Introduction
Because of their lifestyle, injecting drug users (IDU) are at increased risk of premature mortality compared with the general population [1-4]. In the last decade, death rates increased considerably for IDU in countries with a high HIV-1 prevalence amongst this group due to the impact of the AIDS epidemic [5-8].
An increase of non-AIDS deaths concomitant with the introduction of HIV epidemic was first noticed amongst IDU in New York City at the beginning of the 1980s [6,9]. Since then several studies have shown that deaths due to non-AIDS causes are higher among HIV-positive IDU than among HIV-negative IDU [2,4-6,10,11]. Among HIV-infected IDU, mortality before being diagnosed with AIDS (denoted as pre-AIDS mortality) has been found to increase with a decreasing CD4 cell count [10], indicating that not accounting for pre-AIDS mortality could overestimate the AIDS-free survival curve. To date, all studies have evaluated pre-AIDS mortality in seroprevalent cohorts. To our knowledge, neither the probability of dying for the complete course from HIV-1 seroconversion to AIDS has been described, nor has an evaluation of whether the risk of pre-AIDS mortality increases with time since HIV-1 seroconversion.
In the present study, data were merged from IDU with documented intervals of HIV seroconversion obtained from eight cohorts in various European cities. Differences in pre-AIDS mortality rates between these regions, in which drug use, drug policy and access to health care might differ, were studied. The relationship between pre-AIDS mortality and time since HIV-1 seroconversion as well as immunosuppression were also studied.
Methods
Study population
The HIV-1-positive study population comprised 664 IDU for whom the dates of the last negative and first positive HIV-1 test were known. An IDU was defined as a person who, since 1979, had injected drugs before HIV-1 seroconversion.
These IDU originated from eight ongoing prospective studies of HIV infection participating in the European Seroconverter Study: the Valencian HIV Seroconversion Study (n = 246); the Edinburgh Drug Addiction Study and the Edinburgh City Hospital Cohort Study (total number of seroconverters of these two studies combined, 169); the Amsterdam Cohort Study among drug users (n = 99); the Geneva HIV Cohort Study (n = 60); the French SEROCO Study (n = 38); the Scottish National Collaborative HIV testing Study from Glasgow (n = 31); and the Innsbruck AIDS Study (n = 21). The study from Glasgow originally included Scotland-wide HIV-infected individuals, but in the present study predominantly included HIV-sero-converted IDU from Glasgow and surrounding regions because those followed in the Edinburgh cohorts were excluded from these Scotland-wide data.
The original studies started between 1982 and 1988. The design and methodology of each study have been described in detail elsewhere [12-19]. Briefly, participants underwent standardized clinical examination, blood testing and most of them completed questionnaires every 3-6 months. Lymphocyte subsets were determined by flow cytometry. Information on AIDS diagnosis and cause of death were obtained from review of medical records and/or through matching with local and national registers at each study location.
The total study population was composed of IDU who entered the study HIV-negative and seroconverted during follow-up (Prospectively identified seroconverters, 351), and IDU who were HIV-infected at entry into the study but who had earlier blood samples available to determine the interval of seroconversion (retrospectively identified seroconverters, 313). For the latter, HIV-negative blood samples were mostly obtained for reasons unrelated to HIV disease progression, such as hepatitis B testing or knowledge of risk behaviour.
In this analysis, AIDS was defined by the 1987 criteria of the Centers for Disease Control and Prevention (CDC) [20]. The Glaswegian, French and Innsbruck data were combined because of the small number of seroconverters included in each cohort.
Statistical analysis
The expected date of seroconversion was calculated for each IDU by the following procedure: first, the cumulative HIV-1 seroincidence was estimated over calendar time for each cohort separately applying methods for interval-censored data; second, using the cohort-specific seroincidence distribution, the expected date of seroconversion was calculated for each subject conditional upon the date of each subject's last HIV-negative and first HIV-positive test [21-23].
