Secondary Logo

Journal Logo

EPIDEMIOLOGY AND SOCIAL

Variation in estimated viral suppression associated with the definition of viral suppression used

Lesko, Catherine R.a; Chander, Geetanjalib; Moore, Richard D.b; Lau, Bryana

Author Information
doi: 10.1097/QAD.0000000000002579

Abstract

People living with HIV (PLWH) who have a suppressed viral load are less likely to develop complications related to their HIV infection and less likely to transmit the HIV virus [1–5]. Monitoring population levels of viral suppression has become a key indicator of the HIV care delivery system and population health as the final stage of the HIV care continuum. Yet definitions of ‘viral suppression’ can vary. Viral load varies for PLWH, both within and across calendar years. The HIV care continuum attempts to summarize the value of a dynamic, continuous variable within a calendar year into a single, binary variable and there is a necessary loss of information. There have been few studies of the implications of the definition of ‘viral suppression’ used when reporting HIV care continuum(s) [6].

There are three main components of calculating ‘viral suppression.’ First, (and perhaps under-recognized) is defining who will contribute to the calculation. Namely, who is expected to have a viral load measurement during a particular calendar year, what are the implications of excluding persons with missing viral load data, and how long should persons be under observation during the year to contribute to the calculation?

Second, is the threshold below which viral load will be considered suppressed. Common choices are 20, 50, 200, or 400 copies/ml [7]. Lower thresholds are more clinically meaningful for patient care because even low-level viremia (detectable viral load values below thresholds typically considered ‘suppressed’, e.g., 3–1000 copies/ml) is associated with subsequent virological failure (≥1 or ≥2 viral load values above a threshold or a change in treatment regimen) and morbidity and mortality [8–12]. Yet higher thresholds are commonly used to accommodate data from commercial viral load tests with higher limits of detection.

Third, is how to analyze multiple viral load tests per person. Most summaries of the ‘proportion suppressed’ in estimates of the HIV care continuum use the last viral load in a period for each person [13,14]. Others have chosen a random viral load value for patients with more than one measurement [15]. A more stringent definition of suppression would require all viral load values in the period to be below the threshold [16], whereas a less stringent definition would be to classify patients as suppressed if any viral load values in the period are below the threshold. Clearly, to prevent morbidity and ongoing HIV transmission, the goal is that all viral load values in the period are below the threshold [16]. Yet such a requirement discounts progress by classifying individuals as unsuppressed in a year in which they start unsuppressed but quickly become and remain suppressed for the remainder of the year.

Herein, we examined the variability in the proportion of a cohort classified as virally suppressed in a given year according to varying definitions and thresholds for viral suppression. We examined whether estimates of viral suppression based on varying definitions have converged over time, and we described some potential consequences of choosing one definition over another.

Methods

Study sample

We used data from the Johns Hopkins HIV Clinical Cohort (JHHCC). The JHHCC enrolls all patients receiving continuity HIV care at the John G. Bartlett Specialty Practice who consent to share their data (>90% of all those invited). Data are extracted and abstracted from electronic medical records quarterly and include demographics, HIV acquisition risk factors, and laboratory values. For this analysis, we included all persons in the JHHCC who enrolled during the study period (2010–2018) or who enrolled prior to the study period but who had at least one viral load test during the study period.

Our goal was to estimate the proportion of patients ‘in care’ in a given year who were ‘virally suppressed’. The definition of ‘in care’ determines the denominator for calculating the proportion suppressed. We considered two primary definitions of ‘in care’ for a given year: (1) everyone who enrolled in the cohort or had at least one viral load measured in the clinic in the prior year and who was alive at the start of the year; and (2) everyone who had at least one viral load measured in the clinic in the same year. For (2), everyone in the denominator had a value for viral suppression (the numerator for calculating the proportion suppressed) by design. However, persons could be included in the denominator under (2) who did not have a full year of follow-up if their first viral load measurement in the clinic occurred partway through the year. This might have made them more (fewer chances to have an unsuppressed viral load measurement) or less likely (if they did not have a chance to suppress after antiretroviral therapy (ART) initiation) to be virally suppressed. For (1), some patients included in the denominator did not have a value for the numerator; patients may have had a viral load measured in the prior year but not in the year of interest. For these patients we either: first, censored them – that is, excluded them from both the numerator and denominator – which implies that we believe their probability of viral suppression is the same as persons who remain in the clinic; or second, imputed their viral load to be not suppressed, which equates loss to clinic with loss to HIV care and loss of access to ART, and thus loss of viral suppression.

