Date Published: April 16, 2019
Publisher: Public Library of Science
Author(s): Sarah J. Nyante, Richard Biritwum, Jonine Figueroa, Barry Graubard, Baffour Awuah, Beatrice Wiafe Addai, Joel Yarney, Joe Nat Clegg-Lamptey, Daniel Ansong, Kofi Nyarko, Seth Wiafe, Joseph Oppong, Isaac Boakye, Michelle Brotzman, Robertson Adjei, Lucy T. Afriyie, Montserrat Garcia-Closas, Louise A. Brinton, Barbara Fuhrman.
In case-control studies, population controls can help ensure generalizability; however, the selection of population controls can be challenging in environments that lack population registries. We developed a population enumeration and sampling strategy to facilitate use of population controls in a breast cancer case-control study conducted in Ghana.
Household enumeration was conducted in 110 census-defined geographic areas within Ghana’s Ashanti, Central, Eastern, and Greater Accra Regions. A pool of potential controls (women aged 18 to 74 years, never diagnosed with breast cancer) was selected from the enumeration using systematic random sampling and frequency-matched to the anticipated distributions of age and residence among cases. Multiple attempts were made to contact potential controls to assess eligibility and arrange for study participation. To increase participation, we implemented a refusal conversion protocol in which initial non-participants were re-approached after several months.
2,528 women were sampled from the enumeration listing, 2,261 (89%) were successfully contacted, and 2,106 were enrolled (overall recruitment of 83%). 170 women were enrolled through refusal conversion. Compared with women enrolled after being first approached, refusal conversion enrollees were younger and less likely to complete the study interview in the study hospital (13% vs. 23%). The most common reasons for non-participation were lack of interest and lack of time.
Using household enumeration and repeated contacts, we were able to recruit population controls with a high participation rate. Our approach may provide a blue-print for others undertaking epidemiologic studies in populations that lack accessible population registries.
Case-control is one of the most efficient study designs for studying rare diseases, such as breast cancer. There are multiple types of control groups that can be employed in a case-control study, each with advantages and disadvantages . Most prior investigations of breast cancer risk in sub-Saharan Africa have used hospital-based controls [2–8], a combination of hospital and hospital visitor controls [9, 10], or patients with cancers at sites other than the breast [11, 12]. Hospital-based or cancer patient controls can be appealing: there are readily available subject rosters, contact information, and health histories, as well as greater motivation to participate than the general public. At the same time, hospital-based controls may introduce bias if they are not representative of the population from which breast cancer patients are obtained. Furthermore, it can be difficult in a hospital setting to select controls with conditions that are unrelated to the study risk factors of interest. The relationship of hospital controls to the case population base can also be difficult to assess in sub-Saharan Africa, given the often not well-characterized hospital referral patterns, widespread use of traditional healers, self-treatment through pharmacies, and cultural stigmas which may prevent some from seeking care.
The GBHS was approved by the Special Studies Institutional Review Board of the National Cancer Institute (Rockville, MD, USA), the Ghana Heath Service Ethical Review Committee and institutional review boards at the Noguchi Memorial Institute for Medical Research (Accra, Ghana), the Kwame Nkrumah University of Science and Technology (Kumasi, Ghana), the School of Medical Sciences at Komfo Anokye Teaching Hospital (Kumasi, Ghana), and Westat (Rockville, MD, USA). All participants provided written informed consent.
A collaboration between the National Cancer Institute and three major hospitals in Ghana, the GBHS is a case-control study that aims to examine risk associations and estimate the population prevalence of known and novel breast cancer risk factors in a West African population. We prospectively identified a pool of population controls within a defined geographic area prior to the commencement of the study. Our two-stage sampling methodology (sampling enumeration areas and sampling potential controls within enumeration areas) was based in part on a study of benign prostatic hyperplasia and lower urinary tract symptoms conducted previously in Accra that used a three-stage sampling design to select participants . Our generation of a pool of controls prior to the study’s start, rather than recruiting controls after cases have been recruited, meant that control and case groups completed their study participation during the same time period, preventing potential bias that may have occurred if the two groups were interviewed during different time periods. Population-based sampling has been used previously in Ghana for the conduct of health surveys [15–17] and in other parts of Africa to survey health status and chronic disease risk factors, including a case-control study of Burkitt lymphoma [18–25]. The only breast cancer study conducted in sub-Saharan Africa with a similar recruitment approach is the Nigerian Breast Cancer Study, which recruited population controls from a single community . Overall, 83% of potential controls we contacted or attempted to contact participated in the GBHS, including 94% of eligible women. The Nigerian Breast Cancer Study reported similarly high (98%) participation among eligible controls . Our experience demonstrates that population-based research can be conducted in Africa with relatively little loss of the eligible population due to non-participation.
In summary, we successfully recruited population controls in a breast cancer case-control study in Ghana. By enumerating a random subset of the population within the study catchment area, we were able to identify a set of controls that were representative of the study population base, conditional on the control matching factors of district of residence and age. Our experience demonstrates that traditional, population-based epidemiologic methods can be employed in sub-Saharan Africa. The techniques we used may work in other countries with similar infrastructure challenges to increase epidemiologic knowledge of health in underserved populations.