A novel use of AIDS surveillance data to assess the impact of initial treatment regimen on survival

Abstract

The expense in time and money limit the use of randomized clinical trials (RCT) and cohort studies for evaluating long-term AIDS treatment outcomes. We conducted a case-control study to characterize predictors of AIDS mortality after the availability of highly active antiretroviral therapy (HAART) in San Francisco, in which cases were matched with controls on stage of disease, year of AIDS diagnosis and year of HAART initiation. Overall, 266 cases and 1173 controls were included, representing >90% of eligible patients. The class of initial HAART regimen was not associated with mortality. Predictors of mortality were older age ([adjusted odds ratio] AOR 1.23, 95% [confidence interval] CI: 1.13–1.35), public versus private health insurance (AOR 2.80, 95% CI: 1.77–4.42), no versus private insurance (AOR 1.45, 95% CI: 1.02–2.07) and unboosted saquinavir (AOR 2.50, 95% CI: 1.34–4.65). Survival benefit was found in following the 2004 US Department of Health and Human Services preferred treatment guidelines; borderline survival benefits were found in following the guidelines from other years. Similar predictors were found for all-cause and AIDS-specific mortality. Our findings mirrored those of RCT and multi-centre cohort studies, and may be applicable to other settings. Our findings support similar survival benefit to persons initiating HAART with non-nucleoside reverse transcriptase inhibitor- or protease inhibitor-based regimens.

Keywords

surveillance AIDS mortality antiretroviral therapy San Francisco

INTRODUCTION

As of June 2006, the Global Fund for AIDS, Tuberculosis and Malaria provided antiretroviral therapy (ART) to 544,000 people worldwide, with plans to treat 1.6 million by 2009. The US President's Emergency Plan for AIDS Relief (PEPFAR) supported ART for an additional 561,000 people.¹ After five years of expanding access to ART, international agencies are now concerned with evaluating the impact of these efforts on survival. To accomplish this, a low-cost, rapid method to assess the impact of ART on survival is needed.

Despite the need to assess the impact of ART on survival, the costs, time and complexity of conducting randomized clinical trials (RCTs) and cohort studies preclude their frequent and routine use. Few countries targeted by the Global Fund and PEPFAR have conducted RCT or have on-going cohort studies. Moreover, the selection and participation biases inherent in prospective studies limit their generalizability over wide geographic areas, and the use of survival or death as an endpoint is increasingly unfeasible and unethical.

We sought to develop, apply and evaluate a case-control method to assess factors associated with survival using routine AIDS surveillance data and existing sources of vital statistics. In addition to the low cost, this method provides locally relevant information because of the geographic and population basis of these data sources. The objectives were to develop a case-control method, assess predictors of AIDS mortality and compare the results of this approach with RCT, cohort studies and national treatment guidelines. In addition, we sought to test the hypothesis that the initial highly active antiretroviral therapy (HAART) regimen is associated with survival.

METHODS

We conducted a retrospective, matched case-control study using records from the San Francisco (SF) HIV/AIDS Reporting System (HARS) database supplemented with data from local and national vital statistic registries and medical record abstraction. The target population was adults diagnosed with AIDS and treated in SF after the availability of HAART (1996–2004). The primary outcome was death and the main predictor was the initial class of drugs comprising the HAART regimen.

Selection of cases and controls

Study subjects were persons diagnosed with AIDS and reported to the San Francisco Department of Public Health (SFDPH) as of 25 May 2006 using the current Centers for Disease Control and Prevention surveillance case definition: a positive HIV serological test plus either a CD4⁺ T-cell count <200 cells/μL or at least one of 26 AIDS-defining illnesses.² AIDS case reporting was estimated to be >95% complete with a median delay from diagnosis to reporting of two months.³ Eligibility criteria were as follows: (1) at least 13 years of age; (2) resident of SF at the time of AIDS diagnosis: (3) initiation of HAART between 1996 and 2004; (4) use of HAART for at least seven days – this is the minimum time period for treatment and is based on practices of short structure intermittent therapy⁴; (5) initiation of HAART after AIDS diagnosis; and (6) at least one CD4⁺ T-cell count within six months before initiation of HAART. Cases were persons meeting the above criteria documented to have died before 31 December 2004. Controls were persons meeting the eligibility criteria documented to be alive as of 31 December 2004. We selected a date for assessing the availability of data (25 May 2006) beyond the end of the study period (31 December 2004) to minimize misclassification due to delays in reporting AIDS cases and deaths.