Product-limit estimates of the cumulative pre-AIDS mortality and the cumulative AIDS-free incidence following HIV seroconversion were calculated using a competing risk model based on influenced transition probabilities [24]. Kaplan-Meier estimates of the cumulative pre-AIDS mortality stratified by cohort were also calculated. In these calculations, as well as for the calculation of person-years at risk of pre-AIDS dying, subjects were considered at risk from the calculated date of seroconversion for prospectively identified seroconverters through the date of pre-AIDS death, AIDS, loss to follow-up or censoring. For retrospectively identified seroconverters the date of study entry was taken as the first date since seroconversion at which they were included in the risk set (i.e., left-truncation). Depending on the closing date of each data file merged and taking into account reporting delay of registries, the date of censoring (i.e., cut-off date of the analysis) was ultimately 1 August 1995.
To evaluate differences in pre-AIDS mortality rates between the participating cohorts, and to examine the relationship of pre-AIDS mortality with time since HIV seroconversion and CD4 lymphocyte counts, incidence rates and relative risks (RR) with their corresponding 95% confidence intervals (CI) were calculated. To calculate incidence rates, pre-AIDS deaths were divided by person-years according to study site, time since seroconversion (< 2, 2-4, 4-6 and ≥ 6 years) and categories of CD4 lymphocyte counts (≥ 500, 200-500 and < 200 × 106/l). Evaluating site differences, the Edinburgh cohorts were chosen as the reference category because these relatively large cohorts became infected earlier than the other cohorts.
To evaluate the association between CD4 lymphocyte count and pre-AIDS mortality, the risk of pre-AIDS mortality was predicted from the CD4 count at the preceding visit. To limit misclassification of person-time due to temporal changes in this marker, observations were censored maximally 1 year after each visit with a CD4 count measurement.
Gender, age and calendar time were also evaluated as covariates of pre-AIDS mortality. Categories of age were defined by cut-off points at the 33rd and 67th percentile of the age at seroconversion for the total group (23 and 27 years, respectively). Calendar time was divided into the following three periods: < 1990, 1990-1991, ≥ 1992. Both calendar time and age were treated as time-dependent covariates.
Poisson regression was used to test for significance and multivariate analyses [25]. Significance was determined by likelihood ratio tests. Apart from CD4 lymphocyte count, time since HIV seroconversion and site, all univariately associated covariates (P < 0.10), were considered for entry in the multivariate models using a forward stepwise procedure. In multivariate analysis, we tested for interactions between covariates in the final models and considered the confounding effect of covariates. Covariates had been evaluated within each cohort before pooled analyses of the data were performed. Furthermore, bias due to the length of the seroconversion interval, and the interval between sero-conversion and recruitment among retrospectively identified seroconverters were examined by repeating the analysis for subjects with a seroconversion interval of less than 2 years and those recruited before or maximally 1 year after seroconversion. This did not change the main results substantially.
Finally, in addition to studying pre-AIDS deaths from any cause, we repeated our person-time methods separately for pre-AIDS deaths from natural causes (all deaths excluding deaths from overdose, suicide, unintentional injuries and unknown causes) and pre-AIDS deaths due to overdose or suicide. In these cause-specific analyses, subjects who died from causes other than causes of interest were censored at the date of death. The small number of pre-AIDS deaths (n = 6) other than from natural causes, overdose and suicide precluded separate analysis.
Results
Characteristics of the 664 IDU stratified by cohort are displayed in Table 1. The median follow-up time from HIV seroconversion until pre-AIDS death, AIDS, or the end of the study was 4.0 years [interquartile range (IQR), 2.1-6.9 years]. The 313 IDU (47.1%) for whom seroconversion was determined retrospectively had a median lag time between seroconversion and enrolment of 1.6 (IQR, 0.7-2.7) years. Median age at seroconversion ranged from 23.1 years for the Edinburgh cohorts to 29.7 years for the Amsterdam cohort, and median calendar year of seroconversion ranged from 1984 in Edinburgh to 1991 in the Valencian region. For all subjects, the median interval between the last negative and the first positive HIV test was 1.1 year; 296 IDU (44.6%) had seroconversion intervals of less than 1 year, 178 (26.8%) between 1 and 2 years, and 190 (28.6%) of 2 years or more. A total of 108 subjects died, of whom 57 died without AIDS and 50 with AIDS. For one subject who died, the occurrence of AIDS could not be ascertained. Of those not known to have developed AIDS or to have died, 28% of the total population had their last study visit more than 1 year before the cut-off date of the analysis.
