Krentz, Hartmut B. PhD*,†; Gill, M. John MD*,†
The importance of a viral load (VL) in determining an HIV+ individual's infectivity was first reported by Quinn et al1 in 2000 and has since been confirmed elsewhere.2–5 These findings have led researchers to advocate the widespread and increased use of highly active antiretroviral therapy (HAART) as a leading tool to curtail the epidemic. They argue that the extensive use of HAART can reduce the number of highly infectious individuals within a community thereby decreasing new HIV infections.6–13 The concept of community viral load (CVL) has been proposed as a useful surrogate marker for the potential pool of virus in a community available for ongoing transmission and as a means to monitor HIV burden and treatment outcomes.12–16 The CVL has been defined as “an aggregate biological measure of viral load (VL) for a particular geographic location over a specific period of time.”14 Pilot studies from British Columbia,15,17 San Francisco,14 Australia,16 Washington DC,18 and Taiwan19 suggest that in highly focused populations, as CVL decreases from the increased use of HAART, there may be subsequent decreases in new HIV infections. However, as there is often a long time lag between HIV infections and diagnostic testing, tracking new HIV diagnoses and/or infections may be an insensitive method for monitoring the effectiveness of such an approach. The evaluation and validation of CVL to measure the effectiveness of such programs are critical.20
Operationalizing CVL can be difficult.20 The CVL has a number of implicit and explicit assumptions (Table 1) that may or may not be supportable. In any analysis, details on what the CVL represents, which is included in a “community,” how CVL is measured, and over what time period and how “new” HIV diagnoses are defined are all essential study parameters. Recently, Miller et al20 categorized 4 theoretical limitations and potential biases: issues of selection and measurement, the importance of HIV prevalence in determining ongoing HIV transmission, the interpretation of CVL itself, and the ecological fallacy. In addition, the effects of “churn”21,22 (ie, the movement of individuals in and out of a community) also need to be considered as such a movement may alter the composition and integrity of the CVL measurement.
Although the concept of CVL has been discussed since 2000,1 it has only been recently that attempts have been made to standardize the definition. In 2012, the Centers for Disease Control and Prevention (CDC) released their Guidance on Community Viral Load: Measure, Definitions, and Methods for Calculation23 that provided definitions to describe CVL. They proposed 4 measures (ie, Population VL, CVL, In-Care VL, and Monitored VL), and 5 component measures (ie, in-care with undetectable VL, in-care with detectable VL, in-care but no VL, diagnosed but not in-care, and undiagnosed). However, the CDC guidelines caution that, though comprehensive, the first 3 population measures may not be feasible for widespread use due to the difficulty in obtaining reliable VL measures on a consistent basis. They state that missing VL data need to be imputed into the calculations for CVL using supplementary data that are not available in most jurisdictions. Thus, despite these detailed guidelines, operationalizing a workable definition of what CVL actually is, and how it can be effectively used, remains somewhat elusive.
In this study, we document CVL, HAART uptake, levels of VL suppression, and the incidence of new HIV diagnoses over a 10-year period, and then examine CVL in more detail over 12 months (ie, 2009) in our geographically defined population. We examine the effect of churn (ie, patient movement in and out of care) on CVL measurements and new HIV diagnoses to evaluate how movement can influence the interpretation of CVL within any community or population.
Definition and Use of CVL
We use the concept of CVL colloquially rather than using the strict CDC guideline definition to discuss more easily the effects of churn on CVL measurements. The CDC23 defines community VL as a measure that includes 4 of the 5 component measures excluding only those patients who are undiagnosed. To differentiate our usage from the CDC definition, we will italicize our use of community viral load throughout the text. Our study population reflects those patients diagnosed and accessing care at least once (ie, “In-Care Viral Load” population as defined by the CDC) during the study period in the community defined by geography rather than by demographics. Our usage, although not achieving the strict CDC definition of CVL, reflects a more pragmatic application of CVL and is more easily operationalized.
