Introduction
Survival studies of patients with AIDS are an important aspect of monitoring the AIDS pandemic. They provide information on the response of different subgroups to the syndrome, and enable mathematical modellers and healthcare professionals to estimate future needs.
Statistics on AIDS and deaths in Ireland are compiled by the Department of Health. The first case of AIDS in Ireland was identified in the homosexual/bisexual group in 1982. The first case of AIDS in an injecting drug user (IDU) was not found until the second half of 1985. By the end of December 1995, 498 cases of AIDS had been reported. This represents almost 31% of all known HIV-seropositive individuals. A total of 270 deaths due to AIDS occurred in Ireland up to the end of December 1995. A total of 45.5% of these were related to injecting drug use and 10% were related to those who acquired the infection through heterosexual contact.
Most cases of AIDS in Ireland up to December 1995 have occurred among IDU (213) and homosexual men (165). Haemophiliac cases (31) accounted for only 6% of cases. Heterosexual cases made up only 12% of all AIDS cases in Ireland, but they accounted for 29% of female AIDS cases, indicating that heterosexual transmission may be a particularly high-risk acquisition route for females. On the other hand, it is possible that, given the low overall numbers of women with AIDS as compared with men even after excluding homosexuals, a low rate of AIDS due to other exposures, combined with demographic factors, could explain this high percentage of female heterosexual AIDS cases. The number of reported cases of people infected through heterosexual sex doubled in the 3 years from December 1992 to December 1995, however, and many of these heterosexual cases were the sexual partners of HIV-infected IDU.
The problem of IDU acting as a 'bridge' for the virus into the non-drug using heterosexual population is a serious one in Ireland. Their significance as a high-risk group lies in the large number of young people, particularly in Dublin, who have injected drugs since heroin became widely available in the late 1970s and early 1980s [1]. The greatest percentage (57.8%) of Irish AIDS cases occur in the 25-34 year age group, with 83.3% in males and 16.7% in females of all age groups. This gives an overall male : female ratio of 5:1.
HIV infection in Ireland follows a pattern dissimilar to that in the UK but closely resembles that of Spain and Italy. These are the only other European countries to have a higher proportion of AIDS cases in the IDU risk group (66%) than in the homosexual risk group (15-16%) [2,3]. Forty-eight per cent of those who are known to be infected with HIV in Ireland (up to the end of August 1995) have been exposed to it through the intravenous use of drugs, 14% have acquired their status through heterosexual transmission, and 21% have been infected through homosexual intercourse. The other 17% is made up of children, persons with haemophilia and others [4]. In contrast, unprotected homosexual intercourse accounts for the majority (60.5%) of those who are HIV-seropositive in the UK [5].
Once AIDS is diagnosed, the life expectancy of patients is relatively short, although it depends on factors such as age, risk group, and the nature of the opportunistic infection or cancer that triggered the onset of the disease. Most studies to date, of the survival times of AIDS patients have concentrated on homosexual men. In this study, the pattern of survival for patients diagnosed with AIDS in Ireland, the majority of whom are IDU, was analysed. The cohort of 193 patients represents 61% of all those who were diagnosed with AIDS in Ireland up to March 1993, and approximately 65% of the non-haemophiliac cases. The relative influence of sex, age group, risk group, manifestation of disease at diagnosis, and treatment with antiretroviral therapy on survival were also assessed.
Methods and subjects
Data were analysed using the statistical package SAS [6]. Survival curves were investigated using non-parametric methods to obtain life-table or actuarial estimates of survival. The variables of interest were also tested (using the Wilcoxon procedure) for their association with each other. A proportional hazards model was also used to investigate further independent predictors of survival among the covariates [7].
Median conditional survival is defined as the time at which the cumulative probability of survival is 50% (i.e., the expected median survival if all subjects in the cohort were followed until death and no new cases of AIDS were diagnosed). This is distinct from the median observed survival, which is the median value in the distribution of the lengths of survival. Because of the rapid increase in cases from 1989 onwards, the median observed length of survival in the cohort (full group) was shorter than the expected value (Table 1).
