| |
Increased risk of myocardial infarction with duration of protease inhibitor therapy in HIV-infected men
|
| |
| |
AIDS 2003; 17(17):2479-2486
Murielle Mary-Krause a; Laurent Cotte b; Anne Simon c; Maria Partisani d; Dominique Costagliola a; and the Clinical Epidemiology Group from the French Hospital Database
“...The risk of MI in HIV-infected men increased with the duration of PI treatment, while other classes of drug treatment did not associate with MI risk... Longer follow-up under PI is necessary to observe if the risk of MI will continue to increase with the duration of PI exposure, and further studies are necessary to confirm the impact of NNRTI on CHD occurrence as observed in the DAD study..
... CHD risk factors in HIV-positive patients can be considered in two groups: those that are a consequence of treatment and ones independent of treatment. The first includes cholesterol, triglyceride or blood glucose levels, as different studies support a major role for PI in the pathogenesis of metabolic abnormalities such as increases in triglyceride and cholesterol...
.... Our results suggest an increasing risk of MI with the duration of PI treatment among HIV-infected men. However, the increase in life expectancy conferred by HAART clearly outweighs the associated risk of MI. In keeping with preliminary guidelines, the risk of CHD must be taken into account in PI treatment decisions, especially for patients with known risk factors for CHD. Cholesterol, triglycerides and blood glucose levels must be determined before and regularly during HAART in order to diagnose any abnormalities as they occur. If lipid-lowering drug therapy is indicated, it should be limited to those agents with a low risk of interaction with antiretroviral drugs”.
ABSTRACT
Background: In the current context of dyslipidaema, hyperglycaema and lipodystrophia observed among HIV-seropositive subjects, it is important to study the risk of myocardial infarction (MI) in this population. The French Hospital Database on HIV, which includes a large number of seropositive subjects followed for substantial periods, offers the opportunity to analyse the impact of protease inhibitors (PI) on the risk of MI among men.
Methods: Cox model was used to study the risk factors of MI occurrence. Standardized morbidity ratios (SMR) in men exposed to PI were calculated with data from the French general male population (FGMP) of the same age as reference.
Results: Between 1996 and 1999, MI was diagnosed in 60 men among 88 029 person-years (PY), including 49 cases among men exposed to PI. In the Cox model, exposure to PI was associated with a higher risk of MI [relative hazard (RH), 2.56; 95% confidence interval (CI), 1.03-6.34]. The expected incidence in the FGMP was 10.8/10 000 PY. The SMR relative to the FGMP was 0.8 (95% CI, 0.5-1.3) for men exposed to PI for < 18 months (G1), 1.5 (95% CI, 0.8-2.5) for men exposed for 18-29 months (G2) and 2.9 (95% CI, 1.5-5.0) for men exposed for [ges] 30 months (G3). With G1 as reference, the SMR was 1.9 (95% CI, 1.0-3.1) for G2 and 3.6 (95% CI, 1.8-6.2) for G3.
Conclusion: Our results point to a duration-related effect relationship between PI and MI, with a higher MI incidence rate among men exposed to PI for 18 months or more.
Introduction
With the advent of more effective therapies for HIV infection, including protease inhibitor drugs (PI), the morbidity of AIDS-defining illnesses has been reduced and HIV-infected patients are living longer. Although abnormalities of lipid metabolism in HIV-infected patients were described well before the advent of highly active antiretroviral therapy (HAART), several class-specific metabolic effects of PI therapy may have a deleterious effect on the heart, including increased insulin resistance, abnormalities of lipid metabolism, and a lipodystrophy syndrome. HAART has occasionally been implicated in the aggravation of coronary heart disease and other vascular complications. However, the impact of PI on the risk of coronary heart disease (CHD) and myocardial infarction (MI) in particular remains controversial.
The French Hospital Database on HIV (FHDH), which includes a large number of subjects and substantial follow-up, offers an opportunity to analyse the impact of PI on the risk of myocardial infarction. The present study uses the dataset to analyse risk factors for MI, including PI, in HIV-infected men. The incidence rate of MI according to the estimated duration of PI therapy is compared with the expected incidence in the general population of men of the same age.
Statistical analysis of myocardial infarction incidence rates and comparison with the general population
The incidence rate of MI among subjects exposed to PI between 1996 and 1999 was estimated according to the duration of PI therapy. Exposure durations were initially categorized by 6-month intervals. The final analysis was based on three durations: < 18 months (group 1), 18-29 months (group 2) and >30 months (group 3). These particular intervals were chosen on the basis of an adequate number of person-years (PY) and close incidences.
