Out of the 5018 patients, 3158 (62.9%) were women as presented in Table 1. The median age was 37 years (IQR: 18 - 62 years) and 642 (12.8%) patients were older than 50 years. The majority of the patients 1938 (58.9%) were married and 38.6% attended secondary school or higher. 4276 (85.2%) patients were living in Cotonou or nearby. The HIV testing was mainly performed due to clinical suspicion

of HIV for 3712 (74%) patients. The BMI was lower than 18.5 kg/m² for 2966 (59.1%) patients. HIV-1 was the leading type of HIV virus found among them (92%). 3753 (74.8%) presented to care with WHO Clinical stage 4 and 3288 (65.6%) had CD4 count lower than 200 cells/μL. The proportion of patients with a CD4 cell count < 350 cells/μL was 78.6%. The median CD4 count at presentation was 192.3 cells/μL (IQR: 47 - 223). The number of patients by years was not the same for the 12 years. As presented in Table 2, the year 2009 was the most represented with 613 patients and the year 2003 was the less represented with 221 patients.

The prevalence of late presentation (LP) was 84.4% (4233 patients out of 5018) and that of Advanced HIV Disease (AHD) was 62.3% (3128 patients out of 5018) for all the 12 years of the study (Table 2).

The trends of the proportion of patients presenting late to care and those presenting to care with Advanced HIV Disease is presented on Table 2. First the prevalence has decreased during the 12 years. Secondly, we find out that from 2003 to 2009, the prevalence decreased from 97.7% to 77.7% for the late presentation and from 88.7% to 49.9% for the late presentation with Advanced HIV Disease. Between 2009 and 2014, the proportion has remained stable around 80% for late presentation and 50% for AHD.

The factors associated with late presentation to care are presented in Table 3.

Patients aged more than 50 years [Odds Ratio (OR): 3.19; 95% Confidence Interval (CI): 2.32 - 4.14] and those between 40 and 49 years (OR: 3.17; 95% CI: 2.52 - 4.00) were at higher risk of late presentation to care compared to those aged < 30 years. Compared to women, men (OR: 1.85, CI: 1.56 - 2.22) were more likely to present to care later. Not educated patients (OR: 2.13, CI: 1.77 - 2.56) were more likely to present late to care compared to educated. Patients without

an occupation (OR: 0.70, CI: 0.55 - 0.90, p = 0.008) were less likely to present late to care than those with a profession. The location was not associated with late presentation in our study. The reason for HIV screening was significantly associated with late presentation to care. Patients screened after clinical suspicion for HIV were 5 times more likely (OR: 4.85, CI: 3.70 - 6.37, p < 0.001) to be late presenters compared to screening during PMTCT. Another key finding was that the year of presentation to care was also associated with late presentation. In 2003 (OR: 13.97, CI: 5.59 - 34.89, p < 0.001) and 2004(OR: 11.84, CI: 6.24 - 22.49, p < 0.001), patients were respectively 14 and 12 time more likely to be late presenters to care than those presenting to care in 2014.

In multivariate analysis, age, BMI and year of presentation to care were predictive factors for late presentation to care. (Table 4)

In our study, 84.4% of HIV-infected people were late presenters and 62.3% were late presenters with Advanced HIV Disease. This high proportion of late presentation was similar to the 85.4% for late presentation and 63.0% for Advance HIV Disease reported from Nigeria in 2013 [12] . However it was higher than that obtained from European studies, where it varied from 48.1% to 55.2% for Late Presentation and from 29.4% to 37.8% for Advanced HIV Disease [13] [14] [15] . In Benin as well as in other African countries, the issue of late presentation to care is not a restricted HIV concern. It is common with many diseases and is caused by the social and cultural behaviors toward diseases, financial constraints and faith in spiritual treatments [16] [17] . Patients usually seek care from healers first and they only go to hospitals when advanced stages of their diseases are reached [17] . As far as HIV is concerned, stigma, being unaware of HIV- status could also explain these results [18] . Since we performed a retrospective data analysis, we could not assess whether these factors were involved in our study. We could not also explain if there were a delay between the first HIV positive test and the first presentation to care. Actually, it is possible that many of these patients were unaware of their HIV status prior to their presentation care and there were no delay due to the patient himself [19] . Then, the issue to make people infected with HIV be aware of their status earlier is important.