All percentages quoted refer to the group as a whole unless otherwise stated. Reporting of AIDS in Ireland began in 1982 and was based on the clinical case definition established by the Centres for Disease Control (CDC), which was revised in 1985 and in 1987 and further revised in 1993.
The study cohort was made up of 193 patients of St James' Hospital, Dublin, who were diagnosed as having AIDS, between 1986 and the 15 March 1993. (Between 1982 and 1985, there were only 11 cases of AIDS reported in Ireland.) Omitted from the cohort were five patients whose details were in question or unverified. Midpoint intervals were assumed in year and month of diagnosis of AIDS and of death if data were missing (29 subjects with incomplete data). Data were censored in the case of two patients who had not died of AIDS related illnesses, and for two others whose cause of death was unknown. Patients who were alive at the end of the study (30 September 1994) and who had visited the hospital since the 30 September 1993 were assumed to be alive and were censored at the study end date. Patients who were not known to be dead but who had not visited the hospital since before the 30 September 1993 were censored at the midpoint between the date of their last visit and the 30 September 1994. There were 19 patients in this group. Assumption of uniform distribution of death and censoring is clearly not ideal, but was considered reasonable from knowledge of the database, in this instance.
The following major variables were evaluated: sex; age at diagnosis (< 30, 30-34, 35-39, or ≥ 40 years); risk group (indicating probable route of acquisition: injecting drug use, homo/bisexual activity only, injecting drug use and homo/bisexual activity or other risk factor); treatment; year of diagnosis; and manifestations of AIDS at diagnosis. Manifestation of disease at diagnosis was categorised in two different ways:
(i) Using the criteria defined in the protocol of AIDS diseases developed by the CDC (designated ABC+, here), we categorise the cohort into six groups: A, B, C1, C1+D (where the subject is diagnosed with both a C1 and a D disease), D, and E.
(ii) For comparison with previous work [8-11], we also use the criteria used by Rothenberg et al. [9] (designated PCP+, here). These criteria divide the cohort into seven groups: Kaposi's sarcoma (KS) alone; KS and Pneumocystis carinii pneumonia (PCP); KS with other infections or conditions; PCP alone; PCP with other infections or conditions; one other infection alone, i.e., only one condition was recorded as manifestation of disease at diagnosis; and two other infections, i.e., two or more manifestations were recorded.
Both categories of manifestation of disease were based on any manifestations which were recorded within 1 month of AIDS diagnosis.
The male/female ratio of patients in the cohort was approximately 4.2 : 1 (Table 1). The mean age for men was 33 years and that for women was 30 years with all patients between 21 and 55 years of age at the time of diagnosis. The mean age for the total group was 32 and the median age was 31, which is younger than those of all other comparative studies with the exception of a median age of 30 years for Italian and Spanish patients in the AIDS in Europe Study [2,8-14] (see Discussion). The largest age group comprised subjects under 30 years of age, accounting for 38.9% of the cases. A total of 69.3% of all those under 30 years of age were intravenous drug users. The smallest grouping comprised 23 subjects who were 40 years of age and older (11.9%), 21 of whom were homo/bisexual, and 18 of whom died within the study period. The mean age of those in the 40+ age group was 46 years.
With respect to risk group we found that 50.8% of the total group were IDU, 37.3% were homo/bisexual men, 3.6% were both IDU and homo/bisexual men, and 8.3% were classified as 'other', (16 subjects, 10 of whom were female). A total of 83.7% of IDU were under 35 years of age, whereas only 43% of homo/bisexual men were aged less than 35 years. Indeed 29.2% of homo/bisexual men were over 40 years of age (the comparable proportion for IDU was only 2%). The mean age for those whose risk was homo/bisexual activity was 36 years whereas that of IDU was 30 years. Homo/bisexual men therefore tended to be older than IDU. When risk groups other than IDU alone and homo/bisexual activity alone were omitted from the analysis homo/bisexual men were significantly older than IDU [n = 170; χ2 = 37.2; degrees of freedom (df) = 3; P < 0.0001]. Table 2 gives details for age of the different risk groups divided by gender.
