Persistent disparities in antiretroviral treatment (ART) coverage and virological suppression across Europe, 2004 to 2015

Background: Direct comparisons between countries in core HIV care parameters are often hampered by differences in data collection. Aim: Within the EuroSIDA study, we compared levels of antiretroviral treatment (ART) coverage and virological suppression (HIV RNA < 500 copies/mL) across Europe and explored temporal trends. Methods: In three cross-sectional analyses in 2004–05, 2009–10 and 2014–15, we assessed country-specific percentages of ART coverage and virological suppression among those on ART. Temporal changes were analysed using logistic regression. Results: Overall, the percentage of people on ART increased from 2004–05 (67.8%) to 2014–15 (78.2%), as did the percentage among those on ART who were virologically suppressed (75.2% in 2004–05, 87.7% in 2014–15). However, the rate of improvement over time varied significantly between regions (p < 0.01). In 2014–15, six of 34 countries had both ART coverage and virological suppression of above 90% among those on ART. The pattern varied substantially across clinics within countries, with ART coverage ranging from 61.9% to 97.0% and virological suppression from 32.2% to 100%. Compared with Western Europe (as defined in this study), patients in other regions were less likely to be virologically suppressed in 2014–15, with the lowest odds of suppression (adjusted odds ratio = 0.16; 95% confidence interval (CI): 0.13–0.21) in Eastern Europe. Conclusions: Despite overall improvements over a decade, we found persistent disparities in country-specific estimates of ART coverage and virological suppression. Underlying reasons for this variation warrant further analysis to identify a best practice and benchmark HIV care across EuroSIDA.

Background: Direct comparisons between countries in core HIV care parameters are often hampered by differences in data collection. Aim: Within the EuroSIDA study, we compared levels of antiretroviral treatment (ART) coverage and virological suppression (HIV RNA < 500 copies/mL) across Europe and explored temporal trends. Methods: In three cross-sectional analyses in 2004-05, 2009-10 and 2014-15, we assessed country-specific percentages of ART coverage and virological suppression among those on ART. Temporal changes were analysed using logistic regression. Results: Overall, the percentage of people on ART increased from 2004-05 (67.8%) to 2014-15 (78.2%), as did the percentage among those on ART who were virologically suppressed (75.2% in 2004-05, 87.7% in 2014-15). However, the rate of improvement over time varied significantly between regions (p < 0.01). In 2014-15, six of 34 countries had both ART coverage and virological suppression of above 90% among those on ART. The pattern varied substantially across clinics within countries, with ART coverage ranging from 61.9% to 97.0% and virological suppression from 32.2% to 100%. Compared with Western Europe (as defined in this study), patients in other regions were less likely to be virologically suppressed in 2014-15, with the lowest odds of suppression (adjusted odds ratio = 0.16; 95% confidence interval (CI): 0.13-0.21) in Eastern Europe. Conclusions: Despite overall improvements over a decade, we found persistent disparities in country-specific estimates of ART coverage and virological suppression. Underlying reasons for this variation warrant further analysis to identify a best practice and benchmark HIV care across EuroSIDA.

