The protocol for this trial and supporting CONSORT checklist are available as supporting information; see Checklist S1 and Protocol S1.
This study was approved by the Committee on Human Research of the University of California, San Francisco (FWA00000068) and the Ethical Review Committee of the Kenya Medical Research Institute (FWA00002066). The study protocol was also approved by the Associate Director for Science at the US Centers for Disease Control and Prevention. All women participating in the study gave written informed consent for the use of their de-identified data in the evaluation. Participation in the study did not require additional research activities beyond women’s regular antenatal and HIV care. This clinical trial is registered at clinicaltrials.gov NCT00931216.
The primary aim of the Study of HIV and Antenatal care Integration in Pregnancy (SHAIP) is to test if an integrated approach to ANC, PMTCT, and HIV care and treatment provision is an effective service model for low-resource health facilities in rural Nyanza Province, Kenya, with implications for other sites in Kenya and sub-Saharan Africa. In this study, “service integration” has been defined as integrating HIV care and treatment services, including initiation and provision of HAART for eligible women, into existing ANC and PMTCT services, with all services provided by the same health care provider in the same room during ANC visits for the duration of the pregnancy and until a definitive HIV diagnosis of the child (up to 18 months), after which point the woman and infant, if HIV-infected, would be referred to the HIV clinic. Women and infants who presented for HIV care were enrolled on the same day or at the subsequent visit.
We hypothesized that the integrated model would lead to increased uptake of HIV care and treatment for women and infants and thus improved maternal health outcomes and reduced vertical transmission of HIV, as compared to a non-integrated model. While experience and related studies have indicated that integration of services may improve uptake and utilization of HIV care and treatment by pregnant women, full integration needs rigorous evaluation before adoption on a wider scale 
. Thus, in order to ensure provision of the highest quality ANC, PMTCT, and comprehensive HIV care and treatment to pregnant women, a new service model should have strong evidence supporting its effectiveness.
In March 2005, the Kenya Medical Research Institute (KEMRI) and University of California, San Francisco (UCSF), supported by the President’s Emergency Plan for AIDS Relief (PEPFAR)/Centers for Disease Control (CDC), launched Family AIDS Care and Education Services (FACES) in Nairobi and Kisumu, Kenya 
. As one of the PEPFAR local partners in Kenya, FACES works to support the Kenyan Ministries of Health (MoH) in the implementation of quality HIV services including PMTCT at existing public and private health facilities; by providing training, clinical mentorship, logistical support, community engagement activities; as well as employing lay health workers based at peripheral sites and provision of salary support for MoH health workers. FACES currently supports 123 PMTCT sites in Nyanza Province.
As of 2008, HIV prevalence in Nyanza Province was estimated at 16.0% of reproductive-aged women, which is the highest HIV prevalence in the country and more than double the national average 
. The study sites are located in the southern part of Nyanza Province, bordering Tanzania to the south and covering a third of Kenya’s shoreline of Lake Victoria. The main economic activities in this area are fishing and farming.
SHAIP is a prospective cluster randomized trial. This design was chosen because service integration is an intervention that is carried out on a clinic-level, rather than at an individual patient level; it would be practically impossible to randomize individual women at a given clinic to receive either fully integrated (FI) or non-integrated (NI) services. Outcomes for women and infants attending FI clinics will be compared to those of women and infants at NI clinics covering the time period from June 2009 - March 2012. Although clients and providers could not be blinded as to the type of services delivered, study investigators are blinded in terms of knowledge of any outcomes by study arm until locking of the study database.