Incidence data for MI in the general French population of the same age and sex as subjects exposed to PI were obtained from myocardial infarction and coronary death population registries in three French regions (Lille, Strasbourg and Toulouse). These registries, which cover approximately 1 million inhabitants each, participated in the WHO Monica project during the period 1984-1993. Since 1997, all clinically diagnosed MI (lethal and non-lethal) and coronary death have been recorded by systematic review of hospital data and death certificates for men and women aged < 65 years and living in these regions. Age- and sex-specific incidence rates (first events) were computed, and data are available for the period 1997-1998.
The 1996-1999 MI incidence rates estimated from the FHDH dataset were compared with the 1997-1998 incidence rates estimated from the general male population using standardized morbidity ratios (SMR). Then, in order to take into account potentially different CHD risk factor profiles (age, sex, family history of CHD, smoking, hypertension and diabetes mellitus) between the general population and the HIV-infected population, the MI incidence rates in groups 2 and 3 were compared with those in group 1.
Results
MI was recorded in 66 patients. Following the validation survey, six subjects were excluded from the analysis because they had a past history of MI. All the other cases corresponded with the definition. The analysis, therefore, included 34 976 men with a follow-up corresponding to 88 029 PY. MI was diagnosed in 49 subjects among 39 023 PY of exposure to PI. The main characteristics of the patients in the groups with and without MI are shown in Table 1. Significant differences were observed from Cox univariate models in terms of age, the CD4 cell count at entry, and exposure to PI (62%.
|
|
| |
| |
| |
Pts without MI |
Pts with MI |
p-value |
| Number |
34,916 |
60 |
|
| Followup |
33 mos |
28 mos |
|
| Age |
38 yrs |
42 yrs |
0.0009 |
| Transmission group |
| -homosexuals |
48% |
51% |
0.9681 |
| -IVDUs |
22% |
23% |
|
| -heterosexuals |
20% |
18% |
|
| -exposure to blood products |
2% |
1.7% |
|
| -others/unknown |
2.7% |
5% |
|
| First CD4 count |
249 |
202 |
0.0373 |
| Time on ART |
34 mos |
35 mos |
|
| Time on PI |
25 mos |
21 mos |
|
| Pts presenting event before 1st MI or end of followup |
| -CD4 count <50 |
25% |
30% |
0.1808 |
| -AIDS OI |
67% |
65% |
0.4426 |
| ART |
| --nukes |
87% |
95% |
0.1421 |
| --NNRTI |
27% |
21% |
0.37 |
| --PI |
62% |
81% |
0.002 |
|
|
| |
| |
In the multivariate analysis, initial CD4 cell counts, exposure to NRTI and exposure to NNRTI did not influence the risk of MI (Table 2).
Table 2. Multivariate Analysis; Relative Hazard for Development of MI
|
|
| |
| |
| |
RH |
p-value |
| Age (per 10 yr increment) |
1.42 |
0.0067 |
| Initial CD4 count (per log2 unit increment) |
0.94 |
0.3428 |
| ART |
| --nukes |
0.93 |
0.9309 |
| --NNRTIs |
1.38 |
0.3814 |
| --PI |
2.56 |
0.0423 |
|
|
| |
| |
MI was more likely in older patients and those exposed to PI. The risk of MI increased by 42% per 10-year age increment, and more than twofold in patients exposed to PI.
Among the patients exposed to PI for < 18 months, there were 21 MI (25 734 PY); there were 15 MI among those exposed for 18-29 months (9440 PY) and 13 among those exposed for >30 months (3849 PY) (Table 3). The expected incidence of MI in the general male population with same age distribution was 10.8/10 000 PY. In the FHDH dataset, the estimated incidence was 8.2/10 000 PY (95% CI, 4.7-11.7) in patients exposed to PI for < 18 months, 15.9 (95% CI, 7.9-23.9) in those exposed for 18-29 months, and 33.8 (95% CI, 15.4-52.1) in those exposed for >30 months. No significant difference in incidence was observed between the general population and patients treated with PI for < 18 months. The risk of MI was increased, but not significantly, among patients treated with PI for 18-29 months. In contrast, the risk of MI among patients exposed to PI for >30 months was three times that of the general population (SMR, 2.9; 95% CI, 1.5-5.0). Compared with patients exposed to PI for < 18 months, those treated for >18 months were at an increased risk of MI (SMR, 1.9; 95% CI, 1.0-3.1 and SMR, 3.6; 95% CI, 1.8-6.2 for exposure of 18-29 and >30 months, respectively).
