This study examined antenatal ART initiation in eligible pregnant women in Cape Town during 2008. Although overall levels of ART initiation were relatively low, the findings suggest that an integrated model of antenatal ART initiation may be associated with higher ART uptake compared to models that separate ANC and ART services.
The proportion of women starting ART in pregnancy in these data from 2008 (46%) is slightly lower than the corresponding proportion from the same facilities in 2005 (51%). These persistently low levels of antenatal ART initiation point to the ongoing challenges in starting ART during pregnancy. Health services for ART initiation in adults expanded in Cape Town between 2005 and 2008 (doubling from 33 clinics in 2005 to 66 in 2009). However, these general ART services are often not oriented to the needs of pregnant women. First, pregnant women tend to be clinically stable compared to other eligible adults 
and may not receive adequate attention in the general pool of more morbid patients initiating treatment. Second, there are unique psychosocial barriers facing pregnant women starting ART which receive little attention in routine ART counselling models 
Here, the integrated model for starting ART in pregnancy saw a higher percentage of women initiating ART before delivery. In this model, the vast majority of women (95%) who started ART antenatally did so within the ANC, compared to lower proportions in the other models with referral ART services. Previous studies have reported similarly low rates of ART initiation among women referred to centralised ART services 
. It may be possible that with increased distance between antenatal and ART services, factors such as convenience or desire for privacy may adversely affect referral and uptake of ART. For example, qualitative research from Malawi has suggested that pregnant women have a preference for integrated ART services over access to ART in general primary level services where they would be required to mix with HIV-infected men and non-pregnant women with more advanced HIV disease 
These findings also demonstrate that regardless of the model of care involved, late antenatal presentation is a persistent barrier to antenatal ART initiation. In this setting, women presented for care in pregnancy into the second and third trimesters, decreasing the time available for antenatal ART initiation. The phenomenon of late antenatal presentation is a well-known concern in maternal and child health 
, and our findings are consistent with other studies 
. The frequency of late antenatal presentation in this and other settings means that expediting ART initiation in eligible women is critical, as are complementary efforts to encourage women to attend antenatal care earlier 
. The median time to treatment in women who initiated postpartum was 34 weeks, which may suggest identification and linkage to care through baby immunization services. More research is required to understand the reasons for delayed initiation in these women who were not lost to the service. Under the integrated model of care, women started ART more quickly (median delay from antenatal presentation to ART initiation, 36 days) than in other models (median delay, 54 and 59 days at Sites 2 and 3, respectively). Several studies have suggested that each additional week of ART provided before delivery results in a significant reduction in the risk of vertical transmission 
pointing to additional potential benefits to an integrated model of care for PMTCT.
Our findings are particularly important given the interest in universal ART initiation for all HIV-infected pregnant women, regardless of CD4 cell count 
. Implementation of the World Health Organization’s “Option B+” strategy would dramatically increase the numbers of women eligible to start lifelong ART in pregnancy in South Africa and other high-prevalence settings, and would also require new service delivery approaches that can assist in starting ART as quickly as possible during pregnancy. Integration of ANC and ART services presents one valuable strategy to achieve these aims. However, it is important to note that this research focuses on ART initiation, and there are separate concerns regarding treatment adherence and retention in care among women starting ART during pregnancy 
. These issues are likely to persist across models of care, and will require specific attention in the design and operation of integrated or separated ART-ANC services.
The interpretation of these data comes with several limitations. There may be important differences between the three service delivery models other than their approach to antenatal ART initiation and thus it is difficult to infer that the increased antenatal ART initiation at Site 1 is attributable solely to the integration of services. For example, we did not assess whether the time for treatment workup differed between the models. Each ART eligible woman would have required both clinical and psychosocial assessment prior to initiation and it is possible that approaches varied between the models and within the sites depending on service provider. This may have negatively impacted on women who presented in late pregnancy in particular, due to there being little time for work up. Despite this, guidelines did not preclude women in advanced pregnancy (>36 weeks gestation) from initiating ART. Related to this, it is important to note that definitions of service integration related to HIV/AIDS and reproductive health vary widely 
. The integration studied here may not be appropriate in all settings and alternative approaches to integration may be more relevant in other health systems contexts; further research into integrated models of ANC and ART is warranted.
This research was conducted in an urban setting with a high antenatal HIV prevalence and high-volume ANC and ART services, and the findings should be generalized with caution. This analysis was conducted under previous WHO guidelines (with a CD4 threshold for ART initiation of 200 cells/uL) and the numbers of women requiring ART have almost doubled since the implementation of the 2010 WHO guidelines; it is unclear whether such increases would alter the differences shown here between integrated and separated services. In addition, these data come from a retrospective review of clinical records, and hence the accuracy and completeness data may be suboptimal, though there were no differences in the levels of missing data between service delivery models.
In summary, with the efficacy of drug regimens for PMTCT well-established, PMTCT programme impact is dependent on the implementation of appropriate and effective service delivery models 
. This study suggests that integration of ART into routine antenatal care services can lead to significant improvements in ART initiation during pregnancy, increasing both the proportion of eligible women who start ART and the duration of ART received before delivery. While further research is required, integration of ART into antenatal care represents a valuable approach for promoting maternal and child health in the context of HIV/AIDS.