Matching

Cases and controls were matched on the year of AIDS diagnosis, year of HAART initiation and level of CD4⁺ T-cell count within six months prior to treatment initiation (<200 versus 201–350 versus >350 µL/cells). The rationale for these matching criteria was to ensure that cases and controls would be as similar as possible with respect to the stage of disease at diagnosis and would have initiated HAART at approximately the same time. Matching on time helps ensure that the same treatment options were available to both cases and controls. We matched all cases to all available controls in the same stratum; thus, cases were compared with all controls with the same disease severity and treatment options.

Measures

Mortality

We examined death due to all causes, death with AIDS listed as the primary cause and death with AIDS listed as a contributing cause. Given no substantial differences in findings with these different outcomes, we report here the results of all-cause mortality analyses. Vital status was determined by SFDPH personnel using four distinct methods: prospective medical chart reviews, bi-monthly review of the city's vital statistics records, periodic computerized matches with the California and National Death Index (most recently through 2004), and weekly review of obituaries in local newspapers. We collected primary and contributing causes of deaths using the International Statistical Classification of Disease and Related Health Problems, 9th revision (ICD-9) for deaths occurring prior to 1999 and the 10th revision (ICD-10) for deaths occurring in 1999 or after. Deaths attributable to HIV-1 infection or AIDS were coded as 042–044 under ICD-9 and as B20–B24 under ICD-10. AIDS opportunistic diseases, if listed on death certificates, were classified as AIDS related (this prevented misclassification of AIDS deaths as non-AIDS related if ‘AIDS’ was not specifically listed on death certificates).

HAART regimens

Based on the US Department of Health and Human Services (DHHS) guidelines, we defined initiating HAART regimens as follows: (a) any regimen containing at least one non-nucleoside reverse transcriptase inhibitor (NNRTI) in combination with nucleoside reverse transcriptase inhibitors (NRTIs) in the absence of protease inhibitors (PIs); (b) any regimen containing at least one PI in combination with NRTIs in the absence of NNRTIs; (c) any regimen containing at least one NNRTI and one PI in combination with NRTIs; or (d) any regimen containing abacavir (ABC) or tenofovir (TDF) in combination with at least two NRTIs in the absence of both PIs and NNRTIs.⁵ Given the large number of HAART regimens, we collapsed the drug regimens into four classes (NNRTI based versus PI based versus NNRTI plus PI based versus NRTI based). We also collected data on whether a PI-based regimen was unboosted or boosted (i.e. any regimen containing lopinavir/ritonavir (LPV/r) or low-dose ritonavir (RTV) plus another PI was considered boosted).

Data collection

Data were collected in two stages. The first stage comprised data extraction from HARS, including demographic characteristics, HIV-1 transmission risk factors and clinical information. The second stage was performed for quality control. We re-abstracted data on key variables from a random sample of 10% of the study population. These quality control activities were conducted between October and December 2006. Overall, we found a 2.1% misclassification of the year of HAART initiation and a 2.1% misclassification of the initiating drug class. No cases or controls were misclassified.

Statistical analyses

We used conditional logistic regression analysis to test predictors of survival while adjusting for potential confounders and conditioning on matching variables. Bivariate associations of specific risk factors with case-control status were assessed and all significant variables (P < 0.05) were entered into a multivariate model to assess independent associations and potential confounding. Potential effect modifiers were investigated and identified by stratified analysis. A full model with significant covariate and confounding variables was used to test the following hypotheses: (1) whether the initial treatment regimen was associated with survival; (2) whether there was evidence of survival benefit in initiating a preferred versus non-preferred regimen according to US DHHS guidelines; and (3) whether evidence of survival benefit could be observed in direct comparison of efavirenz (EFV) versus nevirapine (NVP)-based regimens. For the second hypothesis, we classified each initial HAART regimen as a preferred regimen or non-preferred regimen based on the year it was prescribed and the DHHS guidelines for initial therapy in effect for that same year. Testing of these hypotheses was repeated using only AIDS-related causes of death as the outcome to assess biases due to the inclusion of non-AIDS-related deaths. Analyses were conducted using STATA (version 8.0, Stata Corp., College Station, TX, USA).