Prior to study initiation, potential sites (government health facilities in the study area) were assessed to determine the size of the facility, staffing, patient load, and available services. Inclusion criteria for sites included the following: 1) providing ANC services, 2) providing HIV testing and ARV services for pregnant women, and 3) an average of at least 20 new ANC clients per month. Study sites included all the government health facilities that met these criteria and were also either providing or scheduled to begin providing HAART services in these districts by June 2007 (excluding Migori District Hospital, which was much larger in terms of staffing and patient volume, and thus very different from other health facilities in the districts, and would have created imbalance in the study arms) (See ). Twelve facilities were categorized by facility type, as either “Health Center or Dispensary” (N
8) or “Hospital” (N
4). Within these strata, each clinic represented a cluster, and was randomized to either control (NI) or intervention (FI) using the ACluster software 
. Prior to beginning study enrollment, each site had to have begun providing comprehensive HIV care and treatment services including HAART; site staff had to have completed the study training program, including research ethics; and a site initiation visit had to have been completed by the external monitor. A map of Greater Migori District, showing the locations of control (NI) and intervention (FI) sites is presented as .
Characteristics of the twelve clinic sites.
also shows baseline measures of the proportions of a) HIV+ women receiving PMTCT prophylaxis other than HAART, b) HIV+ women receiving HAART, and c) HIV-exposed infants receiving prophylaxis, using data from program PMTCT reporting tools for the study sites during the six months prior to study initiation in June 2009 (December 2008–May 2009). These proportions were very similar for FI and NI sites at baseline before study enrollment began: 87.3% FI vs. 86.0% NI for PMTCT prophylaxis received; 11.6% FI vs. 12.3% NI for HAART received; and 95.1% FI vs. 91.1% NI for infant prophylaxis received. It should be noted that these proportions are based on monthly data from clinic registers and may not reflect actual use/adherence.
The study population included all pregnant HIV-positive women, 18 years and older who were not currently enrolled in HIV care and treatment, as well as infants born to the women enrolled in the study. Eligible women were introduced to the study at the ANC clinic and if they were interested in participation they were taken through an informed consent process in a private location by a trained lay health worker. A total of 1,172 HIV-positive pregnant women were enrolled in the study during the period June 2009–March 2011 (See ). Of a total of 1,200 eligible women who were offered study enrollment, 81 (6.8%) declined to participate in the study.
The main study outcome measures are rates of maternal enrollment in HIV care and treatment during the 12-month follow-up period, infant HIV testing uptake by 3 months of age, and HIV-free infant survival at 6 months of age. Maternal enrollment in care is assessed by the existence of a completed HIV care enrollment form for the woman in the FACES electronic medical record system, and by calculating time from the date of study enrollment to the date of enrollment in HIV care. Infant testing uptake by 3 months after the birth is measured using a combination of medical record data and study tracking tools. Mothers were encouraged to bring their infants to the clinic for HIV testing at 6 weeks of age and 6 weeks after the complete cessation of breastfeeding, and the FACES program attempted home visits for HIV-exposed infants who were not brought to the facility for HIV testing by 3–6 months after the birth (see below). HIV-free survival at around 6 months of age is calculated by assessing the numbers of infants who were not lost-to-follow-up and were still alive and had not tested HIV-positive at around 6 months of age, based on data from medical records and home visits after the birth.
Infant HIV-Polymerase Chain Reaction (PCR) testing was done on dried blood spots using Roche DNA PCR version 1.5 (Roche Diagnostic System). Infant samples were collected from skin prick and coated onto filter paper. The spotted filter papers were allowed to dry for at least 4 hours at room temperature and placed in individual zip locked bags containing a silica desiccant. All these samples were then transported to the KEMRI-CDC HIV-R laboratory in Kisumu for PCR testing. This laboratory handles all Infant HIV-PCR tests for Nyanza Province. Specimens for CD4 testing were collected in EDTA whole blood collection tubes. CD4/CD3 fluorescent labeled monoclonal antibodies (Becton Dickinson Biosciences) were used to quantify the CD4 absolute count using BD FACSCount™. Additional outcomes to be assessed for HIV-positive women include: receipt of PMTCT prophylaxis, initiation of HAART (among HAART-eligible women), time to initiation of HAART, changes in CD4 counts, and retention in and adherence to HIV care. Samples for CD4 counts are ideally collected at the time of study enrollment and every six months or more frequently, if clinically indicated. Changes in women’s CD4 cell counts from baseline (study enrollment) to approximately 6 months following study enrollment will be assessed using medical record data. Monthly tracking tools and electronic patient records are used to collect data on enrollment in HIV care, initiation of HAART, retention in care, and number of missed visits. Additional study outcomes to be examined for infants include enrollment in HIV care and uptake of HAART, if HIV-positive.