The burden of late presentation has decreased throughout the 12 years of our study, and patients are presenting earlier to care. This finding was similar to the reports from Nigeria [12] . The changes in guidelines to initiate HIV treatment and the continuous screening campaigns have probably been part of the downward trend. But one important finding from our results was that after 2009, the prevalence was more likely to stabilize around 80%. There was no statistical difference between the prevalence of late presentation from 2009 to 2014. This means

that an important effort should be made to tackle the factors which have never been addressed until now. If this is not done the prevalence will still so high for the upcoming years.

Our results suggested that people aged more than 40 years were 3 times more likely to be late presenters than those under 30 years. Similar results were also found by Jeong et al. [20] . The reason is that the screening campaigns that were carried out targeted mainly young people and offered very little opportunities for old people to be screened early. The reason was that old people were perceived to be at lower risks of being HIV-positive, but they were often infected when they were young and the diagnosis was not made [21] . Moreover, practitioners performed HIV screening for young people than elders when they are hospitalized. This confirms the relevant role of healthcare professionals in recommending HIV testing not only in presence of AIDS-defining events but also without any particular symptom.

Late presentation was higher among men than women, raising the fact that men had fewer opportunities to be tested for HIV than women [22] . Basically, on behalf of the Prevention of Mother-to-Child Transmission (PMTCT) of HIV, systematic screening for HIV was offered to every woman when they were pregnant [23] . When pregnant women were HIV-positive, they were more willing to seek HIV-care early because of the risk of transmission to their child [24] . This explanation is emphasized by our results about the reason for HIV testing, showing that patients diagnosed by clinical suspicion were 5 times more likely to be late presenters than those diagnosed through PMTCT [23] [25] .

The risk of late presentation was higher for the patients with educational level lower than secondary. Wilson and Hoffmann made the same observation in their studies in 2014 [7] [26] . This observation could be explained by the fact that outreach campaigns, were more understandable for those who were enrolled in school and they were therefore screened earlier [27] .

Location was not associated to late presentation in our study. There was no statistical difference between the prevalence of late presentation for patients who were living in Cotonou and those who were living in other cities. This finding is important since it make it clear that the distance to the HIV center and its accessibility are not the issues leading to late presentation. However, in Kenya, Kwobah found that the distance to the health facility was an important factor for late presentation [18] . In this study, the authors considered late presenters as those who presented to care with CD4 cell count ≤ 100 cells /μL. This is probably the reason that explains the differences.

The risk of delaying care when patients were diagnosed with clinical suspicion was 5 times that of PMCT patients. This means that the systematic screening offered to pregnant women for PMTCT is very beneficial for early presentation to care. This practice should therefore be extended beyond the PMTCT framework to all medical consultations so that HIV testing will be offered to anyone who has contact with a health facility. That is known under the name Provider-Initiated HIV Testing and Counselling (PITC).

For the year of presentation to care, patients seen in 2003 and 2004 were 12 and 14 times late presenters than those seen in 2014. This could be explained by the fact that in 2003 and 2004, there was no evidence about the necessity to start earlier the antiretroviral treatment [28] . Furthermore, screening was only based on clinical features and late presentation risk was higher during these years.

One major limitation of this study was the retrospective data collection method. As a fact of matter, it was not possible to assess whether prior knowledge of HIV played a relevant role in late presentation to care. Those who had prior knowledge of the importance of early HIV screening may be willing to seek care earlier than those who had not. But we were not able to assess it. Another backside of the retrospective method was that with this method, we had to deal with missing data for several patients. Also, the CD4 count can be reduced by many opportunistic diseases even though the patient presented earlier to care. Then the definition of late presentation will not be so accurate that it should be. So, regardless of the fact that the definition of the ELPCG was the most commonly used, it exposed our findings to an information bias due to the fact that opportunistic diseases can lower the CD4 count.