The Nakuru eye disease cohort study: methodology & rationale

The sample size of 5000 participants required for the baseline survey was calculated based on an expected prevalence of VA < 6/12 in the better eye due to posterior segment eye diseases (PSED) of 3.0% among those aged ≥50 years, a required precision of 0.5% (i.e. a 95% confidence interval [CI] of 2.5%-3.5%), a design effect of 1.5, and a response rate of 90%. (Epi Info 6.04, Centers for Disease Control and Prevention, Atlanta, GA). We selected 100 clusters each of 50 participants.

Recent census data for Kenya were not available [32], and therefore electoral role lists that were renewed in 2006 in preparation for the 2007 general elections were used as the sampling frame for this baseline survey. The population size was updated for the year 2007 using a population growth rate of 2.7% per year [33]. One hundred clusters were selected with a probability proportional to the size of the population (Figure 1). A cluster was defined as the area served by the polling station.

Households were selected within clusters using a modified compact segment sampling method [34]. Each cluster was divided into segments so that each segment included approximately 50 people aged ≥ 50 years. For instance, if a cluster included 200 people aged ≥ 50 years then it was divided into four segments. One of the segments was chosen at random by drawing lots and all households in the segment were sequentially sampled, until 50 people aged ≥ 50 years were identified. An eligible individual was defined as someone aged ≥ 50 years living in the household for at least three months in the previous year. Age was determined using the subject’s testimony, national identity cards and a calendar of historic events. If the segment did not include 50 people aged ≥ 50 years then another segment was chosen at random and sampling continued until 50 were reached. If after enumerating individual number 49 the next household had more than one person aged ≥50 all were enumerated and invited for examination.

In total, 4381 participants underwent complete (ophthalmic and general) examination at baseline across 100 clusters. The prevalence of blindness was 1.6% (95% CI: 1.2-2.1%) [6].

A pilot follow up retraced 438 participants from 10 of the 100 clusters in 2008, a mean of 1.5 years from baseline, and 408 (79%) were successfully retraced to give an estimated 4.2% loss per year.

Approximately one week before the follow-up examination clinic was planned for a given cluster, a field officer studied the maps of the village and made phone contact with the village chief or guide to arrange the visit. At the planning visit a list of study participants were given to the chief and a local village guide was recruited to assist location of the study participants. At this visit the examination site was established and identification of amenities such as electricity, water and road access were made. Two days prior to the clinic, the field officer reminded chiefs of the visit by phone and notified them and the guide of the advance team’s arrival.

On the day prior to the examination clinic, the Advance Team visited homes of baseline participants and confirmed their identity using National Identity cards. The two advance teams comprised of one nurse, one field officer and a driver or public transport. During this visit they performed the following duties:

On the examination day, the advance team confirmed the identity of participants against their records from the previous day and against data from baseline (age, date of birth, name, and identity cards). Each participant was given a questionnaire, which was completed by the examiners as they move from station to station.

Examinations were performed as per baseline unless otherwise indicated in Table 2.

Details of each examination station are provided below including differences, if there were any, between baseline and follow-up.

A nurse performed and recorded measures of participants: height; weight; waist and hip circumference, and three measures of blood pressure, each 5 minutes apart. In addition, at follow-up, bioimpedence (Tanita Segmental Body Composition Monitor) was performed.

At baseline, capillary blood was taken from all participants for random blood glucose and cholesterol. At follow-up, no blood for cholesterol was collected and in addition, subjects with a random blood sugar greater than 11.1 mmol/L (IDF guidance at time of baseline study), those with known diabetes (regardless of random measure), evidence of diabetic retinopathy on retinal imaging and a subset (chosen randomly within each cluster) with random glucose between 7-11 mmol/L had an additional capillary blood HbA1c (A1C Now+, Bayer).

An interviewer performed the structured interview in the participant’s preferred language covering i) demographic details including; name, year of birth, ethnicity and education level; ii) past medical and ocular history including medical or ophthalmic medication or surgery and relevant family history; iii) relevant risk factors including; smoking and tobacco consumption and alcohol intake; iv) socioeconomic status based on job, housing conditions, ownership of material goods and livestock which is translated in to a score based on previous work in the same population [35]. (See Additional file 1: Appendix for Questionnaire/Data Entry Booklet).

A clinical officer determined whether the study participant:

Visual acuity (VA) was measured using a back-illuminated modified LogMAR reduced tumbling E chart [36, 37], which has been used in previous population based studies [38, 39].

All subjects, regardless of VA underwent autorefraction using the Topcon® Auto refractor RM8800 at baseline and the hand held SureSight autorefractor (Welch Allyn) at follow-up, following manufacturers guidelines. Any subject recording an acuity of ≤24 optotypes, <6/9 equivalent (with or without glasses) underwent best corrected visual acuity. The refraction measure recorded for each eye was mounted in the trial frames using trial lenses (rounded up or down to the nearest 0.25 diopters). Visual acuity was then re-measured to give an estimate of the “corrected visual acuity” in each eye individually. When autorefraction results were not available, the pinhole method was used to estimate corrected visual acuity. A subset of participants also underwent manual refraction by a visiting optometrist for five clusters to validate the accuracy of the autorefractor.