Causes of pre-AIDS death and CD4 counts preceding death
Causes of pre-AIDS deaths are listed in Table 2. Natural causes contributed to 23 pre-AIDS deaths (40.4%). Most of these deaths were attributable to bacterial infections, followed by cirrhosis/liver failure. Twenty-four deaths (42.1%) from overdose occurred and four deaths (7.0%) were attributed to suicide. The documented suicides were committed by electrocution (n = 1), defenestration (n = 1), overdose accompanied by a farewell letter (n = 1), and by unknown methods (n = 1). Remaining causes were accidents (n = 4) and violence (n = 1). Cause of death was unknown for one subject.
Among the 57 IDU who died without AIDS, 33 (57.9%) had a CD4 cell count in the year preceding pre-AIDS death. For these 33 subjects, the median lagtime between CD4 determination and pre-AIDS death was 2.9 months (IQR, 1.2-4.1) and the median CD4 cell count 350 × 106/l (IQR, 193-580).
Cumulative pre-AIDS mortality
Considering pre-AIDS mortality and AIDS as competing risks, at 3 years from seroconversion, 5.8% of all IDU were expected to have died without AIDS and 4.3% were expected to have developed AIDS (Fig. 1). At 5 years since seroconversion these figures were 9.4 and 11.1%, and at 7 years 14.7 and 17.3%, respectively. As shown by the vertical distance between both lines (Fig. 1) at the 7-year timepoint, 68.1% of the population was still alive without AIDS. Median progression time to pre-AIDS mortality or AIDS was 9.72 years. Using this competing risk model to calculate the rate ratio of pre-AIDS dying compared with developing AIDS, this ratio appeared to decrease slightly with increasing time since seroconversion from 1.02 at 2 years to 0.79 at 10 years from seroconversion.
By Kaplan-Meier estimates, considering AIDS not as competing risk but as censoring moment, the probability of pre-AIDS dying within 7 years from seroconversion was 16.0% (95% CI, 12.2-20.8). On comparison, 19.0% (95% CI, 14.8-24.1) developed AIDS within that time period by Kaplan-Meier estimates. Fig. 2 shows the Kaplan-Meier curves of pre-AIDS mortality for each cohort. Progression rates did not differ significantly across sites (likelihood ratio statistic by Cox proportional hazards analysis, P = 0.13).
Mortality rates and risk factors for pre-AIDS mortality
The total number of person-years at risk until pre-AIDS death, AIDS, loss to follow-up or censoring was 2454, resulting in a crude pre-AIDS mortality rate per 100 person-years of 2.32 (95% CI, 1.79-3.01) for all IDU. Table 3 shows the pre-AIDS mortality rates, crude RR and their 95% CI according to variables of interest.
Pre-AIDS mortality from all causes
In univariate analyses studying pre-AIDS deaths from all causes, an increasing trend in mortality with time since HIV seroconversion was not demonstrated, although pre-AIDS mortality rates more than 2 years after seroconversion were somewhat higher than in the first 2 years. Pre-AIDS mortality rates increased with a decreasing CD4 cell count (P = 0.088). The risk of pre-AIDS mortality was about three times higher when CD4 counts fell below 200 × 106cells/l, relative to CD4 counts ≥ 500 × 106cells/l. Since the CD4 cell count might fall temporarily immediately after serocon-version and this might cause some misclassification, we also examined the relationship between CD4 lymphocyte count and pre-AIDS mortality excluding CD4 counts in the first year after seroconversion; RR became 1.44 (95% CI, 0.54-3.83) and 3.30 (95% CI, 1.20-9.08) for a CD4 cell count of 200-500 and < 200 × 106/l, respectively (P = 0.053). The death rate of 1.96 per 100 person-years for time periods with missing CD4 cell counts was very close to the rate for a CD4 cell count ≥ 500 × 106/l. An explanation might be that 41% of the person-years with missing CD4 cell counts took place in the first 2 years following serocon-version when CD4 cell counts were relatively high.
The lowest pre-AIDS mortality rate was found in Valencia, and the highest in Geneva, but both differences did not reach statistical significance when compared with the Edinburgh cohorts. After adjustment for time since seroconversion and CD4 lymphocyte count, the RR for Valencia increased to 0.68 (95% CI, 0.27-1.70) and the death rate for Geneva became significantly different from that in Edinburgh (RR, 2.55; 95% CI, 1.02-6.33). The RR for the Amsterdam cohort and the other cohorts combined did not change substantially after adjusting for duration of infection and CD4 cell count. Additional adjustment for gender and age had only a minor influence on the presented RR for the sites.