The Southern Alberta Clinic Cohort (SAC) follows all HIV+ individuals receiving any form of HIV care in southern Alberta, Canada, with a catchment population of approximately 1.3 million in 2010. The SAC provides exclusive free access to all HIV services (HIV VL tests, CD4 counts, and antiretroviral drugs) available under universal health care. The population followed at the SAC is defined by its isolated geography with the nearest alternate HIV care clinics located 180 miles (300 km) away. The SAC represents a complete geographically defined population in-care within which there are many heterogeneous communities. Patients for this analysis were categorized as “active” (or continuous) from the date when they first access HIV care and remain “active” until they died, transferred HIV care elsewhere outside the region, or were lost to follow-up (LTFU) as defined by the absence of clinic contact (ie, clinic visits, phone calls, HIV-specific laboratory tests, etc) for 12 consecutive months. Patients can only receive specific HIV care in southern Alberta through SAC.
Baseline demographic data (such as date of birth, age, sex, HIV risk factor, and self-reported ethnicity) from the patient's first visit to the SAC along with ongoing care details are maintained in the SAC database. The VL of all active patients (≥16 years of age) were collected per calendar year between 2001 and 2010 to determine the yearly total CVL for the entire HIV population in-care. We also analyzed in detail the VL testing completed in 2009 to determine the effect of churn on the total CVL.
Measuring Community Viral Load
The VL results are recorded as copies per milliliter. For patients with VL < 50 copies per milliliter, we used 50 copies per milliliter as the patient's VL for that test. We first calculated the mean VL per patient by summing up all the patient's VL measurements over a 12-month period and dividing by the number of VL tests. We then determined the total CVL by summing up all the mean VL counts for all the patients. The mean total CVL is calculated as the average of all VL counts within the particular time span and/or patient category. We use both the total and mean CVL as comparative measures rather than as absolute indicators of VL rates in the community as 60%–85% of patients had suppressed viremia (ie, <200 copies per milliliter). We agree with Das et al14 that the median VL is less useful, as in most cases the median would fall below the level of viral suppression for any given time period.
Cut-Off Levels for Virologic Suppression and the Level of Being Undetectable
We use the CDC23 recommended cut-offs to assess the level of HIV infection or the possible transmission potential: ≤50 copies per milliliter = undetectable; ≤200 copies per milliliter = suppressed; >200 copies per milliliter = not suppressed; and >100,000 copies per milliliter = high VL. We determined the proportion of patients above or below these cut-off levels.
New HIV Diagnosis
We report “new HIV diagnoses” in 2 ways. New locally diagnosed HIV patients were defined as having their first HIV-positive test within the region. These new local diagnoses can result from recent infections or are older than previously undiagnosed HIV infections. Although imprecise, we arbitrarily use the CD4 cut-off of ±350/mm3 to indicate a recent or older HIV infection. Individuals diagnosed outside the region may also be considered new infections to the community by bringing HIV into the region. We report both types of new diagnoses. The intensity of HIV testing in Alberta has remained unchanged during the decade.
Annual incidence rates for both new locally diagnosed HIV patients and for those diagnosed elsewhere moving into the area are calculated only for those individuals who accessed HIV care at some point during the listed year and do not include individuals with positive HIV tests that do not access care. Incidence rates were determined by dividing the number of new HIV diagnoses by the total population of southern Alberta multiplied by 100,000. Population statistics were obtained from Government of Alberta, Municipal Affairs.24
To assess the impact of churn on CVL, we grouped all active patients in 2009 into 6 categories: (1) ongoing patients in continuous care throughout the year, (2) new patients diagnosed within the region, (3) patients transferring their HIV care into the region (ie, transfer patients), (4) returning (ie, former SAC) patients reentering care during the year, (5) SAC patients active at the start of the year who subsequently transferred HIV care outside the region (ie, patients who moved), and (6) SAC patients active at the start of the year who became LTFU during the year. Patients who died during the year were excluded.
We compared the CVL between categories using simple descriptive statistics including population totals, mean, and SD.
The use of this administrative data is approved by the University of Calgary Conjoint Medical Committee on Medical Bioethics.