In sensitivity analyses, we restricted the denominator for a given year to persons who (3) enrolled in the cohort or had at least one viral load measured in the clinic in the prior year and were alive at the end of the year; and (4) had at least one viral load measured in the clinic in the same year and were alive at the end of the year. Restricting to persons who were alive at the end of the year may underestimate the transmission potential and overestimate the ART adherence of the same group of PLWH for the year in question, because people with an unsuppressed viral load are more likely to die. Figure 1 is a schematic showing five different hypothetical patients and which of each of these four denominators they would contribute to.

Fig. 1
Fig. 1:
Schematic showing five hypothetical patients and their membership in each denominator considered for calculation of proportion of patients in care in a given year (e.g. 2018) who were virally suppressed.

Historically, HIV care continuum estimates of viral suppression for the United States, based on national surveillance systems, were calculated by looking at the past 12 months of viral load values for patients who were in care in the first 3 months of a given year [most similar to denominator definition (2)] [13,14]. Currently (as of 2019), the denominator for US HIV care continuum estimates of viral suppression is everyone diagnosed before the start of the year and alive at the end of the year (definition 3, as described above). Persons in the denominator with no viral load value for the year are assumed to be unsuppressed [17,18].

Outcome

We explored several different definitions of viral suppression. We considered thresholds of 20 or less, 50 or less, 200 or less, and 400 copies/ml or less. Twenty copies/ml corresponds to the limit of detection of the most common viral load test at the end of the study period; 50 copies/ml corresponds to the limit of detection of the most common viral load test at the start of the study period; 400 copies/ml corresponds to the highest limit of detection of any viral load tests used during the entire study period; and 200 copies/ml is the threshold used to define viral suppression by the Department of Health and Human Services [18]. We classified viral load measurements using a test with a limit of detection that was above the threshold for suppression (e.g. a viral load value of ≤400 when using a threshold of ≤20 copies/ml) as suppressed. This might have overestimated the proportion of the population suppressed using a lower threshold particularly in years in which tests with higher limits of detection were common.

For persons with at least one viral load measurement(s) (labs) in a given year, we used four different approaches to classify them as ‘virally suppressed’ for that year: first, any measurements were below the threshold (or below the limit of detection of the viral load test used); second, the last measurement in the year was below the threshold (or below the limit of detection); third, at least 50% of measurements in the year were below the threshold (or below the limit of detection); or fourth, all viral load measurements were below the threshold (or below the limit of detection).

Statistical analysis

We report the proportion of persons in the cohort classified as suppressed based on the various definitions of the denominator; various thresholds for suppression; and various strategies for summarizing multiple viral loads per year as described above.

To determine whether there are potentially meaningful clinical implications for classifying people as virally suppressed or not, we also calculated the 5-year risk of all-cause mortality associated with each classification of viral suppression. We included persons classified as ‘in-care’ in a given year according to denominator definition (3): persons who had at least one viral load in the prior year who were still alive at the end of the analysis year. We followed these persons from January 1 of the following year to the earliest of death or administrative censoring on 31 December 2018. Because deaths were actively ascertained (through follow-up by clinic staff through the regional health information exchange and emergency contacts, and through regular matches against the National Death Index), there was no lost to follow-up (LTFU) for these analyses. We fit Cox proportional hazards models for time to death with indicator variables for loss to clinic in the year prior to baseline, having any viral load measurement below the threshold, having the last viral load measurement below the threshold, and having all labs below the threshold (such that the hazard ratio associated with stricter definitions of viral suppression is the additional association with mortality beyond being suppressed according to a weaker definition). We presented hazard ratios adjusted only for time in care and calendar year at baseline. In sensitivity analyses, we also estimated hazard ratios adjusted for age, sex, race, HIV acquisition risk group (injection drug use, men who have sex with men, and heterosexual sex; risk groups were not mutually exclusive), time in care, and calendar year at baseline. Adjustment was accomplished using inverse probability of exposure weighting [19,20].

