Background
While infection with HIV continues to be a major global public health issue [
1], antiretroviral treatment (ART) has successfully reduced morbidity and mortality in people living with HIV (PLHIV) [
2‐
4], and prevented onward transmission [
5‐
7]. This provides the basis for the Joint United Nations Programme on HIV/AIDS (UNAIDS) 90–90-90 target for ending the AIDS epidemic by 2030, the first of which aims for 90% of PLHIV to know their HIV status by 2020 [
8]. In addition to reliable data on the number of HIV diagnoses, a robust estimate of the number of undiagnosed HIV infections is required to monitor the first target. This can provide information on current gaps in testing programmes, particularly among key subpopulations, which can inform strategies to reach those undiagnosed [
8].
Robust case surveillance and vital registration data supports back-calculation approaches to estimate the number of PLHIV and undiagnosed population. Several such approaches have been developed in recent years [
9‐
15]. The majority of countries in the European region of the World Health Organisation have produced national estimates using either the HIV modelling tool from the European Centre for Disease Prevention and Control (ECDC) [
16], or the AIDS Impact Model in Spectrum (denoted herein as ‘Spectrum’) [
12]. Both tools are freely available and recommended in Europe. The ECDC modelling tool can model key subpopulations, in addition to the general epidemic [
16].
Norway is a low HIV prevalence country with a longstanding surveillance system. Clinical AIDS diagnoses have been under mandatory nominal reporting to the Norwegian Institute of Public Health (NIPH) since 1983. Diagnoses of HIV were subject to mandatory, anonymised reporting by clinicians and laboratories since 1986, with diagnoses from 1984 and 1985 reported retrospectively. Reporting on HIV diagnoses was made nominal, and the collection of data on CD4 cell count at diagnosis included, from March 2019 [
17]. In Norway, residents diagnosed with HIV can access ART free of charge [
18]. Pre-exposure prophylaxis (PrEP) has been available since January 2017 [
17]. Key risk groups to which specific testing offers are directed include men who have sex with men (MSM), migrants from HIV endemic areas and people who inject drugs (PWID) [
19,
20].
There were 6468 HIV diagnoses reported to the NIPH from 1984 to 2018 [
17]. Initially, the epidemic was characterised by diagnoses amongst MSM and PWID. Diagnoses peaked in 2008 (
n = 299), driven by increased migration from high-endemic areas like Sub-Saharan Africa (SSA) in the early 2000s [
21‐
23] and an increase in diagnoses among MSM. Diagnoses have declined recently (
n = 191 in 2018), driven by decreases among both Norwegian-born (those born in Norway, and/or of Norwegian origin) MSM, and heterosexual migrants who acquired HIV before migration to Norway. Conversely, there has been an increasing number of diagnoses among migrant MSM reported to have acquired infection before migration. A low, stable number of diagnoses continues to be reported among PWID, heterosexual Norwegian-born, and heterosexual and migrant MSM resident in Norway at the time of infection [
17].
There have been few measures of the HIV epidemic in Norway. A back-calculation approach was used to estimate PLHIV and the undiagnosed population up to the year 2000 [
22,
24]. More recent estimates have been generated by UNAIDS, using the ECDC modelling tool to calculate national incidence, and Spectrum to calculate prevalence. The most recent estimate, using surveillance data through 2017, was 5800 PLHIV (uncertainty bounds 5200–6300), however, the input data for these estimates incorrectly included cases previously diagnosed in another country, before re-diagnosis upon arrival in Norway [
15,
25].
We aimed to produce current estimates of the number of PLHIV and the undiagnosed population in Norway, overall and for six key subpopulations, to monitor progress towards the first UNAIDS 90–90-90 target. The six key subpopulations were; Norwegian-born MSM, migrant (those not born in Norway, and not of Norwegian origin) MSM, Norwegian-born heterosexuals, migrant SSA-born heterosexuals, migrant non-SSA-born heterosexuals and PWID.
Discussion
Estimates of the number of PLHIV in Norway by the end of 2018 using the ECDC modelling tool, and Spectrum (with incidence from both CSAVR and the ECDC modelling tool) were all aligned at around 5000 persons. The estimates in this study were generated excluding cases previously diagnosed in another country and re-diagnosed upon migration to Norway and are thus lower than those previously produced for Norway [
25]. General agreement in the PLHIV estimates produced using these two models was also observed in a study from French Guiana [
37].
