Despite the availability and known benefits of combined antiretroviral therapy (cART), a considerable proportion of HIV-infected patients have experienced significant immunodeficiency or disease progression before they present to care 1 2 3 4 5 6 7 . These patients often have poor prognoses 8 , because cART reduces HIV RNA and mortality more effectively when it is initiated early 9 10 11 12 13 . In a recent study conducted in France, six-month mortality among patients who initiated care with advanced HIV disease, defined as CD4 count <200/mm³ and/or AIDS, was 13.6 higher than among patients who initiated care early 2 . Mortality rates remained higher among these patients four years after enrolment. In addition to improving outcomes at the individual level, studies have shown that effective prevention counseling and early cART initiation can also reduce the clinical and financial burden of HIV at the population level by reducing HIV transmission rates 14 15 16 17 18

HIV-infected patients who present to care with advanced HIV disease can be stratified into two groups: (i) patients who discover their HIV infection with an advanced HIV disease and thus initiate care late; and (ii) patients who discover their HIV infection earlier but delay initiating HIV care. Although several recent studies have evaluated the factors associated with presentation to care with advanced HIV disease 1 2 3 4 5 6 7 , few studies have assessed the factors associated with presentation to care at this stage, by delay from first positive HIV test to presentation to care. The risk factors for presentation to care with advanced HIV disease may be different in these two groups.

The objective of this study was to determine the frequency and characteristics associated with late HIV testing with an advanced HIV disease and earlier HIV testing followed by delayed presentation to care with an advanced HIV disease among HIV-infected patients in Brussels (Belgium) and Northern France.

We collected data on 1,819 patients who were followed in Northern France and Brussels from January 1997 to December 2007. The initial characteristics of patients in Northern France and Brussels were the following: median age, 35 years (interquartile range [IQR], 28-43) and 36 years (IQR, 30-43; p = 0.07); proportion of men, 70.6% (860) and 56.2% (338; p < 0.0001); proportion of MSM, 44.9% (541) and 28.3% (156; p < 0.0001); proportion of IDU, 3.4% (41) and 1.8% (10; p < 0.0001); proportion of sub-Saharan African immigrants, 21.8% (265) and 53.1% (319; p < 0.0001); proportion of patients who initiated care in 2003-2007, 43.9% (535) and 54.9% (330), proportion of patients with AIDS-defining diseases at enrolment, 16.0% (195) and 11.1% (67; p = 0.005); and median CD4 count at inclusion, 353/mm³ (IQR, 172-548) and 305/mm³ (IQR, 165-485; p = 0.006) (Table 1).

We estimated the frequency of patients who presented to care with advanced HIV disease among 1,819 HIV-infected patients in the Nord-Pas-de-Calais region (France) and Brussels (Belgium) from January 1997 to December 2007. Moreover, within patients with advanced HIV disease we determined the frequency of "late testing" (i.e.; HIV diagnosis ≤6 months before initial presentation to care) and "delayed presentation to care" (i.e.; HIV diagnosis >6 months before initial presentation to care). Finally, we assessed the potential risk factors associated with "delayed presentation to care"/"late testing" and presentation to care with advanced HIV disease. Of the 570 patients (31.3%) who presented to care with advanced HIV disease, 95 (16.7%) were diagnosed early but delayed initiating HIV care for >6 months, and 475 (83.3%) were diagnosed late. Older patients, sub-Saharan African immigrants, non-immigrant men and patients who enrolled in the study between 1997 and 2002 were more likely to be diagnosed late. Risk factors for "delayed presentation to care" and presentation to care with advanced HIV disease were: age 30-49 years, HIV transmission through heterosexual contact or injection drug use, enrolment in the study between 1997 and 1999, and follow-up in Brussels.

