Avascular necrosis in HIV patients: a case–control study

Introduction

As human immunodeficiency virus (HIV)-related mortality has decreased since the introduction of combination antiretroviral therapy (cART), focus is switching to the morbidity associated with long-term HIV disease and exposure to antiretroviral drugs.^1,2 A growing literature suggests that in the HIV-infected population, the incidence of avascular necrosis (AVN) of bone (also known as osteonecrosis), is 40 times higher than in the general population.^3,4 The reason for this remains unclear. Established risk factors for AVN in individuals without HIV (such as alcohol abuse, corticosteroid use, coagulopathies) may be more prevalent in those with HIV.^4–11 Moreover, it has been suggested that HIV infection itself and antiretroviral therapy (ART), particularly protease inhibitors (PIs), ¹² may increase AVN risk. This study investigates the association of AVN in HIV-positive individuals with possible risk factors including ART.

Methods

We retrospectively reviewed the clinic records of all cases of AVN diagnosed in HIV-positive patients up to July 2009 attending North Middlesex University Hospital, Enfield, London, UK. An AVN case was defined as an individual with HIV infection who had clinical symptoms of, and subsequent radiological evidence of AVN of a joint. Control individuals had no record of clinical symptoms suggestive of AVN and were matched on the following four criteria: gender, age (±2 years), nadir CD4 count by stratum (≤200 or >200 cells/mL) and date of HIV diagnosis (±2 years). For each AVN case, we used a pseudo random number generator (Stata, StataCorp LP, College Station, TX), to select a control from the subset of all HIV-positive patients attending North Middlesex University Hospital who fulfilled the four matching criteria. Clinical and biological data were obtained from the computer database for case and control individuals and verified using written clinical records. For AVN cases, all subsequent imaging was reviewed for evidence of development of AVN at other joints after the initial diagnosis of AVN. Use of corticosteroids was obtained from the clinical notes and pharmacy records. In consonance with the methodology set out in Fessel et al., ¹³ we converted corticosteroid doses to an equivalent prednisolone dose and examined the use of systemic corticosteroids by using two measures. First, the aggregate equivalent dose of prednisolone was taken as milligrams per month of the time between the date of the first dose of corticosteroids and the date of diagnosis of AVN. Second, to look at more recent exposure, we also measured the aggregate prednisolone equivalent dose taken in the 12 months preceding the diagnosis of AVN.

A sample size of 15 cases and 15 controls (unmatched) provides 80% power to detect a binary risk factor as significant if its prevalence is 22% in the population of controls and 78% in cases, and it was hoped that matching would maintain power. Associations with AVN were tested using the exact McNemar’s test for binary potential risk factors and Wilcoxon matched-pairs signed-rank test for continuous factors. A two-tailed p value of 0.05 was considered significant. Analysis was performed using STATA software, version 10 (Stata, StataCorp LP, College Station, TX).

Results

Of 15 symptomatic AVN cases identified, eight were in women (Table 1). Osteonecrosis was confirmed by magnetic resonance imaging (MRI) in eight (53%) cases and radiograph alone in seven (47%) cases. Ten cases were in people of black African origin. The median age of cases at diagnosis of AVN was 39 years (interquartile range [IQR]: 29–44 years). The median nadir CD4 count in cases was 80 cell/mL (IQR: 30–170 cells/μL). The median time since diagnosis of HIV infection to AVN was 2.4 years (IQR: 0.7–4.4 years). The most common joint first affected was the femoral head (80%). Eight (53%) individuals subsequently developed AVN at sites other than that of their initial diagnosis. Six (40%) patients developed AVN at three or more joints.

OPEN IN VIEWER

Table 1.

Clinical characteristics of HIV-positive subjects with avascular necrosis (AVN).

No.	Gender	CD4 nadir (cells/μL)	HIV duration at AVN diagnosis (years)	Sites of initial AVN	Subsequent sites of AVN
1	M	30	2.8	Bilateral femoral heads	Bilateral ankle
2	F	260	0.1	R femoral head	L femoral head
3	F	200	2.3	R femoral head	–
4	M	120	(−0.5) a	Bilateral femoral heads	–
5	F	30	2.4	Bilateral femoral heads	L shoulder
6	M	80	2.5	Bilateral femoral heads	–
7	F	170	4.7	Bilateral ankles	L knee
8	F	5	0.7	Bilateral femoral heads	R shoulder
9	M	37	0.7	Bilateral femoral heads	–
10	M	140	4.3	Bilateral femoral heads	–
11	F	10	1.2	Bilateral femoral heads	–
12	M	60	7.4	R shoulder	Bilateral femoral heads
13	F	260	9.4	Bilateral femoral heads	Bilateral knees
14	F	110	4.4	L talus	–
15	M	20	1.7	L femoral head	R shoulder