Regarding manifestation of disease at diagnosis, PCP was the manifestation with the highest frequency (37%), followed by oesophageal candidiasis in 13% of cases. A detailed protocol definition of AIDS diseases divides patients into five subgroups: A, B, C1, D and E. Subgroup A includes patients with constitutional disease such as wasting syndrome due to HIV. Subgroup B includes patients with neurological disease such as HIV encephalopathy. Subgroup C1 includes patients with symptomatic or invasive disease due to one or more secondary infectious diseases, such as oesophageal candidiasis, PCP or cryptococcosis. Subgroup D includes patients with secondary cancers such as KS and subgroup E includes patients with other clinical findings or diseases, such as cervical cancer. Table 1 shows the distribution of patients with AIDS according to disease group (ABC+) and length of survival.
Using the alternative Rothenberg et al. [9] (PCP+) grouping, one other disease alone was the predominant manifestation of AIDS at diagnosis (48.7% of cases), followed by PCP alone (31.6%), and KS alone (7.3%) (Table 1). The mean age of those whose manifestation of disease at diagnosis was KS either alone or with PCP (18 subjects) was 39 years. Although this contrasts with those (mean age, 32 years) with manifestations of diseases other than the above, it should be noted that the number of KS patients is small; all are homosexual men, with the mean age in this group higher than in the cohort as a whole.
Although diagnosis of AIDS in the cohort began in 1986, only 4.7% of cases were diagnosed before 1988. Between 1989 and 1992, 76.2% of the cohort were diagnosed.
Results
The crude mortality ratio in the cohort was 79.3% after 7.3 years. Sixty-eight per cent of those who were aged 35-39 years had died by the study end date compared with more than 78% of those in each of the other age groups. Overall, 14 persons (7.3% of the full cohort) died within 1 month of diagnosis of AIDS with the percentage of women dying within this short period higher than that of men [five (13.5%) of all women versus nine (5.8%) of all men]. Only three of the 14 were on antiretroviral therapy. Severity of disease at diagnosis was not clear from the data set. Numbers in this group are clearly very small.
Within the first 6 months of diagnosis 37 persons, (19.2% of the full cohort) had died. Of these 37 individuals, 17 (46%) were aged between 30 and 34 years (comprising just under a third of the total cohort aged 30-34), nine (24.3%) were 40 years of age and older, 21 (56.8%) were IDU, nearly a quarter (24.3%) had been diagnosed with a disease from group D, and 21 (56.8%) were on antiretroviral therapy.
Thirty subjects (15.5% of the total group) survived for more than 3 years. Twenty-six of these were male giving a male : female ratio of 6.5 : 1 as compared with 4.2 : 1 for the total group, and half were aged less than 30 years as compared with 38.9% of the total group. Only 13.3% of those in the group, who survived for more than 3 years were aged 30-34 as compared with 28% of the total group.
Thirteen subjects (6.7% of the total group), 10 of whom were male and six of whom were under 30 years of age, survived for more than 4 years. Only one patient survived long enough to be censored at the study endpoint (7.3 years after AIDS diagnosis). Only five (2.6% of the total group) patients survived more than 5 years after AIDS was diagnosed, all of whom were male.
The life-table method provides less crude estimates of long-term survival (Fig. 1). The cumulative probability of survival (mean ± SE) at 1 year was 69 ± 3.3%, and at 5 years (1825 days) was 6.5 ± 2.5% (Table 3). The median conditional probability of survival from the date of diagnosis was 576 days (Table 1). Short-term and long-term survival patterns in this study were distinctive, with 'late' presentation a feature. This agrees with earlier findings on the Irish situation [15]. As currently estimated, the spectrum of survival is bounded by 14 persons who died within 1 month of diagnosis, and one person who survived past the truncation time. Within these bounds, there is considerable heterogeneity. The form of Fig. 1 suggests three main phases with the end of the second year and the end of the fourth year marking the turning points. The phases exhibit no dramatic changes over time, but do appear to indicate that failure rate was slightly more rapid in the early years than in the later years. The final phase, however, is based on survival in only 13 subjects (truncated in Fig. 1 at 2661 days).