Yearly Total Community Viral Loads
Between January 1, 2001, and January 1, 2011, the number of active patients followed increased from 757 to 1423 (Table 2). The mean number of VL measurements per active patient per year remained stable from 3.1 (±0.53) in 2001 to 3.2 (±0.65) in 2010. Ninety-seven percent of all active patients had at least 1 VL test within the calendar year. With the increased use of HAART (62% in 2001 to 81% in 2010), the proportion of all patients with a mean VL of <200 copies per milliliter (ie, suppressed) increased from 49% in 2001% to 72% in 2010 (Table 3). The proportion of patients with mean high VLs (ie, >100,000 copies per milliliter) was variable from a high of 9% in 2002 to 4% in 2010. The mean VL per patient and the total CVL (ie, in-care VL) have remained stable from 2001 to 2010.
The number of new locally diagnosed HIV persons in-care increased from 46 in 2001 to 76 in 2010 with the corresponding incidence rate increasing from 4.4 to 5.8/100,000 (Table 2). The number of transfer patients previously diagnosed with HIV elsewhere but new to the population increased from 38 to 88 (incidence rate increasing from 3.6 to 6.7/100,000). Overall, the yearly number of new patients followed at the SAC (ie, either newly diagnosed or moved into the region) increased from 84 to 164 (incidence rate of HIV patients under care increasing from 8.0 to 12.5/100,000).
Churn and Community Viral Load in 2009
Table 4 shows the effects of churn on CVL for the year 2009. Patients enter the cohort either by having a new local HIV diagnoses (category 2), being diagnosed and treated outside the region and moving into the cohort (category 3), or former SAC patients who returned to HIV care after an absence (category 4). Patients leave the cohort by moving outside the region (category 5), disengaging in-care, or by being LTFU (category 6).
Continuous patients followed during the entire year (category 1) accounted for 78.7% of the population and 81% of all VL tests performed. They, however, only contributed 29.5% of the total CVL. The mean VL per patient was 10,575 copies per milliliter (SD = 0–277 copies per milliliter). Six-eight percent achieved mean undetectable VL throughout the year, and 81% were suppressed.
Individuals newly entering the population (ie, categories 2 and 3) account for 11.4% of all active patients but contribute 43% to the total CVL. New locally diagnosed patients (category 2), although 6.6% of the cohort population, account for 37.5% of the total CVL. The mean VL for these patients was 162,260 copies per milliliter (SD = 213,976); 95% of newly diagnosed patients had mean detectable VLs over the 12-month period in 2009. Patients presenting with a CD4 < 350 mm3 had higher mean VLs (329,856 copies per milliliter; SD = 316,779) than patients presenting with higher CD4 counts (81,886 copies per milliliter; SD = 77,456; data not shown).
Transfer patients (category 3) comprised 4.8% of all patients (43% of new patients) and contributed 5.5% to the total CVL. Thirty-nine percent of these patients had unsuppressed VLs for the year. The mean VL per patient was 34,509 copies per milliliter (SD = 27,458).
Former SAC patients returning to care (category 4), although only 4.2% of the population, contributed 16.6% of the CVL. Eighty-two percent of these patients were not suppressed. Their mean VL was 107,261 copies per milliliter (SD = 96,834). Seventy-five percent of the patients in this category had previously been listed at LTFU from SAC; >80% of the LTFU patients reported that they had remained in the community while they were discontinued from care. For patients who received some care elsewhere outside the region and then returned, 35% had mean VLs >200 copies per milliliter (mean VL = 38,837 copies per milliliter; SD = 25,345).
Patients can also leave the study population by moving or disengaging from care (LTFU). VLs for these patients were obtained before leaving. These patients comprise 5.7% of the total population but 10.9% of the total CVL. Patients who state they are moving and where they will be transferring to (category 5) had mean VLs of 25,337 copies per milliliter (SD = 34,259); however, 51% had unsuppressed mean VLs before leaving. Patients who became LTFU during 2009 (category 6) contributed only 3.1% toward population but 8.5% to the total CVL. When we examined their last VLs from 2008, we found that, before leaving care, they had a mean VL of 90,998 copies per milliliter (SD = 101,723) with 75% having detectable VLs highly suggestive of a major underestimate of LTFUs, true contribution to CVL.