There are multiple definitions of "late presentation" or "delayed HIV diagnosis" in the medical literature. Recently, The UK Collaborative HIV Cohort (UK CHIC) Steering Committee defined presentation at a stage when there is a substantial risk of death (i.e., CD4 counts <200/mm³ or clinically-defined AIDS) as presentation with "advanced HIV disease" and presentation with a CD4 cell count below 350/mm³, resulting in a delay in treatment initiation, as "late presentation" 7 . In this analysis we focused on patients with advanced HIV diagnosis and showed that 31.3% of patients initiated care at CD4 counts <200/mm³ or clinically-defined AIDS. This proportion was found to be similar in Northern France and Belgium, and is comparable to estimates from several recent European studies that used the same definition for "late presentation." In these studies, rates of "late presentation" ranged from 27% to 59% 1 2 3 4 5 6 7 . In our study, 16.7% of those with advanced HIV disease were aware of their HIV status but delayed initiating care for >6 months. Thus, late testing, rather than delay in initiating care after testing HIV positive, is the main determinant of presentation to care with advanced HIV disease. Of note, the proportion of patients who initiated care with an AIDS-defining event was comparable in patients with late testing and those with delayed presentation to care. This shows that late testing is also the main reason of presentation to care with AIDS. Few studies, especially in Europe, have focused their analysis on the group of patients delaying HIV care. In the Italian Cohort Naive Antiretrovirals, 26.1% of all HIV infected patients delayed initiating care for >6 months after HIV diagnosis, and these patients represented 26.1% of patients with <200 CD4 cells/mm³ or clinically defined AIDS upon enrolment 24 . This estimate, using the same definition, is higher than the results from our study in Northern France (11.8% with advanced HIV disease had delayed initiating care) but close to those in the Belgium clinical cohort (25.9% with advanced HIV disease had delayed initiating care)

Consistent with previous studies, our findings show that older patients, heterosexuals, sub-Saharan African immigrants, non-immigrant men and patients who initiated care between 1997 and 2002 were more likely to be diagnosed late with advanced HIV disease 25 26 27 28 29 30 31 32 . The risk of "late testing" increased with age, from an OR of 2.00 among patients aged 30-40 years to 4.07 among patients aged ≥50 years, compared to patients aged <30 years. Previous studies have estimated that the risk of "late testing" in older age groups ranges from 3 to 6.5 2 11 24 33 34 35 36 . A recent review evaluated the psychosocial factors associated with HIV testing in high-income countries. The authors concluded that individuals generally get tested for HIV for the first time when they perceive that they have been at risk 37 . Older individuals and heterosexuals may not feel at risk of HIV infection. Our finding that sub-Saharan African immigrants are at risk of presentation with advanced HIV disease corroborates previous reports that this group faces obstacles to HIV testing. A British study that investigated HIV testing practices among African immigrant communities found that although general HIV awareness was high, perception of individual risk was poor 38 . Furthermore, they found that cultural norms and stigma in immigrant communities often contributed to late presentation and remained a major barrier to HIV testing 11 . Immigrants usually present to care with immediate and specific needs, and the idea of getting tested for a disease before having symptoms, even when the risk of infection is high, is often rejected. Our finding that a large proportion of late HIV diagnoses were among non-immigrant men was also consistent with previous studies 2 . While women are systematically tested for HIV during pregnancy, men are not generally offered HIV tests on a routine basis.

Although several studies have investigated the factors associated with "late presentation" and "late testing," few studies have evaluated the characteristics of patients who get diagnosed with HIV early but delay presenting to care. We show that the risk factors for "delayed presentation to care" and "late testing" among patients who present with advanced HIV disease are different. First, we found that IDUs are more likely to test early and delay presenting to care. This result is consistent with previous studies 24 and demonstrates that IDU present with advanced HIV disease not because they are tested for HIV infrequently, but rather because they are not efficiently linked to care. In fact, IDU regularly receive HIV tests in specialized drug treatment centres. Moreover, because they are considered to be at high risk of HIV infection, physicians frequently offer them HIV tests during regular clinic visits. Despite their tendency to be diagnosed early, however, IDU are difficult to enroll and retain in care 21 25 39 40 41 42 43 .