RESULTS

Of 4350 adult AIDS cases diagnosed in SF between 1996 and 2004, 3698 (85%) had confirmed vital status as of 31 December 2004 with follow-up information on death available through 25 May 2006. A total of 703 AIDS patients died before 31 December 2004 and 2995 were known to be alive as of 31 December 2004. The remaining 652 (15.0%) had unknown vital status (lost to follow-up). A total of 1575 individuals (274 cases and 1301 controls) met study eligibility criteria (Figure 1).

Figure 1

Selection of cases and controls from AIDS cases reported to the San Francisco Department of Public Health HIV/AIDS registry

We successfully matched 1439 individuals, including 266 cases (97.1% of all eligibles) and 1173 controls (90.2% of all eligibles), into 50 strata based on year of AIDS diagnosis, year of HAART initiation and level of CD4⁺ T-cell count before initiation of HAART. There was an average of 4.4 controls per case (range: 0.3–31 controls per case). A total of 128 (9.8%) eligible controls and eight (2.3%) eligible cases were excluded from the analysis because an appropriate match was not available.

Of 266 cases, the median age was 43 years, 89.1% were men, 56.8% were White, 62.8% were gay or bisexual, and 39.9% were uninsured (Table 1). Of 1173 controls, the median age was 37 years, 92.7% were men, 57.8% were White, 76.1% were gay or bisexual, and 51.1% had private insurance. By design, cases and controls were comparable on year of AIDS diagnosis, year of HAART initiation and level of CD4⁺ T-cell count within six months prior to treatment initiation.

Table 1

Demographic characteristics, HIV-1 transmission risk factors, AIDS care, AIDS diagnosis and initial antiretroviral therapy of matched cases and controls, San Francisco, 1996–2004 (n = 1439)

Characteristics	Cases n = 266 No. (%)	Controls n = 1173 No. (%)	Crude OR (95% CI)
Age at HIV/AIDS diagnosis (years)
17–25	11 (4.1)	27 (2.3)	0.65 (0.26–1.62)
26–30	17 (6.4)	114 (9.7)	0.20 (0.10–0.42)
31–35	38 (14.3)	254 (21.7)	0.21 (0.11–0.39)
36–40	52 (19.6)	284 (24.2)	0.27 (0.15–0.48)
41–45	52 (19.6)	216 (18.4)	0.35 (0.19–0.64)
46–50	40 (15.0)	156 (13.3)	0.36 (0.19–0.68)
51–55	29 (10.9)	76 (6.5)	0.56 (0.28–1.13)
56 +	27 (10.2)	46 (3.9)	1.0
Gender
Male	237 (89.1)	1088 (92.7)	1.0
Female	20 (7.5)	64 (5.5)	1.51 (0.87–2.61)
Male-to-female	9 (3.4)	21 (1.8)	1.74 (0.75–4.03)
Race
White	151 (56.8)	678 (57.8)	1.0
Hispanic	37 (13.9)	223 (19.0)	0.93 (0.62–1.41)
Black	61 (22.9)	177 (15.1)	1.74 (1.21–2.49)
API	14 (5.3)	82 (7.0)	0.87 (0.47–1.60)
Native American	1 (0.4)	9 (0.8)	0.43 (0.05–3.54)
Other	2 (0.8)	4 (0.3)	4.77 (0.73–31.14)
Primary risk factor
Gay/bisexual	167 (62.8)	893 (76.1)	1.0
IDU	43 (16.2)	86 (7.3)	2.61 (1.69–4.03)
Gay/bisexual and IDU	43 (16.2)	132 (11.2)	1.65 (1.10–2.48)
Heterosexual	5 (1.9)	34 (3.0)	0.96 (0.36–2.58)
Unknown	8 (3.0)	28 (2.4)	2.14 (0.92–4.97)
Insurance
Private	98 (36.8)	600 (51.1)	1.0
Public	62 (23.3)	118 (10.1)	3.47 (2.32–5.19)
No coverage	106 (39.9)	455 (38.8)	1.46 (1.06–2.01)
Diagnosing facility
Private	115 (43.2)	611 (52.1)	1.0
Public	121 (45.5)	487 (41.5)	1.39 (1.03–1.87)
University/military	23 (8.7)	72 (6.1)	2.02 (1.18–3.45)
Out of jurisdiction	7 (2.6)	3 (0.3)	11.70 (2.89–47.29)
Initial primary AIDS diagnosis
CD4 count <200 or <14%	193 (72.6)	920 (78.4)	1.0
AIDS defining event	73 (27.4)	253 (21.6)	1.27 (0.92–1.75)
Tuberculosis prophylaxis	98 (36.8)	370 (31.5)	1.35 (1.00–1.82)
Initial antiretroviral therapy
NNRTI based	58 (21.8)	435 (37.1)	1.0
PI based	194 (72.9)	681 (58.1)	1.55 (1.07–2.24)
NNRTI + PI based	8 (3.0)	39 (3.3)	1.41 (0.61–3.29)
NRTIs only	6 (2.3)	18 (1.5)	3.37 (1.18–9.60)

