The current study combined data from two cross-sectional studies of HR-HPV in Malian women. Study I recruited urban-dwelling Malian women. The urban study was auxiliary to a study to evaluate the accuracy of visual screening techniques in low-resource settings. A description of that parent study was previously published (see Sankaranarayanan et al.
]). Another publication assessed the validity of using self-collected cervical samples, and presented a basic analysis of correlates of HPV infection in the urban sample only (see Tracy, JK et al.
]). Study II (described in more detail in this paper) recruited rural-dwelling women. Both study protocols were approved by the local institutional review committee in Mali and the Institutional Review Board of the University of Maryland School of Medicine.
Two samples of women were recruited separately: an urban sample from Bamako in March 2007, and a rural sample from the village of Naréna, approximately 100 km (60 mi) southwest of Bamako, in August 2008. The urban sample was population-based, recruited from an age-stratified random sample of women who were selected from a census registry sampling frame. In the rural area, where no census registry existed, women were recruited from the catchment population of a community health center via word-of-mouth advertisement. Participating women in both settings were required to be 15 years or older and provide informed consent. Women under the age of 18 were also required to have verbal consent from a parent or guardian. Women were excluded if they self-reported: (i) history of hysterectomy with removal of the cervix; or (ii) history of anogenital, breast, oral, esophagus, lung, bladder, liver, or cervical cancer.
By local custom, community meetings were held prior to recruitment in both the urban and rural setting. In the urban setting, community recruiters visited the homes of each of the randomly-selected women to invite each to participate in the study. Women who agreed were then assigned appointments at the research clinic, where study personnel explained the study to the women first in groups and then in greater detail individually. An interviewer obtained informed consent in the privacy of an interview room. A printed consent form was translated by a trained investigator into the local language and read to the study participant, and the participant’s signature, index fingerprint, or witnessed verbal affirmation was obtained as indication of consent to participate.
In the rural setting, the study days and times were advertised throughout the community via word-of-mouth. On the study days, potential participants arrived at a community health center, where they underwent an eligibility review and consent process similar to that in the urban setting. The interview was then administered in a semi-private area. Subsequently, the woman entered a private examination room, where a gynecologist performed the pelvic examination and specimen collection.
A standardized interview was used to obtain detailed information on sociodemographic factors and sexual and reproductive history. The self-reported value for age should be considered an estimate, as many women, particularly in the rural setting, may not know their exact age. The questionnaire was derived from a more extensive interview used by the International Agency for Research on Cancer (IARC) during an international multisite international study of the epidemiology of cervical cancer [16
]. Each interview was conducted prior to cervical screening and specimen collection. A sample questionnaire in English is available (see Additional file 1
Clinical examination and specimen collection
Clinical examinations and specimen collection were performed according to standardized procedures [15
]. A clinician inserted an unlubricated speculum and examined the cervix with the aid of a focus lamp. Cervical cells were collected with a Digene female cervical swab collection kit [15
]. Swabs were then placed into a 5-ml vial containing 1 ml of specimen transport medium. Samples were stored at -80°C at a microbiology laboratory in Bamako until shipped to the US for testing.
HPV DNA testing
HPV DNA testing for samples collected from the urban women was performed in the US using Digene Hybrid Capture II (HC2; Digene, Rockville, MD, USA) according to the manufacturer’s protocol. The HC2 is a pooled-probe, nucleic acid hybridization assay with signal amplification that uses microplate chemiluminescent detection. Each sample was determined to be positive or negative for one of 13 high-risk types of HPV, defined as HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59 and 68 [9
]. Samples tested by HC2 were considered positive if the viral load was greater than or equal to the mean viral load of the manufacturer-supplied positive control, which contained 1 pg/mL of HPV DNA, corresponding to 5000 or more viral copies. Samples collected from women in the rural group were sent to a different laboratory in the US for HPV DNA testing, which used HC2, Amplicor (using a similar cutoff based on a manufacturer-supplied positive control), and Linear Array (LA)(both Amplicor and LA by Roche Molecular Systems, Pleasanton, CA, USA). Amplicor is a pooled-probe, polymerase-chain reaction (PCR)-based test that targets the same HPV types as HC2, and LA is a type-specific primer PCR assay for 37 HPV types. Specimens testing positive for any of the 13 HR-HPV genotypes by any test or for HPV-66 (because that type strongly cross-reacts via HC2) by LA were considered infected (HR-HPV+).
Data management and statistical analysis
Interview and clinical examination data were collected using standardized, scannable paper forms, generated by the Teleform™ software system (Cardiff, Sunnyvale, CA, USA).
The variables obtained from the interview were defined as continuous, categorical, or count variables. Average monthly income was reported in West African CFA and was treated as an ordinal variable, with values 1–4 corresponding to increasing income categories.
The urban and rural groups were first compared along sociodemographic and behavioral factors, using Student’s t-tests (continuous and ordinal variables), Pearson’s chi-square tests (categorical variables), Fisher’s exact tests (categorical variables with at least cell of <5 participants), or Poisson regression (counts). We then compared women who were infected (HR-HPV+) to those who were not infected (HR-HPV-), using the same tests. This analysis was then repeated within each geographic setting, with reference categories designated for categorical variables.
To test whether the effects of different variables on HR-HPV infection were different between urban and rural women, we constructed a series of logistic regression models, with HR-HPV status as the outcome variable. Each independent variable was entered into a separate model, along with geographic setting (urban or rural) and a geography×variable interaction term. The significance of the interaction term (or, for variables with more than two categories, the overall significance of the collection of interaction terms) implied how different the effect of the variable was on infection status between the urban and rural women. For categorical variables, this approach produced conclusions that were meaningfully equivalent to those produced by a Breslow-Day test for statistical interaction.
A p-value of <0.05 was considered statistically significant. All analyses were performed using STATA 11 (StataCorp LP, College Station, TX, USA).