Regional differences in rates of HIV-1 viral load monitoring in Canada: Insights and implications for antiretroviral care in high income countries

Background Viral load (VL) monitoring is an essential component of the care of HIV positive individuals. Rates of VL monitoring have been shown to vary by HIV risk factor and clinical characteristics. The objective of this study was to determine whether there are differences among regions in Canada in the rates of VL testing of HIV-positive individuals on combination antiretroviral therapy (cART), where the testing is available without financial barriers under the coverage of provincial health insurance programs. Methods The Canadian Observational Cohort (CANOC) is a collaboration of nine Canadian cohorts of HIV-positive individuals who initiated cART after January 1, 2000. The study included participants with at least one year of follow-up. Generalized Estimating Equation (GEE) regression models were used to determine the effect of geographic region on (1) the occurrence of an interval of 9 months or more between two consecutive recorded VL tests and (2) the number of days between VL tests, after adjusting for demographic and clinical covariates. Overall and regional annual rates of VL testing were also reported. Results 3,648 individuals were included in the analysis with a median follow-up of 42.9 months and a median of 15 VL tests. In multivariable GEE logistic regression models, gaps in VL testing >9 months were more likely in Quebec (Odds Ratio (OR) = 1.72, p < 0.0001) and Ontario (OR = 1.78, p < 0.0001) than in British Columbia and among injection drug users (OR = 1.68, p < 0.0001) and were less likely among older individuals (OR = 0.77 per 10 years, p < 0.0001), among men having sex with men (OR = 0.62, p < 0.0001), within the first year of cART (OR = 0.15, p < 0.0001), among individuals on cART at the time of the blood draw (OR = 0.34, p < 0.0001) and among individuals with VL < 50 copies/ml at the previous visit (OR = 0.56, p < .0001). Conclusions Significant variation in rates of VL testing and the probability of a significant gap in testing were related to geographic region, HIV risk factor, age, year of cART initiation, type of cART regimen, being in the first year of cART, AIDS-defining illness and whether or not the previous VL was below the limit of detection.


Background
Viral load (VL) testing is an essential component of the care of HIV-positive individuals, both with regard to timing of initiation of antiretroviral therapy (ART) and to monitoring of virologic response to combination ART (cART) [1]. The goal of cART is sustained virologic suppression, defined as a VL below the level of detection of the test performed [1]. Guidelines recommend that HIV-positive individuals receive VL testing at intervals of three to four months as standard of care [1]. CD4 count monitoring is important for deciding when to start cART and for determining prognosis, but alone is insufficient as a marker of treatment efficacy as it does not identify individuals experiencing virologic rebound or failure [2]. Early determination of virologic rebound and failure is one of the most crucial components of HIV management as it contributes to the reduction of ART drug resistance [3]. Lastly, VL monitoring has also been shown to promote treatment adherence, which is additionally important for maintaining virologic suppression and reducing the evolution of drug resistance [4].
Access to VL testing has been studied previously. In an Ontario cohort, injection drug use, younger age and residence in Toronto were associated with lower VL testing rates [5]. In another study, drug users were also found to be at risk for irregular VL monitoring [6]. In a study of individuals who initiated ART between 1994 and 2000, individuals with low CD4 counts and high VLs had the highest rates of laboratory testing [7].
In this study, we examine whether there are regional differences in patterns of VL testing among individuals who initiated cART therapy since January 1, 2000 in Canada, where VL testing is available without charge to all HIV-positive residents as part of the provincial universal health insurance plans. Furthermore, we identified demographic and clinical factors associated with suboptimal frequency of VL testing.

Methods
The Canadian Observational Cohort (CANOC) collaboration is a Canadian cohort study of antiretroviral naïve HIV-positive patients initiating cART since January 1 st 2000. The study was established in March 2008 with funding from the Canadian Institutes of Health Research (grant# 711098) and the CIHR Canadian HIV Trials Network (CTN242) and includes cohorts and investigators from across the country (listed at the end of the manuscript). The collaboration is open to all Canadian HIV treatment cohorts with more than 100 eligible patients.

Participating cohorts
Data used in this analysis were from nine cohorts of HIV-positive individuals in British Columbia (BC), Ontario, and Quebec, including the BC Centre for Excellence in HIV/AIDS Drug Treatment Program, Montreal Chest Institute Immunodeficiency Cohort, The Electronic Antiretroviral Therapy, Clinique Médicale l'Actuel, The Canadian HIV/HCV Co-infection Cohort, Ontario Cohort Study, Maple Leaf Medical Clinic, Toronto General Hospital and Ottawa Hospital HIV/HCV Cohort [8].
Patient selection and data extraction were performed at the data centres of the participating cohort sites. In provinces with multiple cohorts, VL data were entered from each cohort site and not from a provincial data source. Non-nominal data from each cohort on a predefined set of demographic, laboratory, and clinical variables were then pooled and analyzed at the Project Data Centre in Vancouver. All participating cohorts have received approval from their institutional ethics boards and governance committees to contribute non-nominal patient-specific data to CANOC. Ownership of individual cohort data remains with the contributing cohort and cohort data can only be used for studies approved by the CANOC Steering Committee.