Background
It is documented that large health inequalities exist across Europe among people living with HIV (PLHIV) as well as for other diseases [1][2][3][4]. In recent years, comparing and characterising differences in healthcare between countries has received growing interest and has become a central component of informing and targeting health policies. Since it was first introduced, the HIV care continuum has been widely adopted as a tool to benchmark the quality of HIV care [2,5,6], and there are several examples of national and local HIV care continua, including a number of European countries [7][8][9][10][11][12][13][14][15][16][17]. The 90-90-90 targets, launched by the Joint United Nations Programme on HIV/AIDS (UNAIDS) in 2014, aim to set goals for improving the HIV continuum of care from diagnosis to virological suppression [18]. More specifically, they state that at least 90% of PLHIV should be aware of their status, at least 90% of people diagnosed with HIV should receive antiretroviral therapy (ART) and at least 90% of those should be virologically suppressed. If these targets are reached by 2020, ending the AIDS epidemic by 2030 should be within reach [18]. It is clear that some countries are close to reaching the targets while others still have a long way to go [2,19,20]. However, in the absence of common definitions for the different steps of the  ART: antiretroviral treatment.
Each bubble represents a country and the area of the bubble is proportional to the number of people under follow-up in each country. Virological suppression was defined as HIV RNA < 500 copies/mL. The two dotted lines indicate 90% of those in care on ART (x-axis) and 90% virologically suppressed among those on ART (y-axis). Individual country estimates based on fewer than 30 people are not shown.
care continuum, country-to-country comparisons have proven difficult [2,6,[20][21][22]. Furthermore, differences in the data sources used to construct the continua of care further complicate international comparisons.
Previous studies comparing the HIV care continuum across countries have been limited by such differences in data collection [19,20], and there is currently a lack of studies with access to internationally comparable data. EuroSIDA has a unique set-up which allows direct comparisons of data between countries. The aims of this study were to characterise country-specific levels of antiretroviral coverage and ART-induced HIV RNA suppression within the EuroSIDA study, and to monitor temporal trends.

Patients
The

Definitions
EuroSIDA participants were followed from recruitment into EuroSIDA until the latest CD4 + T-cell measurement, HIV RNA-measurement, date of most recent visit to clinic, or death. In three a priori defined time periods, Bars indicate 95% confidence intervals. Virological suppression was defined as HIV RNA < 500 copies/mL. a person was considered 'in care' if their first EuroSIDA visit occurred before the end of the time period assessed and their latest recorded visit or CD4 + T-cell count or HIV RNA measurement occurred after the beginning of the time period assessed. People were assessed for being 'in care', 'on ART' or 'virologically suppressed' at the latest of a clinic visit, a CD4 + T-cell or an HIV RNA-measurement in each time period. If neither was available, the midpoint of the period was used. A 12 months window was allowed before the assessment date. This means that if a clinic visit date or a CD4 + T-cell count (and date) was available within the period assessed, an HIV RNA measurement within 12 months prior to that date could be included in analyses. The outcomes of interest were countryspecific percentages of 'on ART', defined as the number of people receiving ART among those in care, and of 'virologically suppressed', defined as the number of people with HIV RNA (most recent measurement) below 500 copies/mL among those on ART. ART was defined as receiving at least three antiretroviral drugs from any class. The cut-off of 500 copies/mL was chosen as not all countries have access to assays with a lower limit of detection of 50 copies/mL. A person in care but with no available HIV RNA measurement within the time period assessed, was considered virologically unsuppressed (missing = failure). People who were followed in EuroSIDA during more than one of the time periods assessed could contribute data to more than one time period. In sensitivity analyses, we assessed the influence of recent recruitment into EuroSIDA and of using an HIV RNA detection limit of 50 copies/mL.

Statistical analyses
In repeated cross-sectional analyses, we compared country-and region-specific percentages of people on ART and virologically suppressed in three time periods: 1 January 2004 to 31 December 2005, 1 January 2009 to 31 December 2010, and 1 January 2014 to 31 December 2015. Country-level estimates were based on pooled results from active EuroSIDA clinics in the country during each time period, and were grouped into regions as listed in Table 1. Because this study had a European focus, data from Argentinian clinics were not   Western Europe reference. Odds ratios are on a log2 scale. Virological suppression was defined as HIV RNA < 500 copies/mL. Total number of people excluded from analysis in panel C: n = 758 (7.6%).   Figure 1C) the highest-ranking country had an ART coverage of 97.0%, and all of those were virologically suppressed. The country with the lowest percentage on ART was a country in Southern Europe with 61.9% on ART among those in care, and the country with the lowest percentage virologically suppressed was a country in East Central Europe, where 32.2% of those on ART were virologically suppressed. Fourteen of 34 countries had levels of virological suppression among those on ART below 90%. In six of 34 countries, more than 90% were on ART and more than 90% were virologically suppressed among those on ART, while 11 of 34 countries reached neither target in 2014-15.  The crude percentage of people who were virologically suppressed among those on ART also increased over time (Figure 2B), and the change over time again varied between regions (p < 0.01 for interaction, Figure  3B).   Figure 4A shows