Table 3. Incidences of MI according to Duration of Protease Inhibitor Exposure
|
|
| |
| |
| Duration |
# of |
# of MI |
Incidence per |
| PI Exposure |
Person yrs |
Diagnosed |
10,000 person-yrs |
| <6 mos |
10,039 |
4 |
3.98 |
| 6-11 |
8,576 |
9 |
10.49 |
| 12-17 |
7,119 |
8 |
11.24 |
| 18-23 |
5,521 |
8 |
14.49 |
| 24-29 |
3,919 |
7 |
17.86 |
| 30-35 |
2,449 |
12 |
49.00 |
| >36 |
1,400 |
1 |
8.82 |
|
|
| |
| |
Discussion
The risk of MI in HIV-infected men increased with the duration of PI treatment, while other classes of drug treatment did not associate with MI risk.
Choosing 1 January 1996 as the starting point for the study shortened the time under NRTI in patients who started treatment before 1996. If time on NRTI was associated with MI (as suggested by David et al. [25] and by the fact that many side effects of HAART, such as the lipoatrophy, are probably in part caused by NRTI [26]), we might be unable to detect it. However, carrying out the analysis with 1 January 1992 as the starting point, showed no association between NRTI and MI but still found PI as an independent risk factor [relative hazard (RH), 1.40; 95% CI, 0.55-3.54 for NRTI; RH, 2.39; 95% CI, 1.36-4.21 for PI].
CHD risk factors in HIV-positive patients can be considered in two groups: those that are a consequence of treatment and ones independent of treatment. The first includes cholesterol, triglyceride or blood glucose levels, as different studies support a major role for PI in the pathogenesis of metabolic abnormalities such as increases in triglyceride and cholesterol. These parameters are known to undergo modifications in HIV-infected subjects and in healthy volunteers receiving PI [6, 27]. The fact that our analysis was not adjusted for cholesterol, triglyceride or blood glucose levels [28-30] is unlikely to introduce a source of bias as these factors are likely to mediate a potential impact of antiretroviral treatment on the MI risk. In this light, it would have been interesting to use those data in a multivariate model to study whether change in these parameters could explain the impact of PI exposure on the risk of MI.
The second category includes known CHD risk factors such as age, familial and personal history of CHD, smoking, obesity and arterial hypertension. The Cox model used here to identify risk factors for MI was adjusted for an important CHD risk factor in the general population, age. Increasing age was found as an independent risk factor for MI in the study population. The relation of PI to MI occurrence (non-adjusted RH, 2.93; 95% CI, 1.48-5.82) did not modify if the Cox model was adjusted for age suggesting than PI was not prescribed according this risk factor.
The Cox models used were not adjusted for the other known CHD risk factors such as familial and personal history of CHD, smoking, obesity and arterial hypertension, which were not recorded in the FHDH cohort. While risk factors for CHD in general are also associated with the risk of MI, it is extremely unlikely that they influenced the decision to prescribe PI in the study population, as was observed for age. Indeed, during the study period, the PI treatment decision was solely based on prognostic factors for HIV disease. However, we could not exclude the possibility that CHD risk factors would be more prevalent in the patients who were exposed to a PI for longer, who would be those who received PI earlier in the study. If this hypothesis was true, one would expect that the risk of MI among those exposed for < 18 months would be higher if they begun HAART in 1996 than if they began later. This was not supported by an analysis of the incidence of MI among people with < 18 months PI exposure according to the year between 1996 and 1998 when the PI-containing regimen was started. Indeed, the incidences per 10 000 PY fluctuated but did not decrease over time (6.4 ± 2.4 in 1996, 10.9 ± 3.4 in 1997 and 8.6 ± 4.3 in 1998). Therefore, this hypothesis was not supported by the data for 1996-1998 and could be excluded as an explanation for the results seen in 1996-1999.
