Most of the literature regarding the natural history of nontraumatic osteonecrosis of the femoral head has concerned symptomatic osteonecrosis1-6. Symptomatic osteonecrosis of the femoral head that is not treated generally progresses to collapse, and eventually painful osteoarthritis develops7-11. However, the fate of untreated asymptomatic osteonecrosis of the femoral head remains unclear.
With the advent of magnetic resonance imaging, it is possible to make an early diagnosis of osteonecrosis before radiographic evidence appears. In particular, magnetic resonance imaging has made it easier to detect asymptomatic lesions in patients with osteonecrosis of the femoral head because most asymptomatic disease is not apparent radiographically11.
Many variables, such as disease stage, lesion size or location, the presence of pain, and radiographic evidence of collapse, have been evaluated as factors in the prognosis of osteonecrosis of the femoral head, and, of these, the extent of the necrotic lesion has been considered the most important11-18.
In the present study, we investigated the fate of untreated asymptomatic osteonecrosis of the femoral head on the basis of the hypothesis that asymptomatic lesions that were smaller than a certain size would not progress to symptomatic disease without treatment.
From January 1982 to March 2005, 312 patients with bilateral nontraumatic osteonecrosis of the femoral head presented to our department with pain in only one hip. The diagnosis of nontraumatic osteonecrosis of the femoral head was made on the basis of typical findings on plain radiographs and/or magnetic resonance images. Patients in whom osteonecrosis had developed after a femoral neck fracture or a hip dislocation were excluded. Of the 312 patients with an asymptomatic hip, 128 were treated with a joint-preserving procedure such as core decompression with or without bone-grafting, multiple drilling, or transtrochanteric rotational osteotomy. The remaining 184 were followed without any treatment for the asymptomatic hip. Included in this group were patients in whom no joint-preserving operation was recommended because they were older than fifty-five years of age as well as patients who were younger than fifty-five years of age and were advised to have a joint-preserving operation but refused it. Seventy-nine hips that remained asymptomatic but were followed for less than five years were excluded, leaving 105 hips to be included in the final analysis. All of those hips were followed for at least five years or until pain developed. Informed consent was obtained from all patients for participation in the study, which was approved by the institutional review board at our institute.
There were eighty-four men and twenty-one women with a mean age of fifty years (range, twenty-three to seventy-three years) at the time of the diagnosis. According to the grading system described by Steinberg et al.15, sixty-two hips (59%) were in stage I (normal findings on radiographs but abnormal findings on magnetic resonance images) and forty-three hips were in stage II (radiographs showing scleroses or cysts but no crescent sign) at the time of diagnosis. The mean duration of follow-up was seven years and eight months (range, ten months to twenty-two years and eight months). The risk factors associated with osteonecrosis were alcohol abuse (fifty-four hips) and corticosteroid intake (twenty hips). Thirty-one hips were considered to have idiopathic osteonecrosis.
The extent of the necrotic lesions was measured on magnetic resonance images when available (seventy-eight hips) or on plain radiographs (twenty-seven hips) at the time of the diagnosis. Details of the method of measurement and its accuracy and repeatability have been previously described19. In brief, we measured the largest mediolateral diameter of the femoral head (R) and the longest mediolateral length of the necrotic lesion (A) on coronal magnetic resonance images or on anteroposterior plain radiographs, and the largest anteroposterior diameter of the femoral head (R') and the longest anteroposterior length of the necrotic lesion (B) on sagittal magnetic resonance images or on frog-leg lateral plain radiographs. The two-dimensional extent of a necrotic lesion was determined with the equation: % area = (A × B/R × R') × 100 (Figs. 1-A and 1-B).
The location of a necrotic lesion was classified as Type A (less than the medial one-third of the weight-bearing area, medial location), B (between the medial one-third and two-thirds of the weight-bearing area, central location), and C (greater than the medial two-thirds of the weight-bearing area, lateral location) with use of the grading system described by Nishii et al.5 on the midcoronal T1-weighted magnetic resonance images (in seventy-eight hips) or on the anteroposterior radiographs (in twenty-seven hips). Twenty-two hips were classified as Type A; twenty-eight, as Type B; and fifty-five, as Type C.
Demographic and clinical differences between hips that remained painless and those that became painful were assessed.
Statistical Analysis
Nominal data were analyzed with use of the chi-square test and continuous data, with the Student t test.
Three subgroups were defined according to the size of the necrotic lesion: hips with necrosis of <30% of the area of the femoral head, those with necrosis between 30% and 50% of the area, and hips with necrosis of >50% of the area. Survival curves for these three groups were calculated with use of the Kaplan-Meier method, with the development of pain as the end point, and the time until pain developed was compared among the three groups with use of the log-rank test. The same analyses were done for the three types of lesion location.