Table 2 shows that a minority of cases had a history of smoking (38%), alcohol use (17%), diabetes (14%), hepatitis C infection (7%), hypercholesterolaemia (17%) and hypertriglyceridaemia (17%). Of all, 13 (87%) cases had received ART prior to the diagnosis of AVN. Nine (60%) cases had received oral or intravenous steroids prior to their diagnosis of AVN. The median aggregate equivalent dose of prednisolone taken between the date of the first dose of corticosteroids and the date of diagnosis of AVN was 93.3 mg/month. Of these, five had received corticosteroids in the 12 months preceding the AVN diagnosis. The median aggregate corticosteroid equivalent dose taken in the 12 months for these cases was 1960 mg. The indications for systemic steroids were the treatment of opportunistic infections (n = 7) or other conditions (asthma, n = 1; dermatitis, n = 1).

OPEN IN VIEWER

Table 2.

Univariate analysis of possible risk factors for AVN.

Characteristic	Cases (n, %)	N a	Controls (n, %)	N a	p
Total	15	15
Black African ethnicity	10 (67)	15	13 (87)	15	0.38
AIDS at diagnosis	7 (47)	15	4 (27)	15	0.51
Evidence of sickle cell b	4 (33)	12	3 (30)	10	1
HBsAg-positive	1 (7)	15	1 (7)	13	1
HCV-Ab positive	0 (0)	10	1 (9)	11	1
Ever smoked	5 (38)	13	3 (23)	13	0.69
Alcohol use	2 (17)	12	5 (42)	12	0.25
Diabetes mellitus c	2 (14)	14	0 (0)	15	0.50
History of steroid use prior to AVN	9 (60)	15	0 (0)	15	0.004
Total cholesterol level of ≥6.2 mmol/L d	2 (17)	12	1 (7)	14	1
Triglyceride level of ≥4.5 mmol/L d	2 (17)	12	0 (0)	14	1
Previous exposure to ART	13 (87)	15	12 (80)	15	1
Previous exposure to PI	10 (67)	15	7 (47)	15	0.38
Previous exposure to NNRTI	10 (67)	15	10 (67)	15	1
Previous exposure to NRTI	13 (87)	15	12 (80)	15	1
Cumulative exposure to ART:	Cases (median weeks, IQR)	Controls (median weeks, IQR)
Any ART	99 (32–188)	15	104 (22–207)	15	0.67
PI	32 (0–107)	15	0 (0–29)	15	0.03
NNRTI	14 (0–109)	15	84 (0–142)	15	0.15
NRTI	72 (23–188)	15	104 (22–207)	15	0.22

A majority of cases (12, 80%) had at least one of the following five established risk factors for AVN in the general population: steroid use, history of smoking, alcohol use, hypercholesterolaemia and hypertriglyceridaemia.

Univariate analysis of possible risk factors demonstrated that the previous use of steroids was strongly associated with AVN (9 cases vs. 0 controls, p = 0.004). In addition, cases had longer cumulative exposure to PI (median exposure 32 vs. 0 weeks, p = 0.03) although exposure to class of ART agent or ART itself was not associated with AVN. There was no association with cumulative exposure to any other ART class or overall to ART. There were insufficient cases to perform a multivariate analysis.

Discussion

In this case–control study, AVN was statistically significantly associated with two characteristics: a history of prior steroid use and cumulative PI use. There was no statistically significant association between AVN and exposure to ART itself or to individual antiretroviral drug class.

History of prior steroid use, which is an established risk factor for AVN in non-HIV-infected populations, was strongly associated with AVN as found in most previous case–control studies in HIV-positive individuals.^4,6,9,11 Our data for steroid exposure are broadly comparable to that of Fessel et al., ¹³ in terms of aggregate corticosteroid dose given per month prior to AVN diagnosis and the dose given in the year prior to AVN diagnosis. In addition, the proportion who had received corticosteroids in the year prior to AVN diagnosis was 5/15 (33%) in our data compared with (3/12) 25% for Fessel et al. Suggestions of how steroid exposure mediates cell death include increased apoptosis through activation of free radicals and inflammatory mediators, ⁴ increased intraosseous pressure due to fat infiltration of bone marrow causing obstruction in small blood vessels ¹¹ or fat embolization in vessels through secondary hyperlipidaemia. ¹¹ Although in other case–control studies, cases were more likely than controls to have hyperlipidaemia,^4,9–11 it only reached statistical significance in one study. ¹¹ In this study, there was no statistical difference in the cholesterol or triglyceride levels between cases and controls (data not shown).