Participants performed the Swedish Interactive Thresholding Algorithm (SITA) STANDARD 24–2. SITA Fast was used to determine the threshold level in all participants having visual field analysis. Appropriate corrective lenses for refractive errors were used when needed. An automated fixation monitor was used throughout. If the SITA fast test was reliable (following manufacturers guidelines) the SITA standard test was performed. If the SITA fast was unreliable then this was repeated once. If it remained unreliable then no further testing was done.

At follow-up, a different strategy for visual field testing was used: All subjects with VA equivalent to > =6/60 Snellen underwent automated visual field testing by a trained visual field technician using the Henson 8000 Visual Field Analyser (TopCon, Inc.) The multiple stimulus suprathreshold test was used following manufacturers guidelines (Screening test - 26 test locations). When one or more spots were missed, the 26-point test was repeated for that eye. If any missed spots re-occurred on the second time of testing the test for that eye was extended to 68 test locations. This machine and strategy were used in preference to the baseline methods due to feedback from both patient’s and tester at baseline. Patient’s found the baseline testing protocol difficult to understand and the time required to complete the test meant only a sub-sample of the population could be investigated.

Undilated (anterior segment) and dilated (posterior segment) slit lamp biomicroscopy examination were performed on all participants by the study ophthalmologists (WM at baseline, AB at follow-up) using a Haag-Streit BD 900 Slit Lamp (BM 900 at follow-up) and Volk condensing lenses (90D at baseline, Superfield and 66D at follow-up).

Pharmacologic dilation of the subject’s pupils was achieved by using tropicamide 1% (Mydriacyl) with phenylephrine hydrochloride 2.5% if needed. Dilation was not performed in subjects deemed at risk of narrow angle closure (inability to visualise at least 180° of posterior pigmented trabecular meshwork on non-indentation gonioscopy [42]). At risk subjects were referred to the Nakuru Eye Unit for prophylactic laser peripheral iridotomies.

Note: The gold-standard for grading of DR, AMD and optic disc changes is based on retinal photographs and not clinical assessment. Clinical examination was performed as a backup to equipment failure and a comparison of clinical and image based grading can be compared in the analysis stage and factored in the scenario by which a number of participants only have clinical grading available.

A patient record was completed for each participant and crosschecked for errors by the project field coordinator. Patient records were scanned to create a digital backup and then data were entered into an EpiData database (with built in range and consistency checks) independently by two data clerks and validated by the study ophthalmologist to reconcile any differences. Further data cleaning and all statistical analyses were conducted using STATA 10.0 (StataCorp LP, Texas, USA).

The visual fields PDF print outs and raw data were sent securely to Moorfields Eye Hospital for grading along with the fundus and anterior segment images at baseline and follow-up. All image and visual field data were backed up on local devices and external hard drives. All images were first examined for quality and categorized as excellent, good, fair, borderline and ungradeable. If the images were ungradeable the clinical diagnosis was used. For gradable images the retina and optic disc were reviewed, and a diagnosis made based on the appearance of the image e.g. diabetic retinopathy, toxoplasmosis, onchocerciasis, age related macular degeneration, myopic fundus, glaucoma, optic atrophy or other retinal pathology. VCDR was measured and all images were graded for the absence/presence and stage of DR and ARMD. The graders graded the images for as many disease categories as possible, and so if it was feasible to grade an image for optic disc abnormality but not for ARMD, then the grader completed the optic disc grading only. A senior grader verified a random 10% of images that were graded as normal as well as all abnormal images to ensure quality assurance. The graders re-graded a random selection of images with a minimum of 14-days interval to allow for intragrader reliability to be established.

Definitions used for analysis are detailed in Table 3.

Participants who were examined at baseline and eligible for follow-up assessment but who did not attend the examination clinic were contacted to determine the reason for their absence. Participants who were not locatable for the examination were categorized as non-responders and their reason for absence determined through available phone contact, neighbors and village guides as, “deceased”, “moved away”, or “unknown”.

All participants identified with treatable disease in the study were offered appropriate care including free surgery and transport to the Rift Valley General Provinical Hospital or St Mary’s mission hospital, Elementita. Specific cases requiring other services were referred to the Kikuyu Eye Unit. A trained ophthalmic nurse or Ophthalmic Clinical Officer (OCO) discussed the diagnosis and the treatment options available to subjects diagnosed with untreatable eye disease. As well as study participants, non-study attendees were examined and treated by the study team.

The study adhered to the tenets of the Declaration of Helsinki and was approved by the Ethics Committee of London School of Hygiene and Tropical Medicine at both baseline and follow-up. Baseline approval was provided by the Kenya Medical Research Institute and the African Medical and Research Foundation (AMREF), Kenya at follow-up. At both phases approval was also granted by the Rift Valley Provincial Medical Officer and the Nakuru district Medical Officer of Health. Approval was sought from the administrative heads in each cluster, usually the village chief. They were also given a copy of the consent form to read and pass on to those in the village.

Informed consent was obtained from all participants. The objectives of the survey and the examination process were explained to those eligible in the local dialect, in the presence of a witness. A subject was examined only after informed consent was obtained. All participants gave written (or thumbprint) consent to participate.