Results

The majority of the 3911 patients who contributed to this analysis were men (65%) or black (75%). At the start of follow-up during the study period, the median age was 48 years [interquartile range (IQR): 40, 54] and median time previously enrolled in the clinic was 0 years (IQR: 0, 11) (0 years corresponds to patients newly enrolling in the cohort during the study period) (Table 1). Patients contributed a median of 6 years (IQR: 3, 9) to this analysis.

Table 1
Table 1:
Characteristics of persons in the Johns Hopkins HIV Clinical Cohort who had at least one viral load test, 2010–2018, and person-years they were included in the denominator for estimating the proportion of the cohort that was virally suppressed.

From 2010 to 2018, there were on average 5923 viral load measurements per year. Patients contributed a median of 2 (IQR: 2, 3) measurements per year in years in which they contributed any measurements. There were no strong temporal trends in the number of viral load measurements per year; median measurements in 2010 was 3 (IQR: 2, 4) and median measurements in 2018 was 2 (IQR: 2, 3). The proportion of all viral load test results more than 400 copies/ml decreased from 21% in 2010 to 12% in 2018 (Table 2).

Table 2
Table 2:
Distribution of all viral load values by year measured on all persons in the Johns Hopkins HIV Clinical Cohort who had at least one viral load test, 2010–2018.

The proportion of patients classified as virally suppressed depended on the definition of viral suppression used (Table 3). Highest estimate was 92.5%, in which suppression was defined as having at least one viral load of 400 copies/ml or less during the year and we censored patients who were lost to clinic. Lowest was 51.8% suppressed, in which suppression was defined as having all viral load measurements during the year of 20 copies/ml or less (or below the limit of detection of the test) and persons who were lost to clinic were assumed to be unsuppressed.

Table 3
Table 3:
Percentage of cohort classified as virally suppressed in a given year, 2010–2018, according to different definitions of viral suppression, assumptions about the viral load of patients lost to follow-up, and thresholds for classifying viral loads as suppressed.

Setting a lower threshold resulted in a lower proportion of patients classified as suppressed in a given year regardless of how multiple viral loads per person were analyzed. For example, defining viral suppression as having all viral load measurements below the threshold, the difference between the proportion suppressed at 400 copies/ml or less versus 200 copies/ml or less was about 2%; the difference between the proportion suppressed at 400 copies/ml or less versus 50 copies/ml or less was about 9%; and the difference between the proportion suppressed at 400 copies/ml or less versus 20 copies/ml or less was about 18% (Table 3). These differences in proportion suppressed associated with the threshold used to define viral suppression were relatively stable over time (Web Table 1, http://links.lww.com/QAD/B754).

Predictably, requiring persons to have all viral load measurements below the threshold resulted in the lowest proportion of persons suppressed, and accepting any viral load measurement below the threshold resulted in the highest proportion of the cohort suppressed. Compared with the proportion of patients with all viral load measurements of 200 copies/ml or less, the proportion with the last lab of 200 copies/ml or less was about 10% higher, and the proportion with any measurements of 200 copies/ml or less was about 13% higher (Table 3). The difference between the proportion suppressed by any labs versus all labs, or by the last lab versus all labs, was greater for lower thresholds (approximately 13% for a threshold of ≤400 copies/ml compared with approximately 23% for a threshold of ≤20 copies/ml).

The absolute differences between proportion suppressed by various definitions also decreased slightly over time, mostly by virtue of increasing viral suppression regardless of the definition used (Fig. 2). (The slight dip in the proportion suppressed in 2013 based on definitions that treated persons without a viral load value in the analysis year as unsuppressed is thought to be associated with some loss of data due to adoption of a new electronic health record during that year.) In 2010, the difference in proportion suppressed at 200 copies/ml or less when defined as any measurements below the threshold [denominator (1), censoring persons who were LTFU] was 87.5% versus only 69.8% suppressed when defined as all labs below the threshold (a difference of 18%). In 2018, the difference in proportion suppressed at 200 copies/ml or less was 95.0% when defined as any measurements below the threshold versus 83.7% when defined as all measurements below the threshold (a difference of only 11%; Web Table 1, http://links.lww.com/QAD/B754).