We found cumulative PI use to be significantly associated with AVN in contrast to other case–control studies where it has been tested.^4,8–10 A case–control study from Aquitaine, France found no association between AVN and ART itself or any individual ART class. ⁴ In that study, as well as two further case–control studies, there was also a trend for AVN cases to be more likely to have received PI as a class compared with controls but this did not reach statistical significance.^8,9 Scribner et al. found an association between AVN and saquinavir use which they were unable to explain. ¹⁰ Hasse et al. found that there was a significantly higher exposure in AVN cases to a drug in a different ART class, efavirenz. ⁸ This was attributed to their cases being more likely to have commenced cART early in the course of disease with subsequent treatment failure and the use of efavirenz in their salvage therapy. Such an explanation is unlikely to explain a PI class effect in this study because ART duration prior to AVN diagnosis (or equivalent follow-up) was not significantly different between cases and controls. A link between AVN and PIs is plausible if mediated by altered lipid metabolism and increased atherogenic and thromboembolic risk. ¹¹ However as explained earlier, no association of AVN with hyperlipidaemia was seen in this study. Nonetheless, with modest power and too few cases to conduct a multivariable analysis, the association between cumulative PI use and AVN shown here must be treated with caution. Hence, we are in agreement with Lawson-Ayayi et al., ⁴ that the importance of PIs and ART is probably limited in comparison with the role of steroid exposure.

In this study, 12 (80%) AVN cases could be accounted for by at least one of five established AVN risk factors: steroid use, history of smoking, alcohol use, hypercholesterolaemia and hypertriglyceridaemia. In 1999, a study of 339 asymptomatic HIV-infected patients screened by MRI showed that 4.4% had hip AVN compared with none in age- and sex-matched HIV-negative controls who were selected because they had no established risk factors for AVN. ¹⁴ This suggests that HIV disease (or its therapy) is an independent risk factor for AVN and may account for at least the remaining 20% of the AVN cases presented here.

Moreover, the fact that AVN may be asymptomatic and relatively common in HIV-positive patients suggests it could occur at multiple sites and present after different latencies at different joints. Whether AVN is caused by an acute event or by a dynamic process related to HIV disease is, however, unclear. In this study, it has not been possible to evaluate this as patients diagnosed with AVN at one joint did not have imaging performed at other sites to exclude further joint involvement at presentation. However, a majority of the cases in this study had AVN at more than one joint at diagnosis, most commonly both femoral heads. More than half went on to develop AVN at sites other than that of their initial diagnosis. In HIV-positive patients, AVN is commonly diagnosed at more than one joint.^8–11,15 In an extreme case, AVN was reported at eight different joints in one HIV-positive individual. ⁵ Therefore, the role of cumulative risks related to HIV disease or its treatment is of interest to physicians.

To our knowledge, this is the first case–control study from the United Kingdom. The HIV cohort attending the North Middlesex Hospital numbers approximately 1000 individuals and mostly comprises individuals of Black African origin living in the UK. Previous case–control studies have been carried out in HIV-positive populations with different characteristics, including a higher proportion of men, intravenous drug use, smokers, alcohol use and white race,^4,6–11 which may have influenced the risk factors that are detectable in the analysis. In comparison to our data, a case series from London of HIV-positive individuals with AVN demonstrated similar levels of prior steroid use, exposure to ART and nadir CD4 count but higher proportions of other established risk factors for AVN than in our cases, such as smoking, alcohol use, hypercholesterolaemia and hypertriglyceridaemia. ¹⁵

This retrospective study could only include information provided in medical records, including tests carried out as part of routine clinical care. This limited the findings in two ways; firstly, some data were missing; secondly, not all possible risk factors for AVN could be examined; thirdly, information about certain risk factors was limited, for example alcohol exposure was not available and, therefore, any alcohol use was used. Also the, results of coagulation studies were not always recorded. Together with the limited number of AVN cases, this affects the statistical power of the study. Furthermore, a major limitation of our study was the lack of a multivariate analysis of the data. We have limited ourselves to a univariate analysis due to the small number of cases. We feel that a multivariate analysis of our data would risk over-fitting the model and results would be of questionable validity. Hence, we acknowledge that results from a univariate analysis are more exploratory and as stated earlier must be treated with caution. Larger prospective studies would be required to further investigate the risk factors for AVN in HIV-infected individuals and the extent to which established risk factors are potentiated in the presence of HIV infection.

In the meantime, physicians should be aware of the risk of AVN with steroid use, the increased risk in the HIV-positive population and the importance of early diagnosis and avoidance of established risk factors in order to prevent further joint involvement if possible.