NNRTI = non-nucleoside reverse transcriptase inhibitor; PI = protease inhibitor; NRTI = nucleoside reverse transcriptase inhibitor; OR = odds ratio; CI = confidence interval; IDU = intravenous drug user

In bivariate conditional logistic regression analyses, AIDS mortality was associated with older age, African American race/ethnicity, injecting drug use, gay or bisexual injecting drug use, tuberculosis prophylaxis, public health insurance or no insurance compared with private insurance, and diagnosis at a public facility, university, military hospital or outside SF's jurisdiction. In bivariate analysis, initiating HAART with a PI-based or NRTI-only regimen was associated with increased odds of death compared with NNRTI-based regimens.

The role of confounding in the apparent association of initial HAART regimen with death was revealed in multivariate analysis (Table 2). We first examined the association between initial regimen and death controlling for older age at AIDS diagnosis, African-American race/ethnicity, injection drug user status, having public or no insurance, and tuberculosis prophylaxis (Model I). Model I showed worse odds of survival with PI-based regimens than NNRTI-based regimens. However, after adjusting for individual antiretroviral drugs (EFV, unboosted saquinavir [SQV] and LPV/r), the survival benefit of NNRTI-based regimens was no longer statistically significant and only older age at diagnosis, public or no health insurance and unboosted SQV remained significantly associated with higher mortality (Model II). That is, after controlling for unboosted SQV, no difference in survival benefit for PI versus NNRTI as initial HAART regimen was detected.

Table 2

Independent predictors of mortality among adult AIDS cases, San Francisco, 1996–2004 (n = 1439)

Characteristic	Crude OR (95% CI)	Model I AOR (95% CI)	Model II AOR (95% CI)
Age (years)
17–25	0.65 (0.26–1.62)	0.56 (0.22–1.42)	0.61 (0.24–1.56)
26–30	0.20 (0.10–.42)	0.17 (0.08–0.37)	0.19 (0.09–0.41)
31–35	0.21 (0.11–.39)	0.19 (0.10–0.36)	0.20 (0.10–0.38)
36–40	0.27 (0.15–0.48)	0.24 (0.13–0.45)	0.26 (0.14–0.49)
41–45	0.35 (0.19–0.64)	0.34 (0.19–0.63)	0.36 (19–0.67)
46–50	0.36 (0.19–0.68)	0.37 (0.20–0.71)	0.40 (0.21–0.76)
51–55	0.56 (0.28–1.13)	0.53 (0.26–1.09)	0.55 (0.27–1.14)
56 +	1.0	1.0	1.0
Gender
Female	1.51 (0.87–2.61)	–	–
Male-to-female	1.74 (0.75–4.03)	–	–
Male	1.0	–	–
African American	1.78 (1.26–2.52)	1.37 (0.94–1.99)	1.42 (0.97–2.07)
Intravenous drug use	1.95 (1.42–2.69)	1.39 (0.96–2.00)	1.33 (0.92–1.92)
Type of insurance
No insurance	1.46 (1.06–2.01)	1.50 (1.06–2.13)	1.5 (1.02–2.07)
Public	3.47 (2.32–5.19)	2.80 (1.78–4.40)	2.80 (1.77–4.42)
Private	1.0	1.0	1.0
Tuberculosis prophylaxis	1.35 (1.00–1.82)	1.17 (0.84–1.62)	1.16 (0.83–1.60)
Initial HAART regimen
NNRTI based	1.0	1.0	1.0
PI based	1.55 (1.07–2.24)	1.52 (1.03–2.24)	1.10 (0.65–1.85)
NNRTI + PI based	1.41 (0.61–3.29)	1.55 (0.64–3.76)	1.25 (0.51–3.07)
NRTIs only	3.37 (1.18–9.60)	2.58 (0.86–7.68)	2.31 (0.74–7.24)
Antiretroviral drug
Efavirenz (EFV)	0.58 (0.38–0.89)	–	0.76 (0.43–1.36)
Saquinavir (SQV)*	2.90 (1.61–5.20)	–	2.50 (1.34–4.65)
Lopinavir/ritonavir (LPV/r)	1.95 (1.10–3.47)	–	1.78 (0.91–3.47)