We performed two different regression analyses. Linear regression analysis (a method used to analyze the relationship between two variables, X and Y) was conducted to determine the relationship between the extent of the necrosis and the time until pain developed in the hips that became painful. Multivariate analyses were performed with use of the Cox proportional-hazards model (a method used to investigate the effects of the predictor variables on the time that an event takes to happen) to identify independent factors with regard to disease progression, defined as the development of pain. These analyses were adjusted for age, gender, body mass index, association with alcohol abuse and corticosteroid intake, and initial stage. The level of significance was set at p < 0.05.
The mean extent of the 105 necrotic lesions was 52.9% (range, 2% to 100%) of the area of the femoral head. Sixty-two hips (59%) became painful and collapsed during the follow-up period, and forty-three hips (41%) remained asymptomatic and without collapse for more than five years (mean, eight years and seven months; range, five to fifteen years).
Of the forty-three hips that remained asymptomatic, thirteen that had been in stage I progressed to stage II; however, eleven hips in stage I and nineteen hips in stage II remained in the same stage without evidence of femoral head collapse at the time of the latest follow-up. All of the sixty-two hips that became painful had progressed to femoral head collapse: eight hips progressed from stage I to III; twenty-six, from stage I to IV; three, from stage I to V; one, from stage I to VI; three, from stage II to III; seventeen, from stage II to IV; two from stage II to V; and two, from stage II to VI. The development of pain had a significant correlation with the occurrence of femoral head collapse (p < 0.0001).
The mean interval between the initial diagnosis and the development of pain was two years and one month (range, one month to eleven years). In the majority (fifty-eight) of sixty-two symptomatic hips, pain developed within five years. The pain developed within one year in twenty-five hips (40%), within two years in forty-one (66%), within three years in forty-seven (76%), within four years in fifty-two (84%), within five years in fifty-eight (94%), and within ten years in sixty-one (98%). (The numbers of hips are cumulative.) A smaller extent of necrosis was found to be significantly associated with a longer time between the diagnosis and the development of pain (p = 0.008, time to pain [in months] = 58.3 — 0.499 × % area of necrosis).
With the numbers studied, age, gender, and body mass index were not found to be significantly associated with the development of pain (p > 0.30 for all, Table I). Cases related to alcohol abuse became symptomatic more often than did those related to corticosteroid intake or idiopathic causes. However, multivariate analysis showed no significant relationship, with the numbers available, between these risk factors and the development of pain after adjustment for other variables, including age, gender, body mass index, initial stage, lesion extent, and lesion location (p = 0.43). No significant difference was found between stage-I hips and stage-II hips with regard to the development of pain (p = 0.58).
The extent of the necrotic lesion was found to be significantly greater in the hips that became symptomatic (p < 0.0001), and adjusted multivariate analyses showed it to be a significant independent predictor of disease progression (defined as the development of pain) (p < 0.0001, Table II). An increase in the extent of osteonecrosis by 1% was found to be associated with a 1.029-fold increase in the risk of pain developing (95% confidence interval, 1.014 to 1.045).
The lesion location was found to be significantly associated with the development of pain (p = 0.001), but it was not found to be a significant independent predictor of disease progression in the adjusted multivariate analysis (p = 0.3). Pain developed in six of the twenty-two Type-A hips, fifteen of the twenty-eight Type-B hips, and forty-one of the fifty-five Type-C hips. With the development of pain as the end point, a comparison of survival curves showed no significant differences among the three types (p = 0.23).
Of the sixty-two hips in which pain developed, forty-six (74%) deteriorated sufficiently to require surgery and forty-four of these were treated operatively; bipolar arthroplasty was performed in three hips and total hip arthroplasty, in forty-one hips. The remaining two patients were awaiting surgery at the time of their latest follow-up. The average time between the onset of pain and surgery was one year and seven months (range, one month to ten years and ten months). Sixteen of the sixty-two hips in which pain developed remained untreated at the time of final follow-up, as the pain was not severe enough to warrant surgery. A significant difference in the mean lesion size was noted between the sixteen untreated hips (53.3% of the area of the femoral head) and the forty-six hips that required surgery (70.9% of the area) (p < 0.0001).
Of the twenty-one hips with a small lesion (<30% of the area of the femoral head), one became symptomatic; of the twenty-four hips with a medium-sized lesion (30% to 50% of the area), eleven became symptomatic; and of the sixty hips with a large lesion (>50% of the area), fifty became symptomatic. With the development of pain as the end point, a comparison of survival curves showed that hips with a small lesion had a significantly longer survival duration than those with a medium-sized or large lesion (p = 0.0096 and 0.0002, respectively) (Figs. 2-A, 2-B, and 2-C). The one hip with a small lesion that became painful had a far-lateral lesion, and the ten large lesion hips that remained painless had a central lesion with normal bone between the necrotic lesion and the articular cartilage (that is, the subchondral bone was spared) (Figs. 3-A, 3-B, and 3-C).