From 2008, the overall number of undiagnosed infections decreased. This reflects a peak in new infections and steadily decreasing time from infection to diagnosis among MSM (see Additional files
5 and
6). Analyses from the Netherlands [
38] and Switzerland [
39] suggest that an increase in infections among MSM up to the mid-2000s was the result of increased high risk sexual behaviour following the introduction of ART, as well as improvements in testing leading to earlier diagnosis, which, alongside better treatment options becoming available [
40], led to an eventual reduction in new infections. This decreasing number of undiagnosed infections also reflects changing migration patterns, with a decreasing number of arrivals from high endemic countries in the latter part of the first decade of the twenty-first century [
21,
23].
Overall point estimates of the proportion undiagnosed in 2018 using the ECDC modelling tool ranged from 7.1% (355 undiagnosed infections) using default model assumptions for CD4 distribution, to 10.2% (520 undiagnosed infections) with the Danish CD4 proxy, which is amongst the lowest in Europe, and comparable to estimates from neighbouring Denmark and Sweden [
13,
15,
41,
42]. National surveillance data and clinical data from the capital, Oslo, support that there is a low proportion undiagnosed in Norway, with a decreasing total number of first-time diagnoses, low proportion of late diagnoses and high proportion achieving viral suppression [
17,
27]. Among Norwegian-born heterosexuals and both MSM subpopulations, estimates did not change noticeably with the use of the Danish CD4 proxy. This may provide greater confidence in estimates for these subpopulations, for which there were < 30 estimated new infection in 2018, undiagnosed infections had decreased in recent years, and who collectively had an undiagnosed fraction of 9.5% (247 undiagnosed infections) using default model assumptions, and 9.7% (254 undiagnosed infections) with the Danish CD4 proxy. Among PWID, the Danish CD4 proxy resulted in slightly higher estimates, although interval estimates were wide. Strategies to prevent HIV transmission in PWID are long-standing in Norway [
43] and estimates indicate the incidence and number of undiagnosed are low. These results collectively allow cautious confidence in concluding that Norway has achieved the first UNAIDS 90-90-90 target.
Among Norwegian-born MSM, the undiagnosed fraction in 2018 was around 4% with approximately 50 undiagnosed infections. Migrant MSM had an undiagnosed fraction of around 13%, with a fractionally higher estimated incidence and number of undiagnosed infections than Norwegian-born MSM, but estimates were low. This highlights the success of the prevention strategy among MSM in recent years in Norway, which has been to increase awareness of HIV infection and condom use, increase HIV testing in higher risk settings or partnerships, rapid initiation of treatment and, more recently, the implementation of PrEP for at-risk persons [
17,
20]. Other European countries have reported higher undiagnosed rates of HIV among migrant MSM [
13,
41], and the proportion of MSM infected overseas, and/or with migrant background is increasing in Norway, as well as neighbouring Denmark and Sweden [
17,
44,
45]. This highlights the importance of continuing to consider migrants of all backgrounds in prevention efforts, particularly in a country which saw a notable influx of migrants from HIV endemic areas in the first decade of the twentieth century. This includes ensuring the accessibility and availability of testing, early treatment and PrEP. Migrants arriving in Norway from HIV endemic areas are offered an HIV test within 3 months of arrival, and as residents of Norway have free access to ART.
Among Norwegian-born heterosexuals, there was a higher number of undiagnosed infections, time from infection to diagnosis and undiagnosed fraction compared to the two MSM subpopulations. National surveillance data suggest that a higher proportion of MSM diagnosed with HIV have been tested on their own initiative or as part of a routine health check, whilst Norwegian-born heterosexuals diagnosed with HIV are more often tested due to the onset of clinical symptoms and signs of HIV infection [
17]. Increased awareness of HIV infection, condom use, and early diagnosis and treatment are the most important preventive measures among heterosexuals. In addition, health care providers should consider informing heterosexual clients who disclose high risk behaviour about PrEP.