Pre-AIDS mortality from all causes was not significantly associated with gender, age and calendar period of follow-up: crude RR for women compared with men, 0.93 (95% CI, 0.53-1.61); RR for those aged 23-27 years, 1.00 (95% CI, 0.41-2.43), and for those aged > 27 years, 1.45 (95% CI, 0.64-3.26) compared with those aged <23 years; RR for 1990-1991, 1.25 (95% CI, 0.63-2.47) and for ≥ 1992, 1.37 (95% CI, 0.73-2.58) compared with <1990. CD4 count appeared to be the only variable independently associated with pre-AIDS death from all causes. None of the remaining covariates appeared to be confounders. CD4 log-transformed and treated as a continuous variable was found to be more predictive than absolute CD4 counts in categories.
Pre-AIDS mortality from natural causes
Both time since seroconversion and decreasing CD4 lymphocyte count were univariately significantly associated with death from natural causes (P= 0.003 and 0.001, respectively). The mortality rates were 0.13 the first 2 years, 0.73 in years 2-4 (RR, 5.59), 1.83 in years 4-6 (RR, 14.0) and 1.54 at 6 years or more after sero-conversion (RR, 11.7; Table 3). RR of 6.99 was found for a CD4 cell count < 200 × 106/l, relative to a CD4 count ≥ 200 × 106/l. Mortality from natural causes was also found to be significantly associated with older age (P = 0.020). RR was 1.75 (95% CI, 0.20-15.7) and 5.36 (95% CI, 0.72-40.2) for subjects aged 23-27 years and older than 27 years, respectively, relative to subjects aged 22 years or less. No significant differences in pre-AIDS mortality from natural causes by gender and calendar period of follow-up could be demonstrated.
In multivariate analysis, both time since HIV serocon-version and CD4 cell count were independent predictors of pre-AIDS mortality from natural causes. Table 4 shows the results when age and site were added to the model. Compared with RR for age obtained in the univariate analysis, RR decreased to non-significant values. As was the case for overall pre-AIDS mortality, Geneva had the highest risk of pre-AIDS death from natural causes. This association did not reach statistical significance, which may have been due to small numbers. Compared with the results from the univariate analysis, RR for time since HIV seroconversion and CD4 lymphocyte count decreased slightly towards 1. However, the risk still increased considerably with decreasing CD4 cell count and ongoing HIV infection, although the risk ≥ 6 years after seroconversion appeared to be smaller than in years 4-6. Log-transformed CD4 values appeared to be more predictive than categorized CD4 count. When replacing CD4 count in categories with log-transformed CD4 in the multivariate model, the elevated risks with ongoing time since seroconversion decreased. The effect was larger for more prolonged duration of infection, with a decrease of 53% for the category ≥ 6 years since HIV seroconversion.
Pre-AIDS mortality due to overdose and suicide
None of the variables studied, including time since seroconversion and CD4 lymphocyte count, were found to be significantly associated with pre-AIDS death from overdose or suicide. Regarding site differences, a relatively low rate of overdose and suicide deaths was found in the Valencia region and a high rate for both the Geneva cohort and the combined cohorts (Table 3). The latter was due to a very high overdose death rate in the Glasgow cohort.
In addition to parallel pre-AIDS mortality from natural causes with progression to AIDS, the person-time methods were carried out for progression from seroconversion to AIDS. The incidence rate was small in the first 2 years following seroconversion (0.92 per 100 person-years) and rose rapidly thereafter (3.08, 4.28 and 7.32 per 100 person-years for years 2-4, 4-6, ≥ 6, respectively; P < 0.001). The risk of developing AIDS was strongly related to a decreasing CD4 lymphocyte count (P < 0.001). The rate was 0.21 for a CD4 cell count ≥ 500 × 106/l, 2.47 for 200-500 × 106/l and 20.30 for < 200 × 106/l. Table 4 shows the results when time since seroconversion, CD4 count, age and site were considered together in a multivariate model. It is clear from Table 4 that the results for pre-AIDS death from natural causes resembled the results for progression to AIDS, although the relationship between outcome and immunosuppression was somewhat stronger for AIDS than for mortality.
