Study design and setting

This was a cross-sectional study of 6-month duration (February to July 2014), conducted in the health district Dschang in the Western Region of Cameroon (Fig. 1).

Fig. 1. Dschang district in the Western Region of Cameroon

Study participants

Sources of participants

According to the data from the regional delegation of health for the Western Region of Cameroon, Dschang health district is the largest health district in the region, with an estimated population of 309,285 inhabitants in 2012, distributed across 22 health areas (19 rurals and 3 urbans) (2012 annual activity report of the regional delegation of health for Western Cameroon). Dschang is home to the largest university in the Region and therefore has a cosmopolitan population reflecting the ethnic diversity of the country. The adult population in urban areas comprises students, traders, civil servants and middle income earners from private sectors while farmers are predominant in rural area. This study was approved by the Cameroon National Ethics Committee, and all participants provided a written informed consent before enrolment.

Eligibility criteria

Eligible participants were adults aged 20 years and above who had been living in the study setting for more than three months. We excluded individuals with serious mental or physical (limb amputation or paralysis) disability, pregnant or breastfeeding women and participants with simultaneous leucocyturia and urine nitrites.

Selection of participants

We used a multi-level cluster sampling including the health area (first level), the village (second level), the neighbourhood (third level) and the household (fourth level). The sequence below was followed to select the clusters and corresponding health areas. 1) We first assumed the number of clusters needed to be 30; 2) We then determined the sampling interval (SI) which corresponded to 10310, by dividing the population of Dschang health district by the number of clusters; 3) We determined the first cluster or random number (RN) by selecting the four last numbers of a randomly selected bank note which corresponded to 1399; 4) We next estimated the cluster number size (C _(n) ) from the formula C _(n) ?=?RN+(n-1)*SI where n is the cluster number; 5) The various health areas (with their corresponding population size) were then sorted in alphabetic order and progressive cumulative population size estimated (see Additional file 1: Table S1); 6) The last step consisted of selecting the health areas. For a health area to be selected, the size of the corresponding cluster number had to be less than the health area population. This action was repeated until the size of the cluster number became superior to the health area population; we then moved to the following health area in the table. The selected health areas and corresponding clusters are presented in Additional file 1: Table S1.

In a selected health area, one village/neighbourhood was randomly drawn when there was more than one regardless of the population size. The starting point was randomly selected from the market, church, health centre or school in the village/neighbourhood. Thereafter, we randomly selected the direction while the side of road was chosen with the aid of coin toss. We entered consecutively in the households where we randomly selected per household a maximum of two adults aged 20 years and above among those who had been living in the household for more than three months. For each household declining to participate, the next household was selected until the total number required for the cluster size was reached. The cluster size ranged between 14 and 15 subjects. In health areas with many clusters, the corresponding villages and/or neighbourhoods were randomly selected as previously described.

Variables of interest

The main outcome of interest in this study was CKD defined by the persistence after 3 months of albuminuria (Albumin/Creatinine ratio???30 mg/g) and/or decreased estimated glomerular filtration rate (eGFR) (<60 ml/min/1.73 m ² ) according to the K/DIGO guidelines [13]. Exposure variables included demographics (age and gender), self-reported existing conditions (hypertension, diabetes and gout), any hypertension (self-reported or screen-detected [i.e. systolic (or diastolic) blood pressure ?140 (90)], any diabetes (self-reported or screen-detected [fasting capillary glucose ?126 mg/dl)], lifestyles (alcohol consumption and smoking), use of nephrotoxins [street medications (western drugs, usually of uncertain origins that are sold in shops and regularly along market streets, instead of pharmacies, and without any control) and herbal medicines], overweight or obesity (body mass index [BMI]???25 kg/m ² ) and blood pressure levels. Potential effect modifier was residency (urban vs. rural).