Fig. 2
Fig. 2:
Proportion of cohort with suppressed viral load (≤200 copies/ml), by year, according to the definition of lost to follow-up employed, Johns Hopkins HIV Clinical Cohort, 2010–2018.

Finally, when we took as the denominator all patients who had at least one viral load in the prior year and censored those who did not have a viral load in the analysis year, we estimated that the proportion suppressed was about 10% higher than when we assumed that all persons who did not have a viral load measurement were not suppressed. This difference was relatively stable over time. When the denominator was only patients who had a viral load in the analysis year, the estimated proportion of the cohort that was suppressed was between the two estimates based on denominator definition (1), and usually closer to the estimate in which patients who were LTFU were censored (Table 3).

In a sensitivity analysis in which restricted the denominators to persons who survived the calendar year of interest, results were very similar (Web Table 2, http://links.lww.com/QAD/B754). Absolute differences in the proportion estimated to be virally suppressed were generally less than 1%.

Among persons in-care according to denominator definition (3), adjusting for time in care and calendar year, and using a threshold for suppression of 200 copies/ml or less, the 5-year risk of mortality was 12.7% [95% confidence interval (CI): 10.4%, 14.8%] for patients who had at least one measured viral load in the year prior to baseline but none were suppressed; 7.8% (95% CI: 6.4%, 9.2%) among patients who were lost to clinic in the year prior to baseline; 7.5% (95% CI: 6.6%, 8.4%) among patients with any lab of 200 copies/ml or less in the year prior to baseline; 7.2% (95% CI: 6.3%, 8.1%) among patients whose last lab in the year prior to baseline was of 200 copies/ml or less; and 6.7% (95% CI: 5.8%, 7.7%) among patients with all labs of 200 copies/ml or less in the year prior to baseline (Web Table 3, http://links.lww.com/QAD/B754). Across all thresholds, having all labs below the threshold was associated with a lower hazard of mortality than just the last lab suppressed, and having the last lab below the threshold was associated with a lower hazard of mortality than having any other lab suppressed (Table 4). This result held even after further adjusting for birth sex, age, race, ethnicity, and HIV acquisition risk group (Web Table 4, http://links.lww.com/QAD/B754).

Table 4
Table 4:
Hazard ratiosa for 5-year all-cause mortality associated with viral suppression status at baseline.b

Discussion

We demonstrated differences in the proportion of our cohort classified as virally suppressed in a given year depending on the definition of viral suppression used. Not surprisingly, a more stringent definition of viral suppression (lower threshold, requiring all labs below that threshold) was associated with a lower proportion of the cohort meeting that definition. More surprisingly, was the magnitude of the difference in the proportion of the cohort classified as suppressed under the least stringent versus the most stringent definition of suppression, which was close to an absolute difference of 40% (92.5% at most versus 51.8% at least). Using the current (as of 2019) definition of viral suppression for US HIV care continuum estimates (denominator definition 3, threshold of ≤200 copies/ml, and persons in the denominator with no viral load value for the year are assumed to be unsuppressed) [17,18], 76.7% of the cohort was virally suppressed in an average year during follow-up. Regardless of the denominator definition used to define proportion suppressed, we confirmed prior work showing that a more stringent threshold for characterizing persons as virally suppressed (lower threshold, continuous suppression) was more meaningful for individuals (lower mortality risk above and beyond any viral suppression); although beyond the scope of this analysis, durable viral suppression is likely (logically) more meaningful for the population with respect to transmission.

There are pros and cons of each of these definitions of viral suppression for monitoring the HIV care continuum. The HIV care continuum is easily communicated and is likely to persist as a primary indicator for public health. However, estimating it necessitates ignoring variability in an individual's viral load over a period and oversimplifying a complex landscape. In particular, a single (e.g. last) snapshot of viral load will overestimate durable viral suppression [21] but underestimate opportunity, that is, persons who have been able to achieve viral suppression, but perhaps not maintain it [22]. For example, in HIV surveillance data in US jurisdictions that required reporting of HIV viral load test results, among persons who had at least one viral load measurements in 2011 and at least two viral load measurements in 2012–2013, their most recent viral load measurement was of 200 copies/ml or less for 83%, but only 62% had all viral load values of 200 copies/ml or less (durable suppression). On average, persons who did not achieve durable suppression spent 60% of their time over the 2 years of follow-up with unsuppressed viral load [16].