Actuarial estimates of survival in each of the major subgroups are shown in Tables 1 and 3. Since 365-day intervals rather than calendar days were used to calculate the survival curves, the numbers of patients surviving per year are approximations. For convenience of presentation, the cumulative probability of survival and its standard error are expressed as percentages.
In this study, the survival curves vary significantly with age at the time of diagnosis of AIDS, with the median survival time ranging from 715 days in those who were aged 35-39 years to 331 days in those who were 40 years of age or older at the time of diagnosis. The subgroup of those patients aged 35-39 years had a more favourable survival pattern with a cumulative probability of survival of 80.25 ± 6.3% at 1 year, when compared with all other patients (P < 0.04, Wilcoxon test of equality over strata). Fig. 2 shows survival stratified by age.
Sex differences are apparent primarily at larger survival times, i.e., no women survived after the fifth year. The difference is not statistically significant, however. There were also no significant differences in survival when patients were stratified by risk group.
The PCP+ classification indicated that subjects whose manifestation of disease was PCP alone (61 subjects) had the most favourable survival pattern, with a median survival time of 696 days and a cumulative probability of survival of 85.0 ± 4.6% at 1 year. This is in agreement with recent findings on improved survival patterns in recent years for patients diagnosed with PCP [2,9,10,12,16]. On the other hand, subjects whose disease at diagnosis was KS either alone or with PCP had the least favourable prognosis (18 subjects, median 363 days, cumulative probability of survival at 1 year 50.0 ± 11.8%). The above difference in median survival times was not statistically significant, however.
Results for the differences in survival time between subjects with KS and those with other diagnostic features were contrary to those found in other studies. In this study KS is the manifestation with the second shortest survival time. Reeves and Overton [8] and Rothenberg et al. [9] agree that KS is the manifestation with the longest survival time between diagnosis and death. The discrepancies between this study and the aforementioned may be explained by the tendency of homosexuals with AIDS to get KS and the older age of homosexual subjects in this study (see patient details). Patients with KS in this case were generally older than those with other manifestations of disease. The anomaly may also be partly explained by the fact that only 19 subjects (9.8%) had KS recorded as a manifestation of disease, a much smaller proportion than in other studies (see Discussion).
The cumulative probability of survival at 1 year among patients, aged 35-39 years, who were initially diagnosed with an opportunistic disease was 84.1 ± 6.5% (P < 0.03, Wilcoxon test of equality over strata), with a median conditional survival of 956 days. This is significantly better than among the remainder of the cohort whose corresponding value was 66.0 ± 3.8%, with a median conditional survival of 547 days.
Zidovudine treatment of AIDS patients was licensed in 1987. Survival for 167 subjects who were treated with zidovudine (ZDV) or didanosine (ddI) or zalcitabine (ddC) (only one subject) at some point during their illness was compared with survival for 26 subjects who were not receiving any antiretroviral therapy (11 of whom died within 1 month of diagnosis of AIDS, and one of whom was treated with co-trimoxazole). Information on two patients was not available. Selection procedures and state of health of individuals on therapy were not specified in the data set, and so it is uncertain whether the two groups (i.e., those on treatment and others) are comparable. The tendency towards late presentation, earlier noted in Irish AIDS patients, must raise a question here.
Survival is calculated without regard to actual duration of therapy. Therapy with ZDV, ddI or ddC was significantly associated with increased survival in the group as a whole and when stratified by the covariates age group, sex, and risk group (P < 0.0002, Wilcoxon test for the association of survival with covariates). There was no significant difference in survival between the sexes for those who were on therapy or for those who were not, although for the latter, women had a longer median survival time (300 days) than men (40 days). Ages were comparable in the therapy and non-therapy groups.