HAART = highly active antiretroviral therapy; NNRTI = non-nucleoside reverse transcriptase inhibitor; PI = protease inhibitor; NRTIs = nucleoside reverse transcriptase inhibitors; OR = odds ratio; CI = confidence interval; AOR = adjusted odds ratio

*Unboosted

Using our case-control method, we tested specific hypotheses that enabled comparison with recent RCT, longitudinal studies and national ART treatment guidelines (Table 3).^2,6–9 First, we tested the hypothesis that persons whose initial HAART regimens were in accordance with the preferred regimens of the US DHHS guidelines would have higher odds of survival compared with persons who initiated HAART with non-preferred regimens. Overall, 580 (51.5%) persons initiated HAART with a preferred regimen between 1998 and 2004. By classifying subjects according to guidelines in effect the year of treatment initiation, we found a borderline survival benefit in following DHHS guidelines (AOR 0.72, 95% CI 0.49–1.05, P = 0.09). When examining survival odds by year, we found a statistically significant survival benefit in following DHHS guidelines in 2004 (AOR 0.03, 95% CI < 0.01–0.81, P = 0.04). The adjusted OR measures from 1998 to 2003 tended towards a survival benefit in following DHHS guidelines, although no other individual year achieved statistical significance. Moreover, the odds of survival improved over time, from 0.99 in 1998 to 0.03 in 2004 (P ≤ 0.01).

Table 3

Summary of additional hypotheses tested

Hypothesis	Sub-analysis No.	AOR (95% CI) for mortality
1. Preferred Regimens* – DHHS guidelines followed according to year of initiation
1. Preferred regimen better survival than non-preferred regimen?	1127	(0.72 (0.49–1.05)
2. By year of HAART initiation
1998	184	0.99 (0.37–2.65)
1999	163	0.71 (0.29–1.74)
2000	183	0.83 (0.32–2.10)
2001	175	0.88 (0.36–2.13)
2002	136	0.37 (0.08–1.83)
2003	146	0.40 (0.09–1.81)
2004	140	0.03 (0.00–0.81)
2. EFV versus NVP
1. EFV better survival than NVP-based regimens?	527	0.87 (0.45–1.71)
2. EFV better survival than EFV plus NVP-based regimens?	355	–
3. EFV better survival than non-NNRTI-based regimens?^†	1264	0.55 (0.33–0.90)
4. EFV better survival than boosted PI-based regimens?	1257	0.71 (0.41–1.23)

DHHS=Department of Health and Human Services; HAART= highly active antiretroviral therapy; EFV = efavirenz; NVP = nevirapine; NNRTI = non-nucleoside reverse transcriptase inhibitor; PI = protease inhibitor

*Source: Guidelines for the use of antiretroviral therapy agents in HIV-infected adults and adolescents. Atlanta: US DHHS (Available at http://aidsinfo.nih.gov/Guidelines/Default.aspx?MenuItem=Guidelines)

^†Non-NNRTI-based regimens include PI and NNRTI-based regimens

The second hypothesis tested was whether persons initiating HAART with an EFV-based regimen had higher odds of survival compared with NVP-based regimens. A total of 344 subjects initiated HAART with an EFV-based regimen and 183 subjects with an NVP-based regimen. In multivariate analysis comparing EFV with NVP, we found similar odds of survival between these two NNRTI-based regimens. However, EFV-based regimens had better odds of survival when compared with non-NNRTI-based regimens (i.e. PI- or NRTI-based regimens) (AOR 0.55, 95% CI 0.33–0.90, P = 0.02), but similar odds when compared with boosted PI-based regimens (AOR 0.71, 95% CI 0.41–1.23, P = 0.22).