Estimates for migrant SSA-born and non-SSA-born heterosexuals were characterised by a high proportion of diagnoses among persons infected before migration to Norway, and results varied depending on the use of the Danish CD4 proxy. Estimates of the number of new infections in these subpopulations do not necessarily reflect incidence in Norway, but rather an extrapolation that depends on patterns of migration. Spectrum accounts for the effect of net migration through the assumption that migrants have the same HIV prevalence as the resident population. For Norway, this may not be the case, as there has been considerable migration from countries with much higher HIV prevalence since the early 2000s [
21‐
23]. For the ECDC modelling tool, in-migration is not taken into account. Also, using CD4 count data in modelling subpopulations with a high proportion infected before migration may overestimate the number undiagnosed, as the time from infection to diagnosis will also include the time before the undiagnosed individual migrated, when they could not conceivably be diagnosed by the receiving country. On the other hand, when CD4 count data are not used (using default CD4 assumptions), estimates of the time from infection to diagnosis are not very robust, as they rely only on data of HIV/AIDS and AIDS diagnoses (ECDC modelling tool) or AIDS deaths (CSAVR).
Ideally, data for both CD4 count at diagnosis, and either precise migration data or an appropriate assumption on migration, should be incorporated to more appropriately generate estimates in populations with high levels of in-migration [
44]. In Norway, the history of anonymised reporting reduces the potential to use precise migration data, while surveillance data on CD4 count were not collected prior to March 2019. However, the recent introduction of nominal reporting, collection of CD4 data, and a planned national clinical HIV registry provide the potential for the inclusion of such data for future HIV diagnoses [
17]. The Danish CD4 proxy may provide a reasonable representation of the historical CD4 distribution in Norway, given the similarities in the HIV epidemics, HIV prevention and control measures, target groups for testing, the accessibility of testing, and the availability and accessibility of treatment in both countries [
19,
20,
25,
27‐
30]. We were unable to adjust for all differences within key subpopulations in Norway and Denmark due to the small number of diagnoses in some groups (for example, see Additional file
1, Table C), thus it will be important to further validate the suitability of the proxy with future surveillance data on CD4 count from Norway and Denmark.
The main strengths of the study are that the input surveillance data were good quality national data with complete reporting, while different commonly used estimation methods have been used, at a time when the use of these tools is rapidly increasing as countries seek to monitor their own progress towards the 90–90-90 target. Also, estimates can easily be repeated to monitor progress in reducing infections and the undiagnosed populations. This study also has some limitations, including the data used for model fitting and the models themselves. It is assumed that reported data are correct, including timing of infection with regards to migration. However, a study in Sweden found that a higher proportion of migrants may be infected after migration than reported [
46]. Also, clustered transmission may contrast with self-reported route of transmission, particularly among migrants from SSA [
47]. There is the potential for duplicate reporting during the study period as HIV reporting was anonymised, but the risk is considered low by national experts, who have rigorously screened HIV notifications and followed up to resolve potential duplicate notifications. In the ECDC model, underreporting of AIDS diagnoses could underestimate time from infection to diagnosis. Conversely, the time from infection to diagnosis may also be fractionally overestimated in groups who test regularly, such as MSM, as the ECDC modelling tool does not account for diagnosis during the first 3 months of infection, which is feasible with current screening tests. Also, the ECDC modelling tool relies on observed data on all-cause mortality and out-migration among those diagnosed. This was unknown for 86% of input HIV diagnoses, thus PLHIV estimates are likely slightly overestimated. Collaboration with other data sources, such as a planned national clinical HIV registry, will help to validate PLHIV estimates, particularly with regards to missing outcome data. Estimates of incidence, time to diagnosis and the number undiagnosed are less affected by missing outcome data, thus one approach could be to model the number of undiagnosed infections using the ECDC modelling tool, and add this to clinical registry data [
45]. In Spectrum, key limitations are the absence of CD4 data to inform the CSAVR model fitting and precise data for the number on treatment over time. While the implied mean CD4 at diagnosis over time – based on the fit to new diagnoses and AIDS deaths – appears plausible, actual data should be incorporated in the future when available. Program data were not available for the number of PLHIV on ART and estimates were instead used. Accurate numbers of those on ART can further improve the Spectrum estimates.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.