Despite HAART use increasing in our population to 81%, and with the proportion of all patients with a suppressed VL increasing from 49% to 79%, total CVL (ie, in-care VL) did not significantly decrease over 10 years. No significant decline in new “local” HIV diagnoses was observed throughout the decade. Our results differ somewhat from those of other studies,12–19 but we suspect that the disparity may be due to the definition of community, different methods for determining CVL, variations on how new HIV diagnoses are reported, and the effects of churn.
Defining community remains problematic and contributes to the potential effect of selection and sampling basis as discussed by Miller et al.20 Most communities are neither isolated nor homogeneous entities. Communities have often been defined by lifestyle (eg, intravenous drug users17), or geographically by city,14,18 province, state, or country.15,16,19 Limiting a community as only those patients under continuous care may reduce some of the bias; however, it may mislead in measuring the preventative effects of HAART by reflecting neither the wider population living with HIV nor the population dynamics. For example, in our study, patients under continuous care (category 1) could be viewed as community, albeit not a homogeneous one. This community has a high proportion of patients on HAART (87%) with only 14% experiencing detectable viremia in 2009. Since 2005, the number of patients in this community increased from 768 to 1110, mean VL decreased from 16,219 to 12,737 copies per milliliter, and total CVL has been relatively stable (11,430,176 to 10,595,519 copies per milliliter; data not shown). However, if this community is used as the measure of successful treatment strategies, it does not account for why new HIV diagnoses have not decreased significantly.
The reporting of CVL also remains an important issue.20 We used the mean VL per patient as suggested by Das et al14 but wonder if this is the appropriate measurement. For example, if all VL measurements for all patients are used, it may better reflect the actual VL in the community but it creates statistical bias as some patients with repeated measurements contribute more to the total CVL than others. Using only the mean VL per patient, as we do here, may reduce bias but may not be representative of the actual VL within the population. Using only the highest VL, a patient reported during a time period (eg, Montaner et al15) may overrepresent CVL given fluctuating annual VLs in many patients. For example, we found that 1 in 5 patients with undetectable mean VLs had at least 1 detectable VL during the year. Miller et al20 point out that the type of CVL measurement selected can underestimate or overestimate the CVL, and which approach (ie, mean VL, most recent VL, and the highest VL) is the better reflection of total CVL remains uncertain.
A further issue is the definition of what exactly constitutes new HIV diagnoses. Ideally, only new locally diagnosed HIV infections should be included as they would reflect the local epidemic; however, new HIV infections and new HIV diagnoses are not necessarily the same. Changing numbers of diagnoses can be the result of greater or lesser HIV testing in the region; they may represent previously undiagnosed chronic HIV infections; or they may represent movement into the community by individuals who acquired HIV elsewhere but were not diagnosed. Conversely, individuals locally infected with HIV may leave the region without being diagnosed and hence are not captured as a new HIV infection. Defining exactly what constitutes a new HIV infection within a community is critical for assessing CVL as numbers can be quite different depending on how, where, and who reports new HIV diagnoses. We looked at all individuals with a new positive HIV test and reported to Public Health in 2009 to see who accessed HIV care. Nineteen of 102 individuals (18.2%) with positive tests did not access care before 2012. Using 102 rather than 83 new locally diagnosed HIV individuals would increase the incidence rate from 6.4 to 7.86/100,000. We do not however know if these were legitimate new diagnoses or previously diagnosed individuals living locally or even in transit taking an HIV test for unknown reasons.
Hence, we found that the number of new HIV infections in our population can vary significantly depending on which definition is used, and this, in turn, impacts CVL calculations. New locally diagnosed HIV patients, although comprising <7% of the total community, contributed 33% of the total CVL. It is well known that acute and early HIV infection is associated with greater transmissibility2,25 contributing a much larger percentage toward the total known CVL. Additionally, patients first diagnosed elsewhere who move into the region, and hence may be considered new HIV diagnoses for the purposes of CVL, contributed an additional 5% of the total CVL. Expanding or contracting the definition of new HIV diagnosis can either support or weaken (ie, underestimate or overestimate) CVL depending on the definition applied.
