The NA-ACCORD is the largest collaborative study of cohorts of HIV-infected persons receiving clinical care in the United States. We estimate that approximately 3% of PLWH-US are enrolled in clinical cohort studies that participate in the NA-ACCORD and that these NA-ACCORD participants are demographically similar to PLWH-US as reported to the CDC from 40 states with stable HIV infection reporting systems as of 2008. Describing longitudinal trends in HIV treatment and related health outcomes in a population similar to that of PLWH-US demonstrates that use of NA-ACCORD data to fill gaps in existing knowledge, together with national surveillance and cross-sectional surveys, can inform progress toward national HIV goals.
Distributions of age structure, percentage of women, and race and ethnicity were qualitatively similar to PLWH-US; because of the large numbers, even small differences were likely to be statistically significant. However, IDUs are slightly underrepresented in the NA-ACCORD compared with PLWH-US (a difference of 8 percentage points). Twelve percent of NA-ACCORD participants could not be classified into an HIV transmission risk group; some of these individuals were likely to be IDUs
Trends in HIV VL could be affected by this underrepresentation because IDUs tend to have poorer clinical response to treatment (21
). The CD4 cell count at death could also be affected because of the increased risk for death at higher CD4 cell counts among IDUs than among other HIV transmission risk groups (24
). Another potential difference to note is geographical coverage of the NA-ACCORD and adjusted CDC surveillance data. NA-ACCORD participants do not live in all 50 states. Adjusted CDC surveillance data for PLWH-US are missing for many states, some of which (such as California) contain large fractions of PLWH-US.
According to the cumulative estimated number of AIDS diagnoses, an estimate for which complete and stable national reporting exists, as of 2009 the 40 states for which adjusted data were available represent approximately 75% of AIDS diagnoses in the 50 states and the District of Columbia (18
). The CDC will report adjusted data on PLWH-US for all states in its 2011 HIV surveillance report, and our comparison can then be updated accordingly. However, these additional national surveillance data on HIV infection should not meaningfully alter our comparison.
We acknowledge that clinical cohorts in the NA-ACCORD do not include HIV-infected adults who are not receiving HIV clinical care. An estimated 75% of PLWH-US successfully link into HIV care within a year of receiving their HIV diagnosis, and 90% link into HIV care within 3 to 5 years after diagnosis (25
). Although only 45% are estimated to be engaged in regular care (≥2 visits annually, ≥3 months apart) (26
), many patients reengage in care sporadically (25
). Thus, we believe that by virtue of its size and demographic characteristics, the NA-ACCORD provides and will continue to provide the most generalizable cohort data available about the clinical epidemiology of Americans in care for HIV infection. In addition, the NA-ACCORD is especially well-positioned to monitor longitudinal trends in the use of and response to HIV therapy, as well as the adoption of HIV-related guidelines and other quality-of-care standards.
We note that our estimate of persons receiving any ART in 2007 was 81%, whereas the self-reported estimate was 85% from the 3643 persons included in the 2007 cycle of the MMP (27
). Although NNRTI-based regimens were consistently the initial therapies of choice from 2000 to 2008, most patients were prescribed PI-based regimens in each of these years. To our knowledge, our findings provide the first publicly available data on national trends in use of ART. These data, in conjunction with those on health outcomes from NA-ACCORD, can be used for clinical and cost-effectiveness modeling to inform improvements in HIV treatment, both now and when future antiretroviral agents become available.
Population-level ecological analyses from San Francisco (28
) and British Columbia (29
) have reported that decreases in new HIV infections paralleled increases in the fractions of HIV-infected patients in these jurisdictions who were treated and who achieved virologic suppression. Our analysis of clinical cohort data demonstrates at a national level that increasing use of HAART is accompanied by decreasing HIV VL in the observed population; such decreases have been shown definitively to reduce the amount of virus in a patient’s plasma and genital secretions (30
) and substantially reduce individual transmission events and new infections (31
). Data from the NA-ACCORD on HIV viral suppression can inform models of transmission dynamics.