Sensitivity analyses
In sensitivity analyses, differences in the availability of HIV RNA measurements seemed to explain some of the differences between regions ( Figure 4C). Thus, the lower odds of virological suppression observed in Northern Europe disappeared when excluding people with missing HIV RNA measurements in 2014-15 (aOR = 0.96; 95% CI: 0.63-1.47) and were attenuated in East Central and Southern Europe. Conversely, the lower odds of virological suppression remained for people in Eastern Europe, even when excluding those with missing HIV RNA measurements (aOR = 0.13; 95% CI: 0.09-0.18).
EuroSIDA is an open cohort, and the latest enrolment wave was 2014-16. To account for differences in ART coverage and virological suppression among these newly enrolled patients, we performed sensitivity analyses excluding 2,893 patients recruited into EuroSIDA in 2014-15. Excluding these patients yielded overall comparable results, although unadjusted point estimates for some individual countries did change (data not shown).

Discussion
In this study we provide a picture of variation in ART coverage and virological suppression in a large number of clinics across 34 countries in Europe including Israel, and describe trends over the last decade. We found that overall, ART coverage increased between 2004-05 and 2014-15, as did the percentage of people with suppressed viral load among those on ART, but these overall improvements covered very large differences in country-specific estimates and trends in ART uptake and virological suppression. We also found that in 2014-15, 14 of 34 countries did not reach the goal of virological suppression in at least 90% of those on ART among those persons included in EuroSIDA.
Our data help illustrate that countries may have very different challenges in improving outcomes along the continuum of care. Following the findings of the Strategic Timing of Antiretroviral Therapy (START) study [25] and the subsequent adaptation of clinical guidelines to initiate ART in any person regardless of CD4 + T-cell count [26,27], we may expect ART coverage to increase in the years to come, although not all countries have adapted the recommendations in national clinical guidelines.
Expanding the use of ART will require substantial funding, especially in countries where ART coverage is low. Further, although countries in Eastern Europe experienced the largest improvements in ART coverage over time, some of these countries also have the highest burden of HIV and are likely to face a range of challenges with scaling up their ART programmes.
We chose to include patients with missing HIV RNA measurements in our analyses, while others have attempted to statistically account for [9,28,29] or exclude [7] those with missing values. Our approach may underestimate the true number of people with suppressed HIV RNA, whereas excluding all with missing HIV RNA measurements is likely to overestimate the percentage of people successfully managed. HIV RNA measurements may be missing for people who receive ART for many reasons, including irregular clinic attendance, patient refusal, poor access or limited resources for laboratory monitoring. Measurements may also be missing at clinics with reporting delays or at sites that have reported incomplete data. We were not able to distinguish between these different causes which may have very different implications for the individual patient. However, it is worth noting that the observed differences in levels of virological suppression were not explained by differences in the availability of HIV RNA measurements. Also, our window for including an HIV RNA measurement was broad, and we find it reasonable to assume that a patient without available HIV RNA measurement within a 2-year period is not successfully managed and may signal a potential opportunity for intervention or a missed opportunity for retaining a patient in care, in particular in the 2014-15 study period. Furthermore, a lack of data may in itself be a sign of possible gaps in performance and may thus equally be a sign of an opportunity for intervention [30].
Our definition of virological suppression was based on a single HIV RNA measurement below 500 copies/ mL. In a previous study we showed that current viral load was as good as repeated HIV RNA measurements to evaluate quality of ART care [31]. This is also in line with strategies recommended by the European Centre for Disease Prevention and Control (ECDC) and UNAIDS [2,18]. However, some may consider this a liberal definition of virological suppression and may argue that confirmed measurements are more reliable. If this is true, we expect that our findings overestimate the percentages of people with virological suppression. Our analyses do not account for virological 'blips', commonly observed in the clinic [32], and we may be underestimating the true number achieving virological suppression. We have no evidence that the frequency of 'blips' would vary between countries. On the other hand, our cut-off for virological suppression was high (500 copies/mL), which should reduce the effect of this bias.