An ascertainment bias could have influenced our results given the way data are collected in the FHDH. Indeed, for patients who started PI-containing regimen in 1996, when PI were first introduced, physicians may have been less aware of having to suspect cardiovascular morbidity than in later years, possibly resulting in underreporting of MI in the earlier period covered by the study. If this ascertainment bias was major, one would expect to observe increasing risk of MI for a given duration of exposure in more recent years of PI institution. As shown for patients with < 18 months PI exposure, and also for analyses with longer duration of exposure stratified by PI starting year, the incidences fluctuated with no consistent pattern whatever the PI exposure duration. Therefore, even if some ascertainment bias did occur, it should not be considered as a major factor in the study. If the literature on the increased incidence of hyperlipidaema and insulin resistance among patients on PI published since 1999 influenced physicians and patients to discontinue PI use and thereby shorten PI exposure, this would actually be a form of negative confounding, biasing the study towards a negative result.
The incidence estimated among subjects not exposed to PI (2.24/10 000 PY; 95% CI, 0.92-3.57) in the FHDH was lower than expected in the general population with same age and sex. That is the reason why no comparison between the incidence among men exposed to PI and men not exposed to PI was made, as there will be a major bias in such comparison. To account for the possibility that MI was more efficiently diagnosed and reported in patients exposed to PI than in other patients, the second analysis was restricted to patients exposed to PI for whom this potential bias could not occur. Comparisons were done with the general population with same age and sex. As CHD risk factors in French HIV-infected patients are not the same as those in the French general population [31, 32], comparisons were also made with HIV-infected men exposed to PI for < 18 months more likely to have similar CHD risk factors.
Both comparisons showed a relationship between the risk of MI and the duration of exposure to PI, particularly when the reference population consisted of patients exposed to PI for < 18 months. The observed effect of PI duration of use could not be explained by those who used PI having more risks for CHD because the risks obtained when the reference population consisted of patients exposed to PI for < 18 months were higher than those obtained when the reference was the general population.
Previous studies have either shown [33-37] or not shown [25, 38-40] a link between the risk of CHD and exposure to PI. Some of these studies suffered from methodological problems, such as the sample size, median duration of PI treatment of < 12 months in two studies, and likely underreporting relative to the general population. A recent study showed no increase in the rate of admission for or death from cardiovascular or cerebrovascular disease between 1995 and 2001 in a large population of HIV-infected US veterans despite the introduction and increasing use of HAART [40]. A direct comparison of these results with those of our study is impossible because the two studies did not examine the same type of events. In our study, we looked at the admissions and the deaths from MI whereas Bozzette et al. [40] only looked at the admissions for cardiovascular diseases. In their study, death or admissions were grouped for cardiovascular and cerebrovascular diseases. However, Holmberg et al. [35], who suggested an effect of PI on the risk of MI, did not highlight any risk for cerebrovascular disease with PI exposure. Moreover, the ICD-9 codes used to define cardiovascular disease by Bozzette et al. [40] included 410, 411, 413, 414, which cover a number of pathologies (while we only used ICD 410); it would be interesting to study the impact of PI on each kind of disease separately. Related to this, it should be noted that a higher, although not significant, risk was observed among subjects exposed to PI who were admitted with cardiovascular diseases (RH, 1.23). In addition, the incidence of admission for cardiovascular diseases > 10/1000 PY observed by Bozzette et al. is much higher than the incidence rate of CHD observed among American seropositive subjects exposed to PI in other studies (5.9-7.1 events/1000 PY) [37, 39].
In the recently presented DAD study [36], the relationship between exposure to antiretroviral therapy including PI or NNRTI and the risk of MI was similar to the one observed in our study. Similarly, an increased risk of CHD was also identified in PI-exposed patients by Moore et al. [37].
If the results quoted in all the available literature are considered, it appears that exposure to PI does increase the risk of MI and this risk depends on the duration of PI exposure. Our results suggest an increasing risk of MI with the duration of PI treatment among HIV-infected men. However, the increase in life expectancy conferred by HAART clearly outweighs the associated risk of MI. In keeping with preliminary guidelines [41, 42], the risk of CHD must be taken into account in PI treatment decisions, especially for patients with known risk factors for CHD. Cholesterol, triglycerides and blood glucose levels must be determined before and regularly during HAART in order to diagnose any abnormalities as they occur. If lipid-lowering drug therapy is indicated, it should be limited to those agents with a low risk of interaction with antiretroviral drugs.
Longer follow-up under PI is necessary to observe if the risk of MI will continue to increase with the duration of PI exposure, and further studies are necessary to confirm the impact of NNRTI on CHD occurrence as observed in the DAD study [36].
|
|
| |
| |
|
|
|