Churn seems to have a substantial impact on the calculation of CVL (Fig. 1). Migrating patients and patients who disengage from care tend to have a higher mean VL22 with fewer patients achieving suppressed VL levels. LTFU patients, for example, had high levels of detectable VL (82%) before leaving care. These individuals do not contribute significantly to the measured CVL in 2009 (3%) and their absence from care actually lowers the CVL, but if they remain in the community, they potentially can contribute significantly to new HIV infections. For instance, we found that 75% of the returning patients who were LTFU had detectable VLs upon returning to care and contributed more to the measured CVL than did patients moving into the region.
Our centralized care structure and geography allow us to examine many of the components of CVL in our population. We, however, like every other center, are limited in determining the “true” CVL in that we do not know who is HIV infected but undiagnosed. Studies26–30 have estimated that between 20% and 30% of all HIV-infected individuals remain undiagnosed. In addition, Kelley et al31 discuss the importance of determining the proportion of persons within the total population that are unsuppressed and how they contribute to CVL; however, how much these individuals are contributing to the CVL is not measurable (by definition as a VL test result is absent) but given that at least 82% of patients with an HIV-positive test accessed HIV care in our study, and that newly diagnosed patients contributed almost 33% of the total CVL that is known, inclusion of these individuals (ie, imputing of VL data) would most likely substantially increase the total CVL. As such, the data we present are an underestimation of the true CVL within the community. We also do not directly address the assumptions listed in Table 1, although testing these assumptions should be a priority before CVL is widely incorporated; the assumptions may or may not, by themselves, be supportable.
Over the past decade, despite increasing use of HAART, and fewer patients in-care with detectable VLs, there has been little or no change in either the total CVL (ie, in-care VL) or the number of new HIV diagnoses. These data may result from differential transmissibility of high-risk individuals as discussed by Miller et al,20 or may also in part reflect the aspects of the “churn” effect. Communities and social networks are not “closed” and are susceptible to movement between communities, however defined.
We strongly agree with more widespread and routine testing for HIV infection coupled with accession of care in a timely fashion. The provision of HAART to reduce VLs is important both for treatment of an individual and to reduce their infectivity to others.31,32 Retention in-care remains a critical issue.33,34 We feel however that, although CVL is an intriguing measurement, the use of CVL as a benchmark indicator of “success” is, at present, premature and must be used with caution. Our CVL results suggest that in developed world populations effort should be focused on finding undiagnosed patients and engaging them in lifelong care.
The authors wish to thank Karen Ko and Danette Mohagen at the Southern Alberta Clinic for their assistance in maintaining the SAC database. They thank Marcy Thompson and Dr. Judy MacDonald for providing surveillance data of HIV tests conducted in southern Alberta. They also thank the anonymous reviewers for their comments and suggestions.
1. Quinn TC, Wawer MJ, Sewankambo N, et al.. Viral load and heterosexual transmission of human immunodeficiency virus type 1. Rakai Project Study Group. N Engl J Med. 2000;342:921–929.
2. Cohen MS, Chen YQ, McCauley M, et al.. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med. 2011;365:493–505.
3. Donnell D, Baeten JM, Kiarie J, et al.. Heterosexual HIV-1 transmission after initiation of antiretroviral therapy: a prospective cohort analysis. Lancet. 2010;375:2092–2098.
4. Sullivan P, Kayitenkore K, Chomba E, et al.. Reduction of HIV transmission risk and high risk sex while prescribed ART: results from discordant couples in Rwanda and Zambia. Abstract 52bLB. 16th Conference on Retrovirus and Opportunistic Infections, Montreal, Canada, February 8–11, 2009.
5. Wang L, Ge Z, Luo J, et al.. HIV transmission risk among serodiscordant couples: a retrospective study of former plasma donors in Henan, China. J Acquir Immune Defic Syndr. 2010;55:232–238.
6. Velasco-Hernandez JX, Gerhengorn HB, Blower SM. Could widespread use of combination antiretroviral therapy eradicate HIV epidemics? Lancet Infect Dis. 2002;2:487–493.
7. Granich RM, Gilks CF, Dye C, et al.. Universal voluntary HIV testing with immediate antiretroviral therapy as a strategy for elimination of HIV transmission: a mathematical model. Lancet. 2009;373:48–57.
8. Garnett GP, Baggaley RF. Treating our way of the HIV pandemic: could we, would we, should we? Lancet. 2009;373:9–11.
9. Wagner BG, Kahn JS, Blower S. Should we try to eliminate HIV epidemics by using a “Test and Treat” strategy? AIDS. 2010;24:775–776.