Data sources and measurement

Data were collected during household surveys by final year’s undergraduate medical students. Demographics, history of existing conditions, lifestyles and data on the use of nephrotoxins were collected during face-to-face interviews with participants. Blood pressure was measured according to the World Health Organization (WHO) guidelines [14] using an automated sphygmomanometer (OMRON HEM705CP, Omron Matsusaka Co, Matsusaka City, Mie-Ken, Japan) on the right arm with participants in a sitting position after 30 min of rest with a cuff size of 23 x 12 cm or larger for obese individuals. Body weight and height were measured three times and their average used in all analyses. For each participant, 3 ml of whole blood was collected from an antecubital vein for serum creatinine and fasting glycemia (after an overnight fast of at least 8 h), and mid-stream second morning urine collected for dipstick, creatinine and albumin tests. Fasting glycemia and dipstick tests were done immediately after sample collection. The remaining sample was transported in ice to the Biochemistry Laboratory of the Yaounde University Teaching Hospital for further processing. Urine dipstick tests used the CombiScreen 7SL PLUS 7 test strips (Analyticon Biotechnologies AG, D-35104 Lichentenfeis, Germany). Fasting glycemia was performed using One Touch Ultra® easy reader® (LifeScan Europe, Cilag GmbH International, Zug, Switzerland). Serum and urinary creatinine were measured with a kinetic modification of the Jaffé reaction using Human visual spectrophotometer (Human Gesellschaft, Biochemica und Diagnostica mbH, Wiesbaden, Germany) and Beckman creatinine analyzer (Beckman CX systems instruments, Anaheim, CA, USA) while urinary albumin was measured using pyrogallol red-molybdate complex with Teco diagnostics tests (Teco Diagnostics, Anaheim, CA, USA). For any participant with positive dipstick [proteine (? trace), blood, leucocytes), albumin/creatinine ratio (ACR) ?30 mg/g and fasting glycemia of at least 126 mg/dl (for unknown diabetes), another test was performed 2 to 3 weeks after to confirm the results. In participants with estimated glomerular filtration rate (eGFR)?<?60 ml/min/1.73 m ² according to the MDRD formula and/or urinary albumin/creatinine ratio (ACR) ?30 mg/g, the chronicity was confirmed on another sample 3 months later.

Definitions and calculations

Estimated glomerular filtration rate (eGFR, ml/min) corresponded to creatinine clearance. For the main analyses, eGFR was based on the four-variable MDRD (Modification of Diet in Renal Disease) study equation; however, for comparison purpose of the baseline estimates, we also derived the eGFR from the Cockcroft–Gault (CG) formula and the CKD-EPI (Chronic Kidney Disease Epidemiology Collaboration) equations [15]–[17]. 24-h albuminuria was estimated from Albumin/Creatinine ratio (mg/g). BMI was estimated as weight (kg)/height (m)*height (m).

Study size

By considering a 10 % prevalence (P) of CKD in adults [1], a precision (I) of 2 %, a correction factor (K) of 2 for the cluster effect, a 95 % confidence interval, the minimal sample size (N) required was 432 subjects using the following formula N?=?[(Z? /2) ² PQ/ I ² ] x K.

Handling of quantitative variables

Age was treated as continuous variable in all analysis while blood pressure, while other quantitative variables were treated both as continuous and categorical variables, based on clinically meaningful stratification. Hypertension was defined as a systolic (SBP) ?140 mmHg and/or a diastolic blood pressure (DBP) ?90 mmHg or use of blood pressure lowering medications. Diabetes mellitus was defined as repeated fasting glycemia???126 mg/dl or use of glucose control agents. A BMI?>?25 kg/m ² was used to define overweight and obesity. CKD was classified based on GFR and albuminuria categories. GFR categories of CKD included: G1 (eGFR???90); G2 (eGFR 60–89); G3a (eGFR 45–59); G3b (eGFR 30–44); G4 (eGFR 15–29) and G5 (eGFR?<?15). Albuminuria categories of CKD were: A1 (<30 mg/g); A2 (30–300 mg/g) and A3 (>300 mg/g). The following formula was used to convert serum creatinine from Jaffe reaction (SCr _Jaffe ) to standardized serum creatinine (SCr _Standardized ) for use in MDRD and CKD-EPI formulas: SCr _Standardized ?=?0.95*SCr _Jaffe – 0.10 [18].

Statistical analysis

Data analysis used SAS/STAT v9.1 software and the survey analysis procedures (‘proc surveymeans’, ‘proc surveyreg’ and ‘proc surveylogistic’) to account for the multilevel sampling design of the study. We have reported the results as means, counts and percentages and the accompanying 95 % confidence intervals. The sampling error was estimated with the use of the Taylor expansion method. Age and sex adjusted logistic regression models were used to investigate the predictors of CKD, CKD stages G3-G4 and albuminuria. A p-value <0.05 was used to indicate statistically significant results. For the main analyses, prevalence and determinants of CKD are based on MDRD derived eGFR. In secondary analyses however, we have also estimated GFR and staged kidney function using the Cockroft-Gault and CKD-EPI equations.

...