Discussion
This study showed a high pre-AIDS mortality rate among IDU in various European sites, with no significant differences between them. A striking result of this study was that pre-AIDS death from natural causes was found to be related with ongoing time since HIV seroconversion and immunosuppression. The increase in these death rates by ongoing time since seroconversion and a decreasing CD4 count showed similarity with the rates observed for AIDS. As already postulated by researchers in New York City in 1988 [6,9], this strongly suggests that pre-AIDS death from natural causes is at least partially the result of HIV disease progression.
Our findings confirmed the results obtained in a recent study in the Amsterdam cohort of drug users [10]. In this previous study, which included seroprevalent subjects, crude RR for pre-AIDS mortality of 1.41 (95% CI, 0.43-4.57) and 2.66 (95% CI, 0.66-10.6) for CD4 count levels of 200-500 × 106 cells/l and < 200 × 106cells/l, respectively, relative to a CD4 count ≥ 500 × 106cells/l were found, but numbers were small. These risk ratios are very close to those for overall pre-AIDS mortality in our study. Since we studied a larger group, we were able to demonstrate a significant association. Furthermore, we could show that this association was even stronger for pre-AIDS deaths from natural causes. In contrast, the association does not hold for pre-AIDS deaths from overdose or suicide.
Because our study population consisted of subjects with well-estimated dates of HIV seroconversion, we were able to demonstrate that the risk of pre-AIDS mortality from natural causes also increased with ongoing time periods since seroconversion. Considering time since seroconversion and CD4 cell count together in a model, immunosuppression was still found to be strongly associated with pre-AIDS death from natural causes. In contrast, after adjusting for log-transformed CD4 count the elevated risk for increasing time since seroconversion reduced markedly, especially with more prolonged duration of infection. This suggests that at later timepoints CD4 count is much more relevant than time since HIV seroconversion. This finding is in agreement with several studies showing that time from HIV infection to AIDS or death due to AIDS are largely explained by CD4 lymphocyte count and not or to a lesser extent by the duration of HIV infection [26-28].
Several studies found an increased overdose/suicide death rate among HIV-positive IDU compared with HIV-negative IDU [2-4,29,30], whereas one study did not [11]. It has been argued that the increased overdose death rate among HIV-positive IDU might be causally related to HIV infection by an increasing probability of lethal outcome from drug use among seropositive IDU compared with seronegative IDU [3]. Another explanation might be that the increased risk among HIV-positive IDU is the result of unrecognized HIV-related diseases [31]. In our study, we did not find any indication that the evaluated overdose and suicide death rate among HIV-positive IDU is a consequence from disease progression, since the rate did not increase with time since seroconversion nor with a decreasing CD4 count. The increased overdose mortality rate among HIV-positive IDU might be explained by a selection of population with a high-risk lifestyle or psychiatric comorbidity that has already resulted in HIV infection. Nonetheless, an association between overdose/suicide mortality and disease progression after AIDS diagnosis might exist, since we studied only pre-AIDS deaths.
We stress that drawing firm conclusions about geographic differences in pre-AIDS mortality among the general population of HIV-infected IDU is hazardous because we studied a selected group, namely cohort participants, and because numbers were small. We can only be certain that pre-AIDS mortality was common among all IDU cohorts studied and that there were no significant differences across these cohorts. The risk in the Geneva cohort tended to be increased for pre-AIDS mortality as well as for progression to AIDS. A possible explanation is that fast progressors were slightly over-represented in this clinic-based cohort in which almost all seroconversions were determined retrospectively. The relatively high overdose death rate amongst the Glasgow cohort might be related to the increase in drug-related deaths in Glasgow in the beginning of the 1990s, which was probably due to mixing heroin with benzodiazepines and alcohol [32,33]. It has been hypothesized that the tendency of a low overdose/suicide death rate in Spain might be explained by a better supportive social network in the Valencian region than in the other regions, but data to test this hypothesis are lacking. In general, drug use, methadone provision and social and medical conditions vary in the different regions involved. For example, methadone has been available in Edinburgh since 1986 but is still very restricted in France; most IDU in Amsterdam do not live with their family in contrast to IDU in the Valencian region, for example. In other words, we cannot exclude the possibility that our results on site differences are confounded by factors that we did not take into account and that are often difficult to evaluate, such as drug use, psychopathology, poverty, methadone provision, social and medical care, study design-related factors (e.g., inclusion criteria, setting of follow-up), and completeness of mortality registers.
