresponding coronal T1-weighted MR image to that shown in a performed in October 2008, showing complete regression

Full size image

During the course of the follow-up, six patients complained of hip discomfort, although four of these patients showed lesion reductions. Their mean lesion volume at the first MRI scan was 23.97 ± 7.71 cm3 (range: 18.32–39.22 cm3) and relatively larger than that of patients without symptoms (p = 0.000).

The lesion size changes between two intervals were evaluated. Compared with the baseline scan, 26 hips did not change and 58 hips improved in 2006. Compared with the scan in 2006, 20 hips showed no change and 64 hips improved in 2008 (p = 0.299).

Among the 26 hips that showed no change in 2006, six did not change and 20 improved in 2008. Among the 58 hips that improved in 2006, 14 did not change and 44 improved in 2008 (p = 0.916).

According to the baseline scans, we classified the hips into two groups, one containing larger lesions with necrotic volumes of ≥15 cm3 and the other containing smaller lesions with necrotic volumes of <15 cm3. In the larger lesion group, three hips showed no change and 15 hips improved, while in the smaller lesion group, one hip did not change and 65 hips improved in 2008 (p = 0.029).

Patients were classified into age groups of >40 years and ≤40 years. Overall, two hips showed no change and 21 hips improved in the >40 years group, while two hips showed no change and 59 hips improved in the ≤40 years group in 2008 (p = 0.301).

There were 27 hips in the male group and 57 hips in the female group. All of the male hips improved, while four female hips showed no change and 52 female hips improved (p = 0.308).

There were six hips at stage II and 78 hips at stage I at the first MRI scan. All of the stage II hips improved in 2008, while four stage I hips showed no change and 74 stage I hips improved (p = 1.000).

Discussion

It is appropriate to establish whether ONFH can spontaneously resolve. Some researchers have reported that the lesion size is unlikely to change during the natural course of the disease [3, 4, 8, 9]. However, other researchers have observed reductions or even complete regression of lesions at the early stage of ONFH over the course of a few years [10, 13, 14]. There are some shortcomings in these studies. One is that the lesion size changes were evaluated on relatively few coronal or sagittal planes, in which the MR images can be influenced by the patient’s position and the MRI protocols, and no volume measurements were carried out. Another is that the follow-up intervals and the numbers of necrotic femoral heads included in these studies were relatively limited.

Yoshida et al. [12] calculated the necrotic volumes of 24 hips showing early-stage ONFH in 13 patients with systemic lupus erythematosus. They reported that hips with a necrotic volume of <25% showed improvement, while the mean volume of necrosis did not change in type C. Takao et al. [11] observed 31 hips with ONFH in 25 patients and found that lesions detected at less than one year after initial steroid treatment can show size reductions using image registration as well as volume measurements.

Compared with these studies, there are some advantages in our study. First, ONFH was detected at a relatively early stage. The mean time was 6.05 ± 1.64 months (range: 2.3–9.0 months) from administration of the corticosteroid to the diagnosis of ONFH. Most of our patients were at stage I. Second, the follow-up periods with MRI were relatively similar. The mean interval from the baseline to the latest scan was 60.19 ± 1.71 months (range: 57–63.2 months). Third, our patients were generally healthy before suffering the disease and did not receive any further corticosteroid treatment.

The mean lesion volumes of ONFH were 10.12 ± 8.05 cm3 at the baseline, 7.82 ± 7.59 cm3 in 2006 and 5.67 ± 6.58 cm3 in 2008, and the overall decrease was significant. The volume changes between the first MRI scan (Vbaseline) and V2006 and between Vbaseline and V2008 were also significant. Although the volume change between V2006 and V2008 was not significant, the first interval was a little longer than the second interval. According to our arbitrary standard, nearly all of the lesions showed reductions and some lesions still showed reductions after 2.5 years. Therefore, there was a strong tendency for the reduction in lesion size to be time dependent. Another observed phenomenon was that the reduction was not continuous. Some lesions that showed reductions in 2006 did not reduce in 2008 and vice versa.

The ratio of the lesion size changes was lower in hips with relatively larger lesion sizes compared with hips with relatively smaller lesion sizes. One reason for this finding may be that the interface between the necrotic and normal areas is relatively smaller than the lesion size. Another reason may be that the reparative capability is limited.

There are some diagnostic criteria for ONFH, but it is hard to fulfill these criteria at the early stage (pre-radiological stage) [20, 23, 24]. If the diagnosis of ONFH is made by abnormal signals on MRI, there may be misdiagnoses [25]. During our follow-up study in SARS patients, there was an interesting finding that a few radiologically normal hips showed abnormal signals, such as band formation or decreased signal areas, on T1-weighted images and normal signals on STIR images. On the follow-up MRI, most of these abnormal signals disappeared on T1-weighted images and still showed a normal presentation on STIR images, while a few new cases occurred. We classified these hips into a non-necrotic group (Fig. 3). Since STIR is more sensitive than other protocols for detecting marrow abnormalities [26], we considered that the band or ring-like zones with decreased signal intensity on T1-weighted images and increased signal intensity on the corresponding STIR images could be used as diagnostic criteria for ONFH in the early stage. According to these diagnostic criteria, there were six hips that showed complete regression of their lesions in our study. Their initial volumes were relatively smaller. The intervals to complete resolution were >29 months.

Fig. 3
figure 3

Non-necrotic femoral head with abnormal band formation. a Coronal T1-weighted MR image at the first MRI scan, showing irregular band formation. b Corresponding STIR image at the first MRI scan, showing no abnormal signals. c Corresponding coronal T1-weighted MR image to that shown in a performed 6 months later, showing no abnormal signals

Full size image

A total of six hips with relatively larger volumes presented discomfort during the course of the disease. Among these hips, two progressed to collapse, while four hips showed lesion reductions. These results coincided with other studies in which larger lesions were found to be more likely to cause aggravation and discomfort, and therefore early surgical intervention may be justified for these patients [1–12, 27, 28].

There were some biases in our study. One was that the patients who underwent surgical intervention for discomfort were excluded from the study. Another was that all of the patients were treated with hyperbaric oxygen therapy and administration of prostaglandin E1, ligustrazine and Salvia miltiorrhiza, which were assumed to improve the blood supply and increase bone formation and angiogenesis in necrotic lesions after the first examination [21, 29, 30]. Therefore, our results cannot be considered to reflect the natural history of ONFH. Although a few studies have reported that non-surgical treatment can reduce the lesion size, their proposals are not generally accepted [20]. The purpose of this study was simply to ascertain whether lesion reduction can occur without surgical intervention. Advanced studies are required to investigate the correlations of the lesion size changes with non-operative treatment and collapse.