The CD4 cell count is an established and useful barometer of immune status in HIV-infected adults. Many recent reports, including previous ones from the NA-ACCORD, have described trends in CD4 cell counts among persons in care in the United States. They usually focus on CD4 cell counts at the time of HIV diagnosis (32
) or at presentation to care (7
) or while the patient is receiving treatment (37
). Although change in mortality is the gold standard for assessing the effects of morbidity and quality of care, we are not aware of any surveillance reports or reports from larger U.S. cohort studies that have looked at CD4 cell count at death. At a national level, we have found a steady increase in immune system preservation at death. We can infer from this finding that immune suppression and associated conditions (opportunistic illnesses) are probably contributing less to mortality, as has been observed in smaller U.S. cohort studies (41
) that have reported a shift in mortality from AIDS-related to non–AIDS-related causes. When we compared deaths in 2000 or 2001 with those in 2007 or 2008 among our study participants, 60% and 53%, respectively, had information about the cause of death. From these limited data, the proportion of deaths with a primary or underlying cause that was not associated with an AIDS-defining illness increased from 52% to 58% (P
= 0.001). Additional data are being collected for a more comprehensive examination of causes of death in the NA-ACCORD.
Our analysis has limitations. First, the NA-ACCORD is limited to adults with HIV infection—adolescent and pediatric information is not available. Second, as a longitudinal cohort study, our observations are obtained from a convenience sample; however, we believe our sample is sufficiently similar in its demographic diversity to the population of PLWH-US and is sufficiently large to adequately represent for monitoring trends. Third, participants were not classified into the multiple transmission risk category of MSM and IDU. All NA-ACCORD participating cohorts contribute a primary single HIV transmission risk category; some cohorts provide additional information on multiple risks for HIV transmission. Participants who are both MSM and IDUs may have been at higher risk for negative health outcomes than those with either risk alone.
Finally, the 3-fold increase in CD4 cell count at death may have been due to a cohort effect. The clinical cohorts contributing to our analyses are dynamic (participants entered and left clinical care during the study period), which reduces the influence of a cohort effect compared with a closed cohort. We observed higher mortality rates among participants with a CD4 cell count less than 0.200 × 109
cells/L, regardless of age; however, differences in mortality rates by CD4 count stratum increased with age (data not shown) (Appendix
, available at www.annals.org
). Thus, the increase in CD4 cell count at death that we observed was probably not due to participants with lower CD4 cell counts dying at younger ages and thereby enriching the remaining cohort with survivors who had higher counts. The Appendix
presents additional data.
Our study’s strengths include the low cost of monitoring trends in HIV treatment and related outcomes in the NA-ACCORD. Participation in this cohort requires no new data collection (only standardized collation of existing data), which lowers the operating cost of the study, speeds dissemination of summary data to within 1 to 4 years after collection of the primary data, and reduces barriers to participation by other cohorts. Cohorts of HIV-infected adults can join the NA-ACCORD at any time, and membership continues to grow. Having more participants will probably increase the generalizability of our collaborative findings. The characteristics of NA-ACCORD participants and PLWH-US will continue to be monitored, and methods will be employed as needed to improve the generalizability of findings to specified target populations (46
). Although cross-sectional probability surveys are designed to produce a representative sample, differential nonparticipation requires statistical correction and greater operating expense and longitudinal data are not available to monitor trends. We believe that longitudinal cohort studies and cross-sectional probability surveys provide complementary information and that both are necessary to fully characterize the U.S. HIV epidemic and inform public health action to optimize the morbidity and mortality of Americans living with HIV infection.
In summary, the NA-ACCORD is uniquely positioned to provide timely longitudinal data on the clinical epidemiology and health of adults living with HIV infection in the United States. To our knowledge, our analysis is the first to provide data about national trends in antiretroviral prescription. We show that from 2000 to 2008, the percentage of U.S. participants in clinical care who were prescribed HAART increased, as did the percentage of all patients who achieved a suppressed HIV VL. Simultaneously, the median CD4 cell count at death increased by 0.149 × 109
cells/L to greater than 0.200 × 109
cells/L. New cohort studies continue to join the NA-ACCORD. We expect their addition to increase the similarity of NA-ACCORD participants to a national probability sample of persons living with HIV infection, and that this collaboration will remain an important means of monitoring trends that document efforts to improve health outcomes for these persons, as articulated in the National HIV/AIDS Strategy (5