Another key finding of this analysis and key consideration in estimating viral suppression is the importance of the definition of the denominator. When we defined the denominator as persons with at least one viral load measurement in the prior calendar year who were alive at the start of the current year, we explicitly link retention in care (a weaker version of retention than typical care continuum definitions) with viral suppression. Persons who do not return to our clinic may either be in care elsewhere or may be out of care. When we censor them, we assume they are in care elsewhere with the same probability of being virally suppressed as persons who remain in care in our clinic. In a study based in a clinic in Georgia that was able to link with statewide surveillance data, 11% of persons lost to the clinic were virally suppressed elsewhere based on a single viral load measurement, and only 3% were virally suppressed elsewhere based on a more continuous measure of viral suppression [23]. Thus the true proportion suppressed using denominator (1) was probably closer to what was estimated assuming that patients who were LTFU were unsuppressed versus censoring them. When we restricted estimation of the proportion suppressed to patients who had at least one viral load measurement in the same calendar year (denominator 2), the impression may be that we are estimating the proportion virally suppressed among those who were retained in care, but patients drop out of care even within the calendar year. Someone who has only one viral load measurement in early January may never return to the clinic but is counted in the denominator, and classified as suppressed or not based on that one viral load value. Using denominator (1) forces us to be explicit about the role that retention plays in viral suppression.

Viral suppression fluctuates within persons both across and within calendar years. Furthermore, the population in whom viral suppression could be monitored fluctuates across time within a population (due to immigration, emigration, new HIV diagnoses, and deaths). Limits of detection of commercial laboratory tests changes over time. And the frequency of viral load monitoring fluctuates across persons and across calendar years. The HIV care continuum attempts to summarize the dynamic state of viral suppression within persons across a calendar year and there is a necessary loss of information. The HIV care continuum is also designed to describe progress in population control and management of HIV infection across calendar time, and thus must accommodate various changes in clinical practices (changing limits of detection, changing guidelines for viral load monitoring). Those accommodations may result in additional losses of information. The definition of viral suppression used in care continuum estimates can provide either an overly optimistic or overly pessimistic snapshot of viral suppression for a cohort. We have demonstrated the plausible range of those estimates. In our data, durable viral suppression was associated with improved survival in individuals; logically, durable viral suppression is more relevant for treatment as prevention. Moving forward, as the prevalence of viral suppression ideally continues to improve, it will be important to report estimates of durable viral suppression, in addition to the standard estimates of suppression.

Acknowledgements

The current study was supported by grants from the National Institutes of Health (U01 DA036935 and K24 AA027483, and K01 AA028193).

Conflicts of interest

There are no conflicts of interest.