10. El-Sadr WM, Affrunti M, Gamble T, et al.. Antiretroviral therapy: a promising HIV prevention strategy? J Acquir Immune Defic Syndr. 2010;55:S116–S121.
11. Mayer KH, Venkatesh KK. Antiretroviral therapy as HIV prevention: status and prospects. Am J Public Health. 2010;100:1867–1876.
12. Montaner JS, Hogg R, Wood E, et al.. The case for expanding access to highly active antiretroviral therapy to curb the growth of the HIV epidemic. Lancet. 2006;368:531–536.
13. Lima VD, Johnston K, Hogg RS, et al.. Expanded access to highly active antiretroviral therapy: a potentially powerful strategy to curb the growth of the HIV epidemic. J Infect Dis. 2008;198:59–67.
14. Das M, Chu PL, Santos GM, et al.. Decreases in community viral load are accompanied by reductions in new HIV infections in San Francisco. PLoS One. 2010;5:e11068.
15. Montaner JSG, Lima VD, Barrios R, et al.. Association of highly active antiretroviral therapy coverage, population viral load, and yearly new HIV diagnosis in British Columbia, Canada: a population-based study. Lancet. 2010;376:532–539.
16. Law MG, Woolley I, Templeton DJ, et al.. Trends in detectable viral load by calendar year in the Australian HIV observational database. J Int AIDS Soc. 2011;14:10.
17. Wood E, Kerr T, Marshall B, et al.. Longitudinal community plasma HIV-1 RNA concentrations and incidence of HIV-1 among injecting drug users: prospective cohort study. BMJ. 2009;338:b1649.
18. Castel AD, Befus M, Willis S, et al.. Use of the community viral load as a population-based biomarker of HIV burden. AIDS. 2012;26:345–353.
19. Fang CT, Hsu HM, Twu SJ, et al.. Decreased HIV transmission after a policy of providing free access to highly active antiretroviral therapy in Taiwan. J Infect Dis. 2004;190:879–885.
20. Miller WC, Powers KA, Smith MK, et al.. Community viral load as a measure for assessment of HIV treatment as prevention. Lancet Infect Dis. 2013;13:459–464.
21. Rebeiro P, Althoff K, Gill MJ, et al.. Retention among North American HIV-infected persons in clinical care, 2000–2008. JAIDS. 2013;62:356–362.
22. Gill MJ, Krentz HB. Unappreciated epidemiology: the churn effect in a regional HIV care programme. Int J STD AIDS. 2009;20:540–544.
25. Hollingsworth TD, Anderson RM, Fraser C. HIV-1 transmission by stage of infection. J Infect Dis. 2008;198:687–693.
26. Campsmith ML, Rhodes PH, Hall H, et al.. Undiagnosed HIV prevalence among adults and adolescents in the United States at the end of 2006. J Acquir Immune Defic Syndr. 2010;53:619–624.
27. Hall HI, Song R, Rhodes P, et al.. Estimation of HIV incidence in the United States. JAMA. 2008;300:520–529.
30. Yang Q, Boulos D, Yan P, et al.. Estimates of the number of prevalent and incident human immunodeficiency virus (HIV) infection in Canada, 2008. Can J Public Health. 2010:101:486–490.
31. Kelley CF, Rosenberg ES, O'Hara BM, et al.. Measuring population transmission risk for HIV: an alternative metric of exposure risk in men who have sex with men (MSM) in the US. PLoS One. 2012;7:e53284. doi:10.1371/journal.pone.0053284.
32. Moore RD, Barlett JG. Dramatic decline in the HIV-1 RNA levels over calendar time in a large urban HIV practice. Clin Infect Dis. 2011;53:600–604.
33. Cambiano V, Rodger AJ, Phillips AN. “Test-and-treat”: the end of the HIV epidemic? Curr Opin Infect Dis. 2011;24:19–26.
34. Metsch LR, Pereyra M, Messinger S, et al.. HIV transmission risk behaviors among HIV-infected persons who are successfully linked to care. Clin Infect Dis. 2008;47:577–584.