The influence of each of the six covariates (sex, risk group, age group, treatment, year of diagnosis and manifestation of disease at diagnosis) in the presence of each of the others was assessed in a single proportional hazards model to predict length of survival (Table 4). The model was constructed using a stepwise procedure (maximum partial likelihood ratio method) with a P value of 0.10 as the entry criterion and 0.15 as the removal criterion at each step. Differences in survival were summarized by calculating relative hazard estimates for the different levels of each variable. In this model, a relative risk of 1 and positive coefficients imply an increased risk and therefore indicates a negative relationship with survival. Antiretroviral drug treatment was the best independent predictor of survival, followed by year of diagnosis and disease group (ABC+). Disease group, however, had a P value of 0.06. Sex, manifestation of disease at diagnosis (PCP+), age group and risk group were not significant predictors and were excluded by the model.
The proportional hazards model also suggests that improved survival was associated with being on anti-retroviral therapy (but note qualifications discussed above) and having a disease from the C1 group (ABC+) (Table 5). Survival was shortest for those patients aged 30-34 and those aged over 40 years, those whose manifestation of disease at diagnosis included KS or other cancer, or those whose disease was neurological. Referent groups were those determined to have the best survival curves by the categories listed in Table 5.
In the overall analysis, the influence of each of the major variables in the presence of the others was tested. Year of diagnosis, and disease group to which the patient belonged at diagnosis, had a significant influence on the survival of the cohort (P < 0.0006 and P < 0.02 respectively, Wilcoxon test for the association of survival with covariates). They also remained significant when the cohort was stratified over sex, risk, age, and manifestation of disease groups.
Age group, disease group and year of diagnosis showed significant differences between strata (P < 0.01, P < 0.002, and P < 0.04, respectively, Wilcoxon test of equality over strata). Patients aged 35-39 years showed longer median survival times (715 days) than all other age groups (median survival = 547 days), (P < 0.04, Wilcoxon test). Subjects aged 30-34 years and those aged 40 and over had the poorest prognosis. Patients whose disease group was opportunistic disease/s alone (CDC IV group C1) had a significantly longer median survival (672 days) than all others (i.e., those whose disease was KS or other cancers, those with neurological disease or those with other diseases), with a median survival = 281 days (P < 0.0002, Wilcoxon test).
Survival did not differ significantly by sex or risk group alone, nor did it differ significantly by manifestation of disease when grouped according to criteria PCP+.
Discussion
To date, a number of studies have estimated survival among AIDS patients with overall median survival ranging from 280 days to 760 days. Survival of patients with AIDS in Ireland has not been dealt with in the published literature up to the present time. Although the AIDS epidemic is still relatively recent in Ireland, this study is exceptional in presenting results on a cohort which represents such a high proportion of all non-haemophiliac AIDS cases in a given country. Despite some differences in cohort composition for this study compared with other recent investigations, the overall survival for these data was found to be comparable to that of Cologne (Germany), the UK, and San Francisco (USA) [16-18], and shorter than that of Washington state (USA) [19], but generally within the ranges quoted for broader European studies [2] (A. Mocroft, personal communication, 1995), in particular those conducted more recently. This comparability, or lack of it, however, may only reflect the degree of maturity of HIV infection and its epidemic in the populations referred to above. Both the cumulative mortality ratio and the cumulative probability of survival, (at 12 months and 3 years) were found to agree closely with that for more recent US and European studies compared with earlier ones [2,9,10,12,20].
This study, with a view to obtaining as complete data as possible, includes diagnoses up to 1993. Survival with AIDS may have improved with experience of treatment, in particular with regard to PCP [2,9,12,16]). For those in this study who received anti-retroviral therapy, however, survival did not differ by gender which is consistent with the findings of Lemp et al. [20].
The epidemiology of AIDS in Ireland has been described [15,21,22] and has been shown to be dissimilar to that in other developed countries. It is different in terms of the distribution of cases by sex, the age groups that are affected and the risk activities that lead to infection with HIV (Tables 1 and 2).