References

1. Cohen MS, Chen YQ, McCauley M, Gamble T, Hosseinipour MC, Kumarasamy N, et al. Prevention of HIV-1 infection with early antiretroviral therapy. N Engl J Med 2011; 365:493–505.
2. Palella FJ Jr, Delaney KM, Moorman AC, Loveless MO, Fuhrer J, Satten GA, et al. Declining morbidity and mortality among patients with advanced human immunodeficiency virus infection. HIV Outpatient Study Investigators. N Engl J Med 1998; 338:853–860.
3. Quinn TC, Wawer MJ, Sewankambo N, Serwadda D, Li C, Wabwire-Mangen F, 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.
4. Wood E, Milloy MJ, Montaner JS. HIV treatment as prevention among injection drug users. Curr Opin HIV AIDS 2012; 7:151–156.
5. Lundgren JD, Babiker AG, Gordin F, Emery S, Grund B, Sharma S, et al. INSIGHT START Study Group. Initiation of antiretroviral therapy in early asymptomatic HIV infection. N Engl J Med 2015; 373:795–807.
6. Miller WC, Powers KA, Smith MK, Cohen MS. Community viral load as a measure for assessment of HIV treatment as prevention. Lancet Infect Dis 2013; 13:459–464.
7. McMahon JH, Elliott JH, Bertagnolio S, Kubiak R, Jordan MR. Viral suppression after 12 months of antiretroviral therapy in low-and middle-income countries: a systematic review. Bull World Health Organ 2013; 91:377E–385E.
8. Hermans LE, Moorhouse M, Carmona S, Grobbee DE, Hofstra LM, Richman DD, et al. Effect of HIV-1 low-level viraemia during antiretroviral therapy on treatment outcomes in WHO-guided South African treatment programmes: a multicentre cohort study. Lancet Infect Dis 2018; 18:188–197.
9. Elvstam O, Medstrand P, Yilmaz A, Isberg P-E, Gisslén M, Björkman P. Virological failure and all-cause mortality in HIV-positive adults with low-level viremia during antiretroviral treatment. PloS One 2017; 12:e0180761.
10. Vandenhende MA, Ingle S, May M, Chene G, Zangerle R, Van Sighem A, et al. Antiretroviral Therapy Cohort Collaboration (ART-CC). Impact of low-level viremia on clinical and virological outcomes in treated HIV-1-infected patients. AIDS 2015; 29:373–383.
11. Laprise C, de Pokomandy A, Baril J-G, Dufresne S, Trottier H. Virologic failure following persistent low-level viremia in a cohort of HIV-positive patients: results from 12 years of observation. Clin Infect Dis 2013; 57:1489–1496.
12. Maggiolo F, Callegaro A, Cologni G, Bernardini C, Velenti D, Gregis G, et al. Ultrasensitive assessment of residual low-level HIV viremia in HAART-treated patients and risk of virological failure. J Acquir Immune Defic Syndr 2012; 60:473–482.
13. Whiteside YO, Cohen SM, Bradley H, Skarbinski J, Hall HI, Lansky A. Progress along the continuum of HIV care among blacks with diagnosed HIV – United States, 2010. MMWR Morb Mortal Wkly Rep 2014; 63:85–89.
14. Bradley H, Hall HI, Wolitski RJ, Van Handel MM, Stone AE, LaFlam M, et al. Vital signs: HIV diagnosis, care, and treatment among persons living with HIV – United States, 2011. MMWR Morb Mortal Wkly Rep 2014; 63:1113–1117.
15. Nance RM, Delaney JC, Simoni JM, Wilson IB, Mayer KH, Whitney BM, et al. HIV viral suppression trends over time among HIV-infected patients receiving care in the United States, 1997 to 2015: a cohort study. Ann Inter Med 2018; 169:376–384.
16. Crepaz N, Tang T, Marks G, Mugavero MJ, Espinoza L, Hall HI. Durable viral suppression and transmission risk potential among persons with diagnosed HIV infection: United States, 2012–2013. Clin Infect Dis 2016; 63:976–983.
17. Li Z, Purcell DW, Sansom SL, Hayes D, Hall HI. Vital signs: HIV transmission along the continuum of care – United States, 2016. MMWR Morb Mortal Wkly Rep 2019; 68:267–272.
18. Centers for Disease Control and Prevention. Monitoring selected national HIV prevention and care objectives by using HIV surveillance data - United States and 6 dependent areas, 2017. HIV Surveillance Supplemental Report 2019; 24. http://www.cdc.gov/hiv/library/reports/hiv-surveillance.html. Published June 2019. [Accessed 9 January 2020]
19. Robins JM, Hernán MA, Brumback B. Marginal structural models and causal inference in epidemiology. Epidemiology 2000; 11:550–560.
20. Sato T, Matsuyama Y. Marginal structural models as a tool for standardization. Epidemiology 2003; 14:680–686.
21. Marks G, Patel U, Stirratt MJ, Mugavero MJ, Mathews WC, Giordano TP, et al. Single viral load measurements overestimate stable viral suppression among HIV patients in care: clinical and public health implications. J Acquir Immune Defic Syndr 2016; 73:205–212.
22. Crepaz N, Tang T, Marks G, Hall HI. Viral suppression patterns among persons in the United States with diagnosed HIV infection in 2014. Ann Inter Med 2017; 167:446–447.
23. Colasanti J, Kelly J, Pennisi E, Hu Y-J, Root C, Hughes D, et al. Continuous retention and viral suppression provide further insights into the HIV care continuum compared to the cross-sectional HIV care cascade. Clin Infect Dis 2015; 62:648–654.
Keywords:

care continuum; HIV; measurement error; survival; viral suppression

Supplemental Digital Content

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.