In this study, AIDS was defined using the 1987 CDC criteria [20]. In 1993, recurrent bacterial pneumonia, pulmonary tuberculosis and cervical neoplasia were added to the AIDS case definition in Europe [34]. Since diagnostic criteria (e.g., for recurrent pneumonia) are very strict, at most only for six (11%) of the 57 pre-AIDS deaths in this study, the cause of death might have satisfied this expanded definition. Of course, some subjects who died before AIDS might have met the criteria of the expanded definition some time before dying, regardless of the cause of death. In general, since pre-AIDS mortality and AIDS do not occur at random (i.e., both pre-AIDS mortality and AIDS are more likely when subjects are immunosuppressed), the introduction of the expanded definition should result in a decrease in pre-AIDS mortality rates. However, because diseases such as infective endocarditis are not defined as AIDS according to this expanded definition and because for example overdose is associated with the lifestyle of IDU, we would still expect to find a relatively high pre-AIDS death rate among this group when we apply the 1993 criteria. Furthermore, the recent introduction of early combination therapy (protease inhibitors and nucleoside combinations), which should be initiated before immunological damage has occurred, might delay the diagnosis of AIDS and also pre-AIDS death from natural causes. Consequently, the number of pre-AIDS death from other causes might increase because of longer exposure to a high-risk lifestyle.
Several limitations to our study should be noted. First, as mentioned previously, our results cannot be generalized, because we studied a restricted group of HIV-infected IDU. IDU not followed in cohort studies might have poorer access to health care resulting in a higher pre-AIDS mortality rate. In addition, other relevant characteristics might differ between cohort participants and non-participants. Second, the statistical power was not large in our study because of the relatively small number of pre-AIDS deaths. Therefore, causes of death were grouped together, although the association between some cofactors and pre-AIDS mortality may not be uniform. By combining overdose and suicide deaths, misclassification is expected to be small because there were only four documented suicides, of which one was an overdose with farewell letter. Furthermore, it is not known whether the documented overdoses were accidental or suicidal. Regarding pre-AIDS death from natural causes, death rates due to bacterial infections have consistently been found to be elevated among HIV-positive IDU in comparison with HIV-negative IDU [4-6,11,29,30]. Studies evaluating cirrhosis/liver failure death rates do not give concordant results; some have found a higher rate in HIV-positive IDU than in HIV-negative IDU [2,4,29,30], but one did not [11]. Since seroprevalent groups were studied, these inconsistent results might be due to differential time since infection. When we calculated death rates by time since seroconversion and immunosuppression separately for deaths due to bacterial infections and deaths from cirrhosis/liver failure, in both cases increasing rates were observed over time and with a decreasing CD4 count (data not shown). Finally, although assignment of the cause of death represented the best judgement of the investigators at each site based on all available information, the source of information (clinical diagnosis, post-mortem results, death certificate) differed across and within sites. Hence, some misclassification might have been present, but we feel that it is very unlikely that this has seriously biased our results.
We found strong evidence that pre-AIDS mortality among IDU underestimates the burden of disease in this population. Implications are that in survival analysis to AIDS or death from AIDS, pre-AIDS deaths from natural causes should not be treated as random censoring events because this might result in substantially biased progression estimates. Pre-AIDS death from overdose and suicide does not violate the requirement for random censoring. Those applying back-calculation methods to estimate the HIV incidence from seroprevalence and AIDS surveillance data as well as those using models such as Markov models to estimate the AIDS incubation time should consider all-cause pre-AIDS mortality.
Although there are not many well-defined cohorts of IDU, further studies among large groups of HIV-sero-converted IDU are needed to confirm our results and to study the impact of the introduction of the expanded AIDS case definition on pre-AIDS mortality. Pre-AIDS mortality and its association with disease progression should also be studied amongst other populations with a relatively high likelihood of pre-AIDS death, such as HIV-infected haemophilia patients [35] and HIV-infected individuals in sub-Saharan African countries [36].