Some major strengths of this study should be mentioned. Firstly, we had access to data from a large number of countries, including some with no national data collection structure. Secondly, and in contrast to previous studies [2,19,20], we were able to compare data directly across countries as data was collected in a uniform manner in all countries and across all three time periods. Previous studies have not been able to describe temporal trends, largely due to insufficient follow-up and changes in definitions over time [33]. Finally, we had access to complete data on ART coverage among those followed in EuroSIDA, which is not often readily available in registry studies.
A limitation of our study is that data were based solely on clinics contributing to the EuroSIDA study, which may not be representative of nation-wide care.
EuroSIDA is an open cohort and therefore, differences over time may also reflect recruitment patterns over time. Although not a recommended standard treatment strategy in current guidelines [26,27], switching ART in stable patients with undetectable viral load to dual therapy regimens has gained some ground in recent years [26,34,35]. Our definition of ART does not capture patients on such regimens, which may lead to an underestimation of ART coverage. Also, we did not take eligibility for ART into consideration when describing trends in ART uptake over time. In this context it is important to underline that this study did not aim to evaluate adherence to guidelines. Furthermore, it is worth keeping in mind that, while guidelines for when to start ART have changed over time, the goal of virological suppression among those on ART has not.
Partly due to our study design, we used the definition 'on ART among those in care' rather than among those diagnosed. Compared with the UNAIDS 90-90-90 definition of 'on ART' [18], we expect that our estimates of ART coverage are higher, as the denominator includes only those in care, and the true betweencountry variation in ART coverage may thus be even more pronounced than what we observed. On the other hand, our definition allowed us to directly compare performance at the clinic level without considering the impact of variation in linkage to care.
Between-country variation in HIV treatment and care is likely to reflect a complexity of underlying reasons including differences in patient populations, patient management, healthcare structures, policies, health expenditure, varying local treatment guidelines and more. In addition to presenting the unadjusted, country-specific estimates that are usually presented in cascades of care, we evaluated the odds of receiving ART and being virologically suppressed on ART across regions. In these analyses, we chose to adjust for basic patient characteristics that were expected to vary across regions and over time. One advantage of the unadjusted snapshot approach is that the estimates are easy to interpret and to communicate, e.g. to policymakers. However, we believe that the adjusted analyses add to the understanding of how much of the regional variation may be attributed to differences in patient characteristics and how much may be attributable to differences in the quality of care, which could be targeted in health policy interventions. For example, in unadjusted estimates, people in Eastern Europe were less likely than people in Western Europe to receive ART. However, after adjustment, the odds of receiving ART were higher in Eastern Europe, indicating that some of the differences in ART coverage in the two regions may be explained by differences in patient characteristics. Needless to say, our analyses are only one step towards understanding the complex interplay of factors that lead to variation. However, we believe that our study may help identify potential gaps in care and may help frame new questions that will give us a better understanding of the causes of variation in the quality of HIV care. One gap in care within the individual regions has been identified for people who inject drugs [36]. In this context, our findings emphasise the need to continue monitoring the response to the HIV epidemic and to construct high-quality data collection that may serve as a platform for both local monitoring and international comparisons.

Conclusion
We were able to directly compare data from a large number of clinics across Europe, including some countries that do not have national registries. We found persistent between-country disparities in the level of ART coverage and virological suppression, as well as the rate of improvement over the last decade. EuroSIDA will continue the surveillance of changes and variation in countries' performance in the 'test and treat' era. Current EuroSIDA work aims to explore the underlying reasons for the observed variation, with the goal to identify a best practice and to benchmark HIV care.