Eligibility Criteria
Eligibility criteria for inclusion in CANOC were documented HIV infection, residence in Canada, aged 18 years and over, initiation of three or more antiretroviral drugs for the first time (i.e. ART-naïve cART start) after January 1, 2000, and a viral load measurement and CD4 cell count within 6 months of the start of therapy. To be included in this analysis, individuals had to have at least one year of follow-up. Viral load measurements were available both before and after starting cART.

Statistical Methods
Three measures were used to assess the frequency of VL testing. First, the primary outcome of interest was defined as a gap between VL tests in excess of 9 months, corresponding to at least three missed or delayed tests assuming the optimal frequency between tests is three months. This was felt to be a clinically important gap in VL testing. Second, the annual rate of VL testing was calculated by dividing the total number of tests for an individual by the duration of follow-up for that individual in years. Third, the time interval, defined as the number of days between two successive VL tests for a subject, was examined.
Demographic and clinical characteristics such as gender, race, HIV risk factors, age, geographic region, CD4 count and type of cART regimen were compared among regions with chi square tests for categorical variables and Wilcoxon rank sum tests for continuous variables.
The proportion of individuals with at least one suboptimal interval and the annual rate of VL testing experienced were compared among regions and by demographic and clinical characteristics with the chi square test and Wilcoxon rank sum test respectively. Generalized Estimating Equation (GEE) logistic regression models, which account for correlation among multiple observations within subjects, were used to determine factors associated with the occurrence of a suboptimal testing interval [9]. An exchangeable correlation structure was assumed for this model. As a form of sensitivity analysis, we repeated the analysis by defining a suboptimal interval as six months or more, corresponding to at least two missed or delayed tests if the optimal frequency between tests is three months.
The relationships between the length of the VL intertest interval and individual characteristics were examined using GEE linear regression models. Plasma VL levels, CD4 counts, characteristics of the antiretroviral regimen and the calendar year of VL tests were treated as time-varying covariates in all regression models. Covariates with a p < 0.10 in the univariate regression models were considered as candidates for inclusion in the multivariable model.

Data cleaning
When there was a gap between two consecutive VL tests in excess of nine months, the accompanying CD4 test dates were examined to validate the VL test dates. Viral load tests within 8 days of each other and with results within 0.1 log 10 copies/mL were considered to be duplicate measurements. In such cases, the first VL date was kept and the average of the VL measurements was assigned as the VL value.

Results
Study cohort and overall VL testing 3,648 subjects met cohort inclusion criteria. Demographic characteristics are described by region in Table 1. The median number of VL tests and the median length of follow-up were 15 (interquartile range (IQR) [9][10][11][12][13]

Analysis of the annual rate of VL testing
In univariate analyses, higher annual rates of VL testing were associated with residence in BC, age > 40 years, white race, male gender, later year of cART initiation, pretreatment VL ≥ 10 5 copies/mL, an HIV risk category of "men who have sex with men" (MSM), history of an AIDS-defining illness, boosted-protease inhibitor (PI)based cART regimen, not being co-infected with Hepatitis C and not being an injection drug user (IDU) ( Table 2).

Analysis of gaps of greater than nine months and six months
Of the 3,648 patients eligible for the analysis, 26% and 51% of the population had experienced at least one nine-month and one six-month gap during their followup, respectively. Proportions of patients with at least one nine-month gap and with at least one six-month gap are shown by demographic and clinical characteristics in Table 2. Results of univariate GEE logistic regression models are shown in Table 3. In the multivariable GEE logistic regression model (Table 4), gaps of both nine and six months were significantly more likely to occur in Ontario and Quebec and among IDUs. Gaps of both nine and six months were significantly less likely to occur among older individuals, among MSM, in recent calendar years of VL test, among individuals on cART, in the first year of cART and if the VL had been suppressed at the previous visit.