In practice, clinicians should be aware of the risk of premature death among HIV-infected IDU who do not yet have AIDS. IDU who have low CD4 counts are at increased risk for premature death from natural causes. For these HIV-infected IDU, clinical check-ups should be intensified in order to detect diseases earlier. Increased attention to bacterial infections is warranted. Both CD4 count and time since seroconversion may be useful for considering specific interventions (next to HIV therapy) directed at the prevention and early treatment of bacterial infections [e.g., vaccination and (prophylactic) antibiotics].
Acknowledgements
The authors thank A.M. Richardson (Edinburgh), R.B. Geskus, L. Gerber, J.S. Bax, E.J.C. van Ameijden, H.J.A. van Haastrecht (Amsterdam), B. Hirschel, H. McArdle (Geneva), J. McMenamin (Glasgow) and the partners in the Valencian HIV Seroconversion Study (FIS 95/0228, 95/1688) for their contributions to this study. The authors also thank the laboratories collaborating with the original studies for determining lymphocyte subsets, the clinicians and health workers who contributed by collecting data, and the participants for their ongoing participation.
References
1. Perucci CA, Davoli M, Rapiti E, Abeni DD, Forastiere F: Mortality of intravenous drug users in Rome: a cohort study. Am J Public Health 1991, 81:1307-1310.
2. Zaccarelli M, Gattari P, Rezza G, et al.: Impact of HIV infection on non-AIDS mortality among Italian injecting drug users. AIDS 1994, 8:345-350.
3. Eskild A, Magnus P, Samuelsen SO, Sohlberg C, Kittelsen P: Differences in mortality rates and causes of death between HIV positive and HIV negative intravenous drug users. Int J Epidemiol 1993, 22:315-320.
4. Goedert JJ, Pizza G, Gritti FM, et al.: Mortality among drug users in the AIDS era. Int J Epidemiol 1995, 24:1204-1210.
5. Selwyn PA, Hartel D, Wasserman W, Drucker E: Impact of the AIDS epidemic on morbidity and mortality among intravenous drug users in a New York City Methadone Maintenance Program. Am J Public Health 1989, 79:1358-1362.
6. Stoneburner RL, Des Jarlais DC, Benezra D, et al.: A larger spectrum of severe HIV-1 related disease in intravenous drug users in New York City. Science 1988, 242:916-919.
7. Galli M, Musicco M, COMAT Study Group: Mortality of intravenous drug users living in Milan, Italy: role of HIV-1 infection. AIDS 1994, 8:1457-1463.
8. Orti RM, Domingo-Salvany A, Muñoz A, Macfarlane D, Suelves JM, Antó JM: Mortality trends in a cohort of opiate addicts, Catalonia, Spain. Int J Epidemiol 1996, 25:545-553.
9. Des Jarlais DC, Friedman SR, Stoneburner RL: HIV infection and intravenous drug use: critical issues in transmission dynamics, infection outcomes, and prevention. Rev Infect Dis 1988, 10:151-158.
10. van Haastrecht HJA, van Ameijden EJC, van den Hoek JAR, Mientjes GHC, Bax JS, Coutinho RA: Predictors of mortality in the Amsterdam cohort of human immunodeficiency virus (HIV)-positive and HIV-negative drug users. Am J Epidemiol 1996, 143:380-391.
11. Selwyn PA, Alcabes P, Hartel D, et al.: Clinical manifestations and predictors of disease progression in drug users with human immunodeficiency virus infection. N Engl J Med 1992, 327:1697-1703.
12. Hernández-Aguado I, Bolumar F: Determinants of HIV-1 infection in intravenous drug users in Valencia, Spain, 1987-1991. Int J Epidemiol 1993, 22:537-542.
13. Skidmore CA, Robertson JR, Robertson AA, Elton RA: After the epidemic: follow up study of HIV seroprevalence and changing patterns of drug use. BMJ 1990, 300:219-223.
14. Brettle RP, Bisset K, Burns S, et al.: Human immunodeficiency virus and drug misuse: the Edinburgh experience. BMJ 1987, 295:421-424.
15. van den Hoek JAR, Coutinho RA, van Haastrecht HJA, van Zadelhoff AW, Goudsmit J: Prevalence and risk factors of HIV infections among drug users and drug-using prostitutes in Amsterdam. AIDS 1988, 2:55-60.
16. Ledergerber B, Von Overbeck J, Egger M, Luthy R: The Swiss HIV Cohort Study: rationale, organization and selected baseline characteristics. Soz Praventivmed 1994, 39:387-394.