Previous studies of asymptomatic osteonecrosis of the femoral head have demonstrated various rates of disease progression5,20-28. Differing opinions regarding the fate of asymptomatic disease have been attributed to several reasons, including differences in study populations, diagnostic modalities, follow-up periods, and the definition of disease progression5,27. In the present study, 105 asymptomatic hips with early stages of disease were not treated. Sixty-two hips were followed until pain developed, and the other forty-three remained painless without collapse for five years or more. Therefore, the overall rate of progression of asymptomatic osteonecrosis in this study was 59%. These results are comparable with the overall rate of 66% (317 of 480) in the previous reports5,20-28 describing progression of asymptomatic disease. In our study, pain developed in most hips (83%; fifty of sixty) in which the lesion involved >50% of the femoral head, usually within three years (76%; forty-seven of sixty-two hips).
The cohort of this study did not comprise all patients with asymptomatic osteonecrosis of the femoral head. Patients younger than fifty-five years of age were included only when they had refused joint-preserving treatment. To minimize the effect of this selection, we recommended joint-preserving treatment to all patients younger than fifty-five years of age, regardless of the size of the necrotic lesion.
The extent of a necrotic lesion is an important determinant of prognosis11-18. The present study showed that, among the various factors evaluated, the extent of the necrotic lesion was the most important for predicting the prognosis of asymptomatic osteonecrosis of the femoral head. In this study, the rate of disease progression, defined as the development of pain, was 5% for small necrotic lesions (<30% of the area of the femoral head), 46% for medium-sized necrotic lesions (30% to 50% of the area), and 83% for large necrotic lesions (>50% of the area). While Hungerford and Jones11 and Steinberg et al.15 defined small lesions as <15% of volume and large lesions as >30% of volume, we defined small lesions as <30% of area and large lesions as >50% of area. We used percent of area because the measuring method was simple.
The results of this study suggest that the development of pain is related to the occurrence of femoral head collapse, a finding consistent with those reported by Nishii et al.5. In addition, we found that the extent of the lesion was related to the time that it took for pain to develop and the need for surgery. Pain developed more slowly and rarely progressed enough to require surgery in hips with smaller lesions.
We found that age, gender, body mass index, selected risk factors associated with osteonecrosis, and initial radiographic stage were not associated with the development of pain. These results are similar to those of earlier studies regarding the natural course of early-stage disease, whether asymptomatic or not5,12,18,24,25. In the present study, the extent of the lesion appeared to be more important than the initial radiographic stage as a factor in the fate of osteonecrosis of the femoral head, particularly in its earlier stages, an observation that is supported by the literature on the topic11,13,16,17,25.
Exceptional cases, in which a small lesion became painful or a large lesion remained painless, suggest that the location of the lesion is another important prognostic factor. In this study, lesion location was closely related to the development of pain, but it was not a significant prognostic factor for disease progression. This may be due to the fact that larger lesions usually extended more laterally in the femoral head.
In the current study, one small necrotic lesion that became painful was located far laterally in the femoral head. Nishii et al.5,18 reported a poor prognosis for lesions in this location despite a small size. Moreover, the ten large necrotic lesions that remained painless in our study were located in the central portion of the femoral head, and a substantial amount of subchondral bone above these lesions had been spared. This finding is consistent with reports by Ohzono et al.2 and Ito et al.25, who also noted that centrally located lesions rarely progressed to collapse despite a large size.
Many authors have reported that small necrotic lesions remained asymptomatic for several years5,23-25,27. The results of the present study strongly suggest that asymptomatic necrotic lesions involving <30% of the femoral head typically have a benign course for more than five years and do not need any treatment unless they are located far laterally. When pain developed in our series, it almost always did so within five years after the diagnosis. We are of the opinion that our duration of follow-up of small lesions that remained painless was adequate as it ranged from five to fifteen years and averaged seven years and ten months. Hernigou et al., in two separate studies26,28, reported that the majority of small asymptomatic osteonecrotic lesions became symptomatic; 88% of forty hips became symptomatic within a mean of six years and eight months after the diagnosis and 91% of 121 hips did so within three years after the diagnosis. The reason for this difference between their results and ours is uncertain but may be related to different study populations and different durations of follow-up. A larger proportion of the patients in one of their studies26 (eleven of forty compared with twenty of 105 in our study) had steroid-related disease and were followed for a mean of eleven years, and, in their other study28, all patients had sickle cell disease and were followed for a mean of fourteen years.
We intend to continue observing the patients who were enrolled in the present study closely. Nevertheless, we believe that no treatment is necessary for asymptomatic necrotic lesions when they involve <30% of the area of the femoral head, as calculated with our method. 