Analysis of the time interval between successive tests
In univariate GEE linear regression models, covariates which were associated with a decrease in the number of days between VL tests included age, male gender, later year of initiating cART, being within the first year of cART, receiving PI-boosted cART regimen, higher baseline VL and having been diagnosed with an AIDSdefining illness (

Discussion
There was considerable regional variation in annual rates of VL measurement documented in this Canadawide study. This variation remained significant even after adjusting for demographic variables such as age and HIV risk factors and clinical variables such as year of cART initiation, type of cART regimen, being in the first year of cART, AIDS-defining illness and whether or not the previous VL was below the limit of detection. In Ontario and Quebec, VL was measured quarterly on average. In BC, rates of VL measurement were even more frequent than guidelines suggest, with an average measurement frequency of almost five times annually. There are a number of possible explanations for the regional differences in rates of viral load measurement. Some of the difference in measurement rate was due to regional differences in demographic factors such as the proportions of IDUs, who typically have less frequent viral load testing, or pregnant women, in whom viral load is monitored more closely. Further differences could be due to variation in rates of VL blips among regions, after which a repeat VL measurement is typically ordered. Recent data has documented differing rates of blips by VL assay [10,11] and this may explain the higher rates of testing in BC. Regional differences in VL testing policies may also have an impact. While there are no differences among provinces in the official guidelines for the frequency of VL measurement, it is possible that there are differences in the implementation of the guidelines. In Ontario, a VL will not be performed by the laboratory if one has been done within the last 14 days. Furthermore, rates of VL measurement may be higher in British Columbia due to the fact that all antiretroviral drug distribution and VL testing is coordinated through a single center in the province. Lastly, differences in participation rates among provinces in research studies, which may require more frequent VL testing may explain some of regional variation.
Effective therapy should result in at least a 90% or 10-fold (1.0 log 10 copies/mL) decrease in plasma VL in the first month and suppression to below 50 copies/mL by 24 weeks, depending on the pretreatment VL level [1]. Current guidelines suggest once VL suppression to below 50 copies/mL is confirmed, it should be assessed at regular intervals (e.g. every 3 or 4 months). Isolated episodes of low-level viremia ("blips") are not necessarily predictive of subsequent virologic failure, but consistent elevations of VL above 50 copies/mL meet a strict definition of virologic failure. Emergence of a detectable VL in a previously suppressed patient (i.e. previously consistently < 50 copies/mL) mandates re-evaluation of the case, including repeated testing to confirm whether this represents a "blip" or virologic failure. Confirmed VL rebound should prompt a careful evaluation of regimen tolerability, drug-drug interactions, resistance and patient adherence. CD4+ cell counts should generally be assessed in concert with VL.
In terms of the demographic findings, our findings are similar to those of the study conducted in Ontario several years ago and could likely be generalizable to other settings with universal health care programs. In settings with user-pay programs for virologic monitoring, it seems likely that socioeconomic factors will have a considerably greater influence on rates of testing. In resource limited settings, availability and travel distances are likely to remain significant barriers to regular monitoring of VL levels.
Our results have important implications for guidelines and for research into the monitoring of VL. For patients for whom VL is measured less frequently, virologic rebound will be detected later on average and patients will remain on failing regimens longer than is necessary. In some circumstances, however, it is safe to measure VL less frequently [12]. A reduced monitoring schedule would result in cost savings due to both the cost of the laboratory testing and the cost of the physician time. As ART is a life-long commitment, cost considerations are not insignificant. Further, less frequent VL monitoring would reduce the inconvenience to the patient, which may improve overall compliance. However, while it may be safe and cost effective to extend the period between VL measurements from three to six months in patients with virologic suppression, periods in excess of nine months are less easily justified and place patients at unnecessary risk. Steps should be taken to facilitate more careful VL monitoring in groups at high risk for gaps in testing. A strength of our study is that VL monitoring is provided free of charge to all HIV-positive individuals in Canada, so that we were able to examine correlates of rates of VL testing in the absence of financial barriers to testing. Further, by combining data from nine sites across Canada, we achieved a large sample size as well as a good representation by region, gender, race, and other demographic characteristics. CANOC represents about a quarter of all HIV-positive people in the country who are currently on ART and approximately half of those who have initiated on more modern regimens since 2000 [13]. This is the largest sample of HIV-positive people on ART compiled in Canada and represents one of the most representative samples put together in a high-income country. One limitation of this dataset is that this information is not currently available on a nation-wide basis, so regional differences in provinces other than BC, Ontario and Quebec could not be examined. A further limitation is that data were missing on race and HIV risk factors for many patients.
Future plans for investigation include further examination of the reasons for the regional differences in rates of VL testing. Also, an economic analysis to determine if the less frequent testing found in Ontario and Quebec is adequate in terms of clinical outcomes and could result in financial savings would be informative. Particularly, an analysis of VL testing frequency in patients with full virologic suppression where less testing may be adequate would be of interest and its impact on cost savings could be important.

Conclusions
In our setting, with universal health care and similar regional guidelines for viral load testing, significant variation in rates of VL measurement and the probability of a significant gap in testing were related to geographic region, age, HIV risk factors and clinical variables such as year of cART initiation, type of cART regimen, being in the first year of cART, AIDS-defining illness and whether or not the previous VL was below the limit of detection.