17. Carré N, Deveau C, Belanger F, et al.: Effect of age and exposure group on the onset of AIDS in heterosexual and homosexual HIV-infected patients. AIDS 1994, 8:797-802.
18. Goldberg DJ, Emslie JA, Smyth W, Reid D: A system for surveillance of voluntary HIV testing: results of the first 2 years 1989-1990. AIDS 1992, 6:495-500.
19. Zangerle R, Fuchs D, Rossler H, et al.: Trends in HIV infection among intravenous drug users in Innsbruck, Austria. J Acquir Immune Defic Syndr 1992, 5:865-871.
20. Centers for Disease Control: Revision of the CDC surveillance case definition for acquired immunodeficiency syndrome. MMWR 1987, 36:1S-15S.
21. Jongbloed G: The Iterative Convex Minorant Algorithm for Nonparametric Estimation. Technical Report. Delft: University of Technology; 1995.
22. Groeneboom P, Wellner JA: Information Bounds and Non-Parametric Maximum Likelihood Estimation. Berlin: Birkhäuser Verlag; 1992.
23. Prins M, Veugelers PJ, European Seroconverter Study and the Tricontinental Seroconverter Study: Comparison of progression and non-progression in injecting drug users and homosexual men with documented dates of HIV-1 seroconversion. AIDS 1997, 11:621-631.
24. Andersen PK, Borgan O, Gill RD, Keiding N: Statistical Models Based on Counting Processes. New York: Springer; 1993.
25. Anonymous: EGRET User's Manual. Seattle: Statistics and Epidemiology Research, Corp.; 1992.
26. Phillips AN, Pezzotti P, Cozzi Lepri A, Rezza G, Italian Seroconversion Study: CD4 lymphocyte count as a determinant of the time from HIV seroconversion to AIDS and death from AIDS: evidence from the Italian Seroconversion Study. AIDS 1994, 8:1299-1305.
27. Shi M, Taylor JMG, Currier RJ, et al.: Replacing time since human immunodeficiency virus infection by marker values in predicting residual time to acquired immunodeficiency syndrome diagnosis. J Acquir Immune Defic Syndr Hum Retrovirol 1996, 12:309-316.
28. Phillips AN, Sabin CA, Elford J, Bofill M, Janossy G, Lee CA: Acquired immunodeficiency syndrome (AIDS) risk in recent and long-standing human immunodeficiency virus type 1 (HIV-1)-infected patients with similar CD4 lymphocyte counts. Am J Epidemiol 1993, 138:870-878.
29. van Haastrecht HJA, Mientjes GHC, van den Hoek JAR, Coutinho RA: Death from suicide and overdose among drug injectors after disclosure of first HIV test result. AIDS 1994, 8:1721-1725.
30. van Ameijden EJC, Vlahov D, van den Hoek JAR, Flynn C, Coutinho RA: Pre-AIDS mortality and morbidity among injection drug users in cohort studies in Amsterdam and Baltimore. XI International Conference on AIDS. Vancouver, July 1996 [abstract ThC220].
31. Jones ME, Brettle RP, Busuttil A, Inglis MH, Bell JE, Willocks L: Sudden death in HIV-infected drug users. Presumptive overdose in an HIV-positive narcotic addict. J Infect 1993, 27:79-81.
32. Hammersley R, Cassidy MT, Oliver J: Drugs associated with drug-related deaths in Edinburgh and Glasgow, November 1990 to October 1992. Addiction 1995, 90:959-965.
33. Frischer M, Goldberg D, Rahman M, Berney L: Mortality and survival among a cohort of drug injectors in Glasgow, 1982-1994. Addiction 1997, 92:419-427.
34. Ancelle-Park RA: European AIDS definition. Lancet 1992, 339:671.
35. Darby SC, Ewart DW, Giangrande PLF, Spooner RJD, Rizza CR: Importance of age at infection with HIV-1 for survival and development of AIDS in UK haemophilia population. Lancet 1996, 347:1573-1579.
36. Mulder DW, Nunn AJ, Kamali A, Nakiyingi J, Wagner HU, Kengeya-Kayondo JF: Two-year HIV-1-associated mortality in a Ugandan rural population. Lancet 1994, 343:1021-1023.
© Lippincott-Raven Publishers.