Sample Size and Power Calculations
From preliminary site assessments of the clinics conducted in 2006, it was estimated that 6,564 new ANC clients would be seen in a 12-month period at the 12 study sites. It was expected that an average of 80% of pregnant women would agree to HIV testing. At the time when the sample size calculations were performed, the most recent estimate of HIV prevalence among pregnant women in Nyanza Province was 25% 
. These assumptions resulted in estimates of approximately 1,313 women testing HIV-positive with a previously unknown HIV-serostatus, and 3,938 being HIV-negative after the first ANC visit. Assuming that 90% of the women who tested HIV-positive agreed to participate in the study, we estimated being able to enroll 1,182 HIV-positive pregnant women in the study over a period of 12 months.
Power calculations were made based on estimated differences in vertical transmission rates in the two arms of the study, accounting for stratification, based on prior published transmission rates using similar antiretroviral protocols in Thailand and Cote-D’Ivoire for reference 
. With a sample size of 12 clusters, 591 HIV-positive women per study arm, and an average of 98 HIV-positive women per health facility (average in the Hospital stratum of 146 women per facility and average in the HC/Dispensary stratum of 75 women per facility), we calculated that we would have 80% power to be able to detect a minimum odds ratio of 2.02 for vertical transmission, if we assumed a conservative intracluster correlation coefficient (ICC) of .01. We also calculated our minimum detectable effect size with a range of ICCs 
and found that the minimum detectable odds ratio ranged from 1.81 for a very low ICC of .005, to 3.75 if the ICC is as high as .05.
With 99% power, this sample size allows us to detect a mean difference of 50 cells in CD4 counts between groups, using an estimated standard deviation of 142 cells and an intra-cluster correlation coefficient of .01.
Our power will be 90% if the ICC is .025, and 80% if the ICC is as high as .037. This sample size also provides us with exceptional power to examine the other study outcomes including women’s enrollment in HIV care, pediatric HIV testing uptake, and loss-to-follow for those who enroll in HIV care.
Description of Procedures
Health care providers at both FI and NI facilities were trained and supported to provide high quality ANC, PMTCT, and HIV care and treatment services throughout the duration of the study. Refresher trainings were provided on a regular basis, and were quickly organized in response to significant staff turnover at a site and/or changes to the Kenyan national PMTCT guidelines. Study-specific training sessions for the FI and NI clinics included study procedures, correct filling of patient medical record forms, clinic flow and logistics for FI or NI services, informed consent procedures, and participant eligibility criteria. Clinical trainings covered the components of high quality ANC, PMTCT, and HIV treatment for pregnant women; early infant diagnosis; and care and treatment for HIV-exposed infants and HIV-infected children.
The package of ANC, PMTCT, and HIV care and treatment services offered at the sites has followed current Kenyan national guidelines in both study arms for the duration of the study, with the only difference in care being the clinic location of HIV care and treatment services for pregnant and postpartum women (ANC clinic versus the HIV clinic). At the NI clinics, HIV-positive women were given ARVs for PMTCT, clinical staging to determine eligibility was conducted, and a specimen was taken for a CD4 count. They were also given a referral form to enroll in the HIV care and treatment clinic. At the FI clinics, HIV positive women were provided all ANC, PMTCT, and HIV services in the ANC clinic including HAART for women who were eligible. (See and ).
Clinical services provided in the ANC Clinic at Fully Integrated sites and Non-Integrated sites.