Second, although we found that rates of "late testing" increase with age, we did not find the same trend in the "delayed presentation to care" group. Patients aged 30-49 years were at increased risk of "delayed presentation to care" and presentation to care with advanced HIV disease compared to patients aged <30 years, but patients aged >50 years were not. Unlike patients aged >50 years, those aged 30-49 years may seek an HIV test because of a perceived risk, but then delay presenting to care to avoid HIV-related stigma. The number of patients >50 years in our cohort was small, however, so any interpretation of this result should be made with caution. Finally, we found that patients who were followed in Brussels were twice more likely to be diagnosed early and delay presenting to care than patients who were followed in Northern France. Of note, as stated above, the proportion of patients diagnosed early with presentation to care at advanced HIV disease in Brussels was close to estimates in the Italian Cohort Naive Antiretrovirals 24 . This finding may be related to different testing and counselling services and/or linkage to care services in different countries. Girardi et al. for example have shown in their study that having no counselling at the time of first positive test were associated with a higher probability of delayed presentation to care 24 , but unfortunately this information was not available in our datasets. Differences may be also explained by differences in patient characteristics. Our findings may be for example explained by the higher proportion of sub-Saharan African immigrants in Brussels when compared to Northern France. In multinomial logistic regression analysis, we adjusted our results for patient characteristics, including patient origin, but we may not have adjusted our results for all potential confounders. Differences between our study and the Italian study maybe also related to difference in the time period during which the analysis was conducted. As seen in previous studies, in our study rates of "late testing" and "delayed presentation to care" decreased overtime 24 33 . This trend is likely related to the increasing availability of more effective and less toxic cART regimens, as well as improvements in HIV care.

Our study has several limitations. First, our results cannot be generalized to locations outside of Northern France and Brussels, because the risk factors associated with "delayed presentation to care" and "late testing" may depend on population characteristics and local HIV screening practices. Although most of our results are consistent with previous studies, some risk factors for "delayed presentation to care" and "late testing" may depend on geography: patients in Brussels were more likely to be diagnosed early and delay presentation care compared to patients in Northern France. Second, although most data were collected prospectively, the study was designed after data collection had ended. We were therefore unable to assess the impact of variables such as incarceration, homelessness, illegal immigrant status, unemployment status, HIV testing history, psychological condition, counselling and community support on outcomes. These variables might influence rates of "delayed presentation to care" and "late testing." Third, the delay from initial HIV diagnosis to initial presentation to care, which was used to differentiate patients in the "late testing" and "delayed presentation to care" groups, was often self-reported and thus subject to recall bias. Finally, we may not have had enough statistical power to evaluate some of the risk factors associated with "delayed presentation to care", because the number of patients who were diagnosed early and delayed presentation to care was small.

Despite the wide availability of cART in Western Europe, access to care with advanced HIV disease remains common in France and Belgium, with almost one third of HIV-infected patients initiating care at CD4 counts <200/mm³ and/or AIDS in recent years. This study shows that, although late HIV testing is the most frequent cause of delayed presentation to care, low rates of linkage to care also lead to late presentation. In our study, 17% of patients delayed initiating care after being diagnosed with HIV. Because the factors associated with "delayed presentation to care" differ from those associated with "late testing," different strategies should be developed to optimize early initiation of HIV care in each group. One method for increasing the proportion of patients who are diagnosed early is to broaden HIV screening practices. In France, the Haute Autorité de Santé 44 45 recently recommended one-time universal, routine, voluntary HIV screening in the general population, and more frequent screening in higher-risk groups. General practitioners in particular should be encouraged to provide HIV tests on a routine basis, in order to improve access to care at the individual level and limit the spread of infection at the population level 46 47 . Interventions to improve linkage to care should be implemented alongside HIV screening strategies, particularly in marginalized groups such as IDU, who tend to be diagnosed early but to delay presentation to care.

AIDS: Acquired immune deficiency syndrome; cART: combined antiretroviral therapy; CI: confidence interval; HIV: human immunodeficiency virus; IDU: injection drogue use; IQR: interquartile range; MSM: men who have sex with men; OR: odds ratio.

YY has received honoraria for presentation at workshops and consultancy honoraria from Bristol-Myers Squibb, Gilead, Glaxo-SmithKline, Merck, Pfizer, Roche and Tibotec. In Brussels, Tibotec has helped the implementation of the database. Other authors declare that they have no competing interests.

BN, JS, BV and YY were involved in the overall study design and protocol development. BN, AV, PB, FB, PC, KC, CF, HM and HG collected the data and participated to their interpretation. BN and JS performed the statistical analysis. BN, HM, BV and YY participated in writing the manuscript, which all authors have reviewed and approved.