The importance of age on overall survival and survival for various AIDS-defining diseases has been the subject of considerable recent work (U. Baslev, T. Perneger, personal communications, 1995) and provides qualified support for our findings for the Irish data. In particular, the age profile of Irish AIDS patients reflects a consistently higher percentage of young patients compared with some US, Australian and UK investigations, [8-11,13,14]. Mean and median ages are also consistently lower in this study compared with US, Australia, UK and most European countries, with the exception of Spain and Italy [2,8-11,14]. Of all the cases of AIDS in this study, 66.8% of cases were found to be under 35 years of age, 90.2% were aged 40 or younger and 97.4% were aged 50 or younger. This low proportion of older patients is not found in other studies. For example, one third of the New York City cohort [9] were aged 40 and older compared with only 11.9% of this study.
This difference in the ages of patients with AIDS in Ireland compared with the UK and the US is understandable when the method of infection of these patients is considered. The high proportion of patients whose risk activity included injecting drug use and the generally young age of IDU clearly influences the mean age of persons with AIDS in Ireland and in this study. Supportive evidence of the relatively young age of IDU with AIDS is found in Glasgow, Scotland (mean age, 21 years) [23], and Dublin, Ireland (mean age is 25 years for IDU in treatment) [24]. The Dublin report states that preliminary information from the Pompidou Group [25] shows Dublin with the lowest mean age for census clients (December 1990) at 25.7 years. Of the other cities in the study, the mean age for clients was almost 6 years older in Copenhagen, Denmark and 8 years older in Amsterdam, The Netherlands. The report points out that, although census data are not in all cases representative for cities, they do indicate a trend of older long-term users in those cities like Amsterdam, Copenhagen and Stockholm (Sweden) with a longer history of problem drug use. The report further states that available data for all treated drug users (as distinct from census clients only) show current Dublin users to be younger than most of their European counterparts, with the exception of drug users in Lisbon (Portugal), Glasgow and Edinburgh (Scotland).
Our study observed a higher rate of parenteral drug use but a smaller percentage of homosexual or bisexual men compared with the same risk factors seen in the other studies mentioned. Our cohort enumerated 91 (47.2%) active drug users and seven (3.6%) who were former drug users, in contrast with only 72 (37.3%) who described themselves as homo- or bisexual. IDU (active and former) outnumbered homo- or bisexual men by 1.36 : 1. As a corollary, the male: female ratio was approximately 4.2 : 1. This study therefore included a larger proportion of female subjects than any of the other studies [2,8-11,17,26] to which this report refers, with male: female ratios ranging from 9 : 1 to 32 : 1.
The median survival time for patients who were infected by way of homosexual practices was 548 days in this study, 517 days in Europe [2], and 578 days for those diagnosed between 1988 and 1990 in San Francisco [18]. Survival times for those infected in this way and those infected through injecting drug use were also comparable; this is in agreement with results of the recent AIDS in Europe study [2], although the effect on survival of injecting drug use-related AIDS has been subject to some contention [10,27]. It appears, therefore, that survival with AIDS in Ireland is similar to that in other developed countries, even though Ireland has one of the highest percentages of drug-related AIDS cases among those countries who report to AIDS Surveillance in Europe [28].
Marked differences in the median survival of patients whose manifestation of disease at diagnosis was KS alone were found for this study compared with others. A median survival time of 395 days was found for this group in the Irish data, compared with that of 645 days in the UK [13] and 750 days in New York City [9]. In fact KS is the manifestation with the second shortest survival time for our data, which is contrary to findings for a number of other studies, [8-10,13], but is in apparently good agreement with those of the Australian study [11]. It should be noted, however, that there were only 14 (7.3%) such patients in our investigation, 13 of whom were homo/bisexual men.
The natural history of AIDS in this study as indicated by the survival times is therefore similar to that in other Western countries, with main discrepancies as noted above. We found survival curves to vary significantly with age at the time of diagnosis of AIDS, with the median survival time ranging from 715 days in those aged 35-39 years to 331 days in those aged 40 or older. Indications that those over 40 years have the worst prognosis is supported by much previous work [8-10,20,29-32]. The poor prognosis found in this study for those aged 30-34 years is striking, however, as witnessed by the median survival time, the survival time at 1 year, and the very high mortality during the first 6 months after diagnosis for this age group. The reason is not immediately apparent, but follow-up studies may provide further insight. Overall, the range of results in previous studies may be attributed in part to factors such as a wide variation in sample size; differences in the starting date (i.e., date of diagnosis) used for survival curves; variation in the type, timing, and analysis of diagnostic categories; and differing approaches to criteria for diagnosing AIDS, including adherence to the case definition of the CDC and the use of serologic evaluation.