DISCUSSION

We developed and applied a case-control method to identify predictors of survival among persons with AIDS using routinely collected surveillance data. This practical approach can be applied widely in the developed world and may be adapted to the developing world where health information systems are being created to track persons receiving ART. We validated our approach by comparing our results with the findings of RCT, cohort studies and national treatment guidelines.^2–7

Our case-control approach confirmed the associations of AIDS mortality and older age at diagnosis and lack of heath insurance, as previously observed in cohort studies.^10–12 We also confirmed the association of African-American race/ethnicity and increased AIDS mortality found previously in SF and elsewhere in the USA.^13,14 As in these other studies, the association was confounded by insurance status, injection drug use and competing causes of death. Our finding of increased mortality with SQV is consistent with the EuroSIDA Cohort Study and the Danish Cohort Study.^3,15 In fact, the use of SQV was removed from first-line therapy due to the high pill burden and the inferior viral load suppression compared with other PIs.¹⁶ We also provide evidence that national guidelines on initial regimens are valid. We found that the US DHHS guidelines for initial regimens improved as more effective treatment options became available, from nearly equal odds of survival following the guidelines of 1998 (AOR of 0.99) to the significant survival benefit following the guidelines of 2004 (AOR of 0.03). Comparable to another recent RCT, we found no difference in the odds of survival between persons who initiated HAART with either a PI- or NNRTI-based regimen.⁵ Finally, similar to the 2NN RCT, our results showed comparable efficacy of NVP- and EFV-based HAART regimens.⁶

RCTs remain the gold standard for demonstrating drug efficacy and will be needed to evaluate new therapies. As more antiretroviral drugs become available and the number of possible different combinations grows exponentially, it will not be feasible to make comparisons of all possible combinations within RCT or longitudinal cohort studies. There are several advantages to our case-control approach: (a) it uses death as an endpoint; (b) it is rapid and inexpensive as investigators do not have to wait for deaths to accrue; (c) it uses a population base of all AIDS cases and AIDS deaths in a defined geographic area – a study population that may be more representative of the general clinic population than the participants in RCT and cohort studies; (d) it can simultaneously compare many possible antiretroviral regimens; and (e) it addresses the challenge that RCT and cohort studies are increasingly unable to follow subjects to death as survival is prolonged by increasingly effective ART. The largest observational cohort comparing risk of death by specific ART combination to date is the EuroSIDA cohort with 11,930 patients and a total of 188 deaths and 51,176 person-years of patient experience collected as of December 2005.³ The longest RCT (FIRST) ever reported has a median follow-up of five years and 125 deaths.⁴ In comparison, our case-control study included 266 deaths. Our method may be most applicable to resource-poor settings where long-term cohort studies and RCT are not feasible. In fact, comprehensive monitoring and evaluation systems are being implemented in conjunction with increased access to ART that provide quarterly counts of AIDS patients receiving ART and those known to have died. Our method can be adapted to this setting by randomly sampling deaths and comparable controls and abstracting clinical information at the reporting facilities.

A major potential limitation is the inability to adjust for all potential confounders, particularly those that are not recorded, or difficult to define. In particular, the choice of HAART regimen is not random but is based on multiple sources of information, clinical judgement and typical practices at different times. Bias may result if certain regimens are given preferentially to sicker patients. This bias would reduce the apparent benefit of these regimens in our analysis. We attempted to control for this type of bias by restriction, matching on stage of disease, and adjusting for potential confounders in the analysis phase.

This study is also limited in its implications for HIV-1-infected persons who do not have AIDS, i.e. some persons may have started HAART before the diagnosis of AIDS and would not have been included. California recently began case reporting all persons diagnosed with HIV-1 infection, as is done in most other states. Our method can also be applied to assess predictors of survival for HIV-1-infected persons who do not have AIDS.

We were not able to assess the impact of treatment adherence and subsequent changes to initial HAART regimens on survival as these measures are not consistently collected in routine AIDS surveillance. With improved data collection, future analyses may be able to investigate the impact of subsequent regimens on AIDS survival. Studies assessing the impact of switching to a sequential PI- or NNRTI-based HAART regimen on survival will enable better care of HIV-1-infected persons. Our method may also be used to assess the impact of drug resistance mutations on survival, as drug resistance testing is increasingly being performed in routine clinical care.¹⁷

In summary, we describe the novel use of surveillance data to assess the impact of initial HAART regimens on survival. Despite potential limitations, our method offers a tool for national programmes to evaluate the impact of initial HAART regimens on survival within defined populations. While our approach cannot replace RCT and cohort studies, we believe it will complement results from clinical trials and assist with reducing death and improving the quality of life of persons living with HIV/AIDS in both the USA and abroad.