As mentioned previously, over the course of the study, national guidelines concerning infant feeding, CD4 count monitoring, and HAART initiation were updated. The sites complied with each set of changes shortly after the national guidelines on PMTCT were announced with the exception of daily NVP for the infant, which was rolled out to the study sites in October 2010. The changes were implemented at all study sites simultaneously.
Data Collection Methods
Quantitative data on women and infants were collected prospectively as part of their routine clinic care. FACES uses an electronic medical record system using the Open Medical Record System (OpenMRS®) platform. In the study area, paper medical record forms were filled out by clinical staff at the health facilities. Data clerks entered these forms into laptop computers on a regular basis, and subsequently these data were merged with the larger FACES database. This dataset includes detailed information on socio-demographics, health indicators (CD4 counts, WHO Staging, vertical transmission, etc.) and use of specific HIV and maternity services by HIV-positive pregnant women who attend FACES-supported sites. For each study participant, data were abstracted from the FACES database until her infant was approximately six months old (12 months after study enrollment). In addition, study registers were used to monitor study recruitment, enrollment in HIV care, infant HIV testing uptake and results, and for follow-up of women who do not return to the health facility for a return visit.
Regardless of study facility or arm, mothers were encouraged to bring their infants to the clinic for HIV testing at 6 weeks of age and 6 weeks after the cessation of breastfeeding. In the initial study period, if a woman had not brought in her child for testing by the time the child was 7 months old, a FACES staff member conducted a home visit to ascertain whether the infant had been tested or not and to encourage the family to bring the infant in for testing, and for the mother and other family members to enroll in HIV care and treatment. Starting in November 2010, FACES began to support home-based testing services for HIV-exposed infants at the 12 facilities engaged in the SHAIP trial. FACES staff were trained to offer dried blood spot (DBS) infant testing during a home visit for HIV-exposed infants who had not been brought to the clinic by the age of three months. During these home visits, mothers are asked for permission to test their infant for HIV at the home using a heel-stick DBS or are referred to the clinic.
Site assessments of the clinics were performed every six months to monitor service delivery parameters and various aspects of service integration, including fidelity to the randomized service model. These assessments were also used to identify needs for refresher trainings due to staff turn-over and to make other changes such as addressing insufficient supply of commodities such as HIV test kits and condoms.
External monitoring for this study was conducted by FHI 360, including initial site assessment visits, interim monitoring visits, and study close-out visits. Internal monitoring such as form completion and quality control of data entry were conducted on a monthly basis by the study team. Data collection was completed as of March 31, 2012 and the database was closed as of June 29, 2012.
Analyses will examine the effects of integration on service uptake and retention at each step of the PMTCT cascade, as well as effects on vertical transmission of HIV and maternal health outcomes. Analyses will mainly be performed at the cluster (health facility) level and any individual-level analyses will take into account cluster effects 
. Primary analyses will be based on the “intention-to-treat” (ITT) principle, with facilities randomized to the intervention group classified as delivering “fully integrated services” regardless of whether or not individual women actually received integrated services. In addition, we will perform an according-to-protocol (ATP) analysis, taking into consideration the assessments of integration performed every six months at each site. Assessment data will be used to compute an “integration index” which takes into account various aspects of the integration of services including the location and functioning of trained staff, services, infrastructure, equipment, and supplies. This index will be used to assess variability in the service delivery models for both intervention and control clinics.
In this initial paper, we compared participant characteristics between participants enrolled at intervention (FI) and control (NI) sites using proportions for categorical variables and means or medians (as appropriate) for continuous variables. An assessment of the distribution (normally or skewed) of continuous variables was performed. Socio-demographic and clinical characteristics of participants enrolled at intervention and control sites were compared using methods to account for the similarity (intracluster correlation) between participants enrolled at each site. Statistical tests used included chi-square tests for categorical variables and Student’s t-tests for continuous variables, and were adjusted for clustering effects using the clttest and clchi2 routines in Stata version 11 (StataCorp., College Station, TX, USA).