In summary, the patient cohort in this study comprised 65% of the total number of non-haemophiliac AIDS cases reported in Ireland up to March 1993. It differed from those of other studies in that it had a higher proportions of females (5.2%), IDU (51%), and younger subjects aged 34 years or younger, (67%). Cohort composition does not seem to have affected overall survival, with patterns found to be broadly similar to those of other developed countries, but is reflected in some unusual subsidiary features.
From a public-policy point of view, mortality information is particularly important to both healthcare planners and disease modellers, who need the information to accurately project the needs of the future AIDS population. As prophylaxis and treatment for opportunistic diseases improve, and as ever larger numbers of patients receive antiretroviral therapy with zidovudine and other drugs still under development, it can be expected that the number of persons living with severe HIV-related immunosuppression and /or AIDS will increase.
The impact of prolonged survival of people living with AIDS on the health care system needs evaluation. In particular, the observed increases in survival with AIDS, noted in most recent studies [16,17,19,26,33], mean that it is important to begin to view AIDS not as an acute condition but as a chronic condition. Thus, the development of appropriate home- and chronic-care services, and long-term HIV prevention programmes will need to be addressed.
Acknowledgements
We are grateful to M. Dorr and M. Tynan for their assistance in following up queries on the database and, in particular, we should like to acknowledge the very helpful comments and suggestions of the referees in terms of clarifying essential points of this paper.
References
1. Dean G, Bradshaw JS, Lavelle P: Drug Misuse in Ireland 1982-1983: Investigation in a North Central Dublin Area. Dublin: Medico-Social Research Board; 1983.
2. Lundgren JD, Pedersen C, Clumeck N, et al.: Survival differences in European patients with AIDS, 1979-89. BMJ 1994, 308:1068-1073.
3. European Centre for the Epidemiological Monitoring of AIDS: AIDS Surveillance in Europe. Quarterly Report No 41. Paris: Hopital National de Saint-Maurice; 31 March 1994.
4. National Virus Reference Laboratory: Monthly AIDS statistics. NVRL Report; June 1994.
5. AIDS and HIV-1 infection in the United Kingdom: monthly report. Comm Dis Rep 21 October 1994, 4 No 2.
6. SAS Institute Inc.: SAS/STAT User's Guide and SAS Technical Report P-179, Additional SAS/STAT Procedures, Release 6.03. Cary: SAS Institute Inc.; 1988.
7. Hopkins A: Survival analysis with covariates: Cox models. In BMDP Statistical Software. Edited by Dixon WJ. Berkeley: University of California Press; 1985:576-594.
8. Reeves GK, Overton SE: Preliminary survival analysis of UK AIDS data. Lancet 1988, i:880.
9. Rothenberg R, Woelfe M, Stoneburner R, Milberg J, Parker R, Truman B: Survival with the acquired immunodeficiency syndrome: experience with 5833 cases in New York City. New Engl J Med 1987, 317:1297-1303.
10. Seage GR, Oddleifson S, Carr E, et al.: Survival with AIDS in Massachusetts, 1979-1989. Am J Public Health 1993, 83:72-78.
11. Whyte BM, Swanson E, Cooper DA: Survival of patients with the acquired immunodeficiency syndrome in Australia. Med J Aust 1989, 150:358-362.
12. Lundgren JD, Barton SE, Katalama C, et al.: Changes in survival over time after a first episode of Pneumocystis carinii pneumonia for European patients with acquired immunodeficiency syndrome. Archiv Intern Med, 1995, 155:822-828.
13. Marasca G, McEvoy M: Length of survival of patients with acquired immune deficiency syndrome in the United Kingdom. BMJ 1986, 292:1727-1729.