Footnotes

ACKNOWLEDGEMENTS

The authors would like to thank the San Francisco AIDS surveillance staff for their hard work, and Ron Brookmeyer, Haitao Chu and Lawrence Moulton for their helpful suggestions.

Funding: Sanny Chen was supported by a pre-doctoral grant from the National Institutes of Health through the Ruth L. Kirschstein National Research Service Award (NIH: 1F31AI068607).

References

The Global Fund to fight AIDS, Tuberculosis and Malaria. Investing in Impact: Mid-year Results Report 2006. Geneva. pp. 1–80.: Available at http://www.theglobalfund.org/en/files/about/replenishment/progress_report_midyear_2006.pdf (last accessed 22 January 2007)

Castro

, Ward

, Slutsker

, Buehler

, Jaffe

, Berkelman

. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR 1992;41:1–19

Schwarcz

, Hsu

, Parisi

, Katz

. The impact of the 1993 AIDS case definition on the completeness and timeliness of AIDS surveillance. AIDS 1999;13:1109–14

Fauci

. Host factors in HIV disease. Program and Abstracts of the XIII International AIDS Conference; 9–14 July 2000; Durban, South Africa. Session TuOr36

Guidelines for the use of antiretroviral agents in HIV-infected adults and adolescents. Atlanta: US DHHS, 2004. Available at http://aidsinfo.nih.gov/contentfiles/AdultandAdolescentGL.pdf (accessed 25 January 2007)

Olsen

, Gatell

, Ledergerber

, Risk of AIDS and death at given HIV-RNA and CD4 cell counts, in relation to specific antiretroviral drugs in the regimen. AIDS 2005;19:319–30

MacArthur

, Novak

, Peng

, A comparison of three highly active antiretroviral treatment strategies consisting of non-nucleoside reverse transcriptase inhibitors, protease inhibitors, or both in the presence of nucleoside reverse transcriptase inhibitors as initial therapy (CPCRA 058 FIRST Study): a long-term randomized trial. Lancet 2006;368:2125–35

Van Leth

, Phanuphak

, Ruxrungtham

, Comparison of first-line antiretroviral therapy with regimens including Nevirapine, Efavirenz, or both drugs, plus Stavudine and Lamivudine: a randomized open-label trial, the 2NN Study. Lancet 2004;363:1253–63

The World Health Organization. HIV/AIDS Care and Treatment: Guide for Implementation. 2004 Geneva. pp. 65 – 68. Available at http://www.wpro.who.int/NR/rdonlyres/49BC4F23-0DCC-4D28-A50A-4F755BC475EE/0/HIV_AIDS_Care_and_Treatment_Guide_for_Implementation.pdf (accessed 23 May 2007)

10.

Viard

, Mocroft

, Chiesi

, Influence of age on CD4 cell recovery in human immunodeficiency virus-infected patients receiving highly active antiretroviral therapy: evidence from the EuroSIDA study. J Infect Dis 2001;183:1290–4

11.

Bacchetti

, Osmond

, Chaisson

, Survival patterns of the first 500 patients with AIDS in San Francisco. J Infect Dis 1988;157:1044–7

12.

McFarland

, Chen

, Hsu

, Schwarcz

, Katz

. Low socioeconomic status is associated with a higher rate of death in the era of highly active antiretroviral therapy, San Francisco. J Acquir Immune Defic Syndr 2003;33:96–103

13.

Jain

, Schwarcz

, Katz

, Gulati

, McFarland

. Elevated risk of death for African Americans with AIDS, San Francisco, 1996–2002. J Health Care Poor Underserved 2006;17:493–503

14.

Nash

, Katyal

, Shah

. Trends in predictors of death due to HIV-related causes among persons living with AIDS in New York City: 1993:2001. J Urban Health 2005;82:584–600

15.

Jensen-Fangel

, Pedersen

, Nielsen

Use of the protease inhibitor Saquinavir hard gel in human immunodeficiency virus-infected patients in the early period of highly active antiretroviral therapy: does it affect long-term treatment outcome?

Scand J Infect Dis 2003;35:743–9

16.

Bartlett

. In: Dunning

, ed. Pocket Guide to Adult HIV/AIDS treatment, Baltimore: Johns Hopkins University, 2005: p. 42

17.

Kuritzkes

, Boyle

, Gallant

, Squires

, Zolopa

. Current management challenges in HIV: antiretroviral resistance. AIDS Read 2003;13:133–42