14. Centers for Disease Control: Update: acquired immunodeficiency syndrome - United States. MMWR 1986, 35:757-766.
15. Murphy M, Misquito N, Pomeroy L, Mulcahy F: Admissions for HIV-1 related disease in a Dublin hospital 1987-1990. Int J STD AIDS 1991, 2:436-439.
16. Peters BS, Beck E J, Coleman D G, et al.: Changing disease patterns in patients with AIDS in a referral centre in the United Kingdom: the changing face of AIDS. BMJ 1991, 302:203-207.
17. Faetkenheuer G, Buehler A, Salzberger B, et al.: Survival and predictors of outcome. In 7th International Conference on AIDS. Florence, June 1991 [abstract M.C. 3202].
18. Osmond D, Shiboski S, Moss AR: Changes in AIDS survival from 1984 to 1990 in the San Francisco general hospital cohort study. In 7th International Conference on AIDS. Florence, June 1991 [abstract W.C. 43].
19. Lafferty WE, Glidden D, Hopkins SG: Survival trends of people with AIDS in Washington State. Am J Public Health 1991, 81:215-218.
20. Lemp GF, Hirozawa AM, Cohen JB, Derish PA, McKinney KC, Hernandez SR: Survival for women and men with AIDS. J Infect Dis 1992, 166:74-79.
21. Dunne MT: A Modified Mathematical Model for HIV Transmission, AIDS and Intervention Strategies in Ireland. MSc Thesis 1995, Faculty of Computing and Mathematical Sciences, Dublin City University.
22. Comiskey CM: Mathematical Models for the Transmission Dynamics of HIV and its Progression of AIDS in Ireland. PhD Thesis 1991, School of Mathematical Sciences, Dublin City University.
23. Robertson R, Bucknall A, Wiggins P: Regional variations in HIV antibody seropositivity in British intravenous drug users. Lancet 1986, i:1435-1436.
24. O'Hare A, O'Brien M: Treated Drug Misuse in the Greater Dublin Area 1991. Dublin: Health Research Board; 1993.
25. Hartnoll R: Treatment Reporting Systems and the First Treatment Demand Indicator, Progress Report, P-PG/Epid 93; 6. Strasbourg: Council of Europe; 1993.
26. Rasch MC, Weverling GJ, de Vries E, Frissen PHJ, Weigel HM: Survival differences between intravenous drug users and homosexuals with AIDS, 1992. In 8th International Conference on AIDS/3rd STD World Congress. Amsterdam, July 1992 [abstract PuC 8180].
27. Batalla J, Gatell JM, Caylla JA, Plasencia A, Jansa JM, Parellada N: Predictors of the survival of AIDS cases in Barcelona, Spain. AIDS 1989, 3:355-359.
28. European Centre for the Epidemiological Monitoring of AIDS: AIDS Surveillance in Europe. Quarterly Report 1993. Paris: Hopital National de Saint-Maurice, France; 1993, vol. 37.
29. Whitmore-Overton SE, Tillet HE, Evans BG, Allardice GM: Improved survival from diagnosis of AIDS in adult cases in the United Kingdom and bias due to reporting delays. AIDS 1993, 7:415-420.
30. Altes J, Riera M, Villalonga C, Salas A, Rifa J, Pico G: Survival with AIDS in Balearic Islands, Spain (1985-1990). In 8th International Conference on AIDS/3rd STD World Congress. Amsterdam, July 1992 [abstract PuC 8195].
31. d'Arminio Monforte A, Mainini F, Bini T, et al.: Survival differences for 547 AIDS cases in Milan. J Acquir Immune Defic Syndr 1992, 5:1276-1277.
32. Morlat Philippe, Chene G, Laxoste D: Are opportunistic infections independent prognostic factors of survival in AIDS patients? A cohort study, Bordeaux, 1985-1990. In 7th International Conference on AIDS. Florence, June 1991 [abstract M.C. 3190].
33. Merrick S, Walsh J, McKinley P, Giordano M, Jacobs J: Gender and stage at presentation as factors in the clinical course of AIDS. In 8th International Conference on AIDS/3rd STD World Congress. Amsterdam, July 1992 [abstract PuC 8139].
© Lippincott-Raven Publishers.