Population and data collection
A cross-sectional study was undertaken in January and February 2007 in four Saharawi refugee camps near Tindouf in the Algerian desert. The total population was estimated at approximately 165,000 persons. The target population was limited to children (6-14 years old) and women (15-45 years old). The sample size calculation was based on an estimated goitre prevalence of 50%, an absolute precision of ± 5% and a 95% confidence interval. This corresponded to approximately 400 children and 400 women, as determined with EpiInfo Statcalc; version 6.04b .
Each of the four refugee camps was organized into six administrative zones called "dairas", and it was assumed that each daira (24 in total) had approximately the same number of inhabitants. To achieve a total sample size of 800 individuals, about 34 participants (17 children and 17 women) were included from each daira from randomly selected households. The average household had 8 individuals in the target age groups, so about four households per daira were needed to reach the required number. All eligible children and women in each selected household were included. The total study sample included 421 children and 405 women from 92 households. The response rate was 96%, and the main reason for not participating was absence on the day of visit.
Ethical approval for the study was given by the Norwegian Regional Committee for Medical Research Ethics and by the Saharawi health authorities. Informed written consent was obtained from the chief medical officers in the camps. Informed oral consent was given by the women and the parents of the children included in the study. It was emphasised that refusal to participate in the study would have no negative effects on their entitlement to food aid or other services.
Assessing thyroid volume
Tvol was assessed by two trained health workers (examiner 1 and 2): a trained nurse who completed a short course in ultrasound measurements at the Txagorritxu hospital in Vitoria-Gasteiz in Spain prior to the study and a local doctor who was trained by the nurse at the study site. Examiner 1 measured 406 individuals, and examiner 2, 418 individuals. A portable ultrasound (Sonosite Titan) equipped with a 38 mm 5-10 MHz linear transducer was used for the thyroid measurements. Subjects were examined in a supine position with extended cervical spine. Maximum perpendicular depth (anteroposterior, AP diameter) and width (mediolateral, ML diameter) were measured with electronic callipers on a transverse image of the largest diameter. The maximum lobe length (craniocaudal, CC diameter) was measured on a longitudinal image. Two images on a dual screen were used if the lobe extended beyond the 38 mm transducer measurement width. The transducer was kept perpendicular to the skin. Nodules and/or cystic areas were included in the volume determination. Thyroid volume was estimated according to the method of Brunn et al. . By regarding the two lobes without the isthmus as corrected rotation ellipsoids, the volume of the thyroid gland was calculated by adding the volume of the right and left lobe, each calculated as follows: Tvol lobe = AP diameter × ML diameter × CC diameter × 0.479. A thyroid enlargement in a child was regarded as a goitre if the thyroid volume was above the 97th percentile of the age and sex-specific international reference values for thyroid volumes in an iodine-replete population . Thyroid enlargement was defined as a Tvol exceeding 18 ml for women, which corresponds to mean + 3SD in iodine sufficient populations . Height and weight were measured by standard anthropometric techniques . Body weight was measured using a UNICEF electronic SECA 890 (Hamburg, Germany) to the nearest 0.1 kg. Height was measured to the nearest 0.1 cm using a portable stadiometer. Spot urine samples were aliquoted and stored at 5°C until analysis. Water samples were collected in every household. To determine the urinary iodine concentration (UIC) and the iodine concentration in the household drinking water, samples were analysed according to the Sandell-Kolthoff reaction .
Intra- and inter-observer agreement
To assess intra- and inter-observer agreement of the ultrasonic measurements, a methodological sub-study was conducted on 12 children (7 to 14 years old), who were selected from the study population, but were not part of the main study. Each field worker measured the 12 children twice using the same apparatus. Thus, a total of 48 Tvol measurements were taken. To minimize any influence of the fieldworkers' memory, the time between measurements of the same individual was increased by measuring all subjects once in consecutive order from 1 to 12 before repeating the procedure. Since each Tvol measurement involved the evaluation of up to six images per person (an image set) - 4 transverse images (two diameters on each lobe, right and left sides) and up to 2 longitudinal images - it is unlikely that the fieldworkers would remember the specific measurements of any individual.
Clinical evaluation of ultrasound images
A sample of 56 images (28 from each examiner) of all diameters was sent to an external ultrasound expert at the Txagorritxu hospital in Vitoria-Gasteiz, Spain, who inspected the images visually. According to the expert evaluation, the ultrasound measurements of examiner 1 were satisfactory. The AP diameters measured by examiner 2 were consistently too short, nor were they perpendicular to the ML diameter. The CC diameter measurements were also evaluated, but not commented upon in the expert evaluation.
Validation study and re-measurement of ultrasound images
Based on the study of intra- and inter-observer agreement and the clinical evaluation, the decision was made to print and re-measure the images taken by examiner 2. The images of examiner 1 had been found to be satisfactory, so examiner 1 also performed the new measurements. Only 272 of the 418 image sets for examiner 2 had been stored and could be retrieved. New measurements of the AP and ML diameters were taken. The original CC diameters were retained because the quality of the stored images was generally insufficient for taking new measurements.
Data on Tvol, UIC, and iodine in drinking water did not adhere to a Gaussian distribution. Descriptive statistics were therefore reported as the median value with percentiles (P25 and P75). and the difference in Tvol between the samples of the two examiners was tested using the Mann-Whitney test. The proportion of goitre in women and children was compared with a Chi-square test. All P values were two-sided, and a 5 percent significance level was used.
The intra- and inter-observer agreement in the measurements of the thyroid diameters and thyroid volume were presented as intraclass correlation coefficients (ICC) based on a one-way random effects model for single measurements taking the absolute agreement between the measurements into account . The ICCs express proportions of variance, and when a high proportion of the total variance is accounted for by intra- or inter-observer variability, the corresponding ICC and agreement in the measurements will be low. For comparability with other studies, the inter-observer variability was also calculated as the absolute value of the difference between the measurements of examiner 1 and examiner 2 expressed as a percentage of the mean of the two measurements for each individual . This percentage was presented as the sample mean value (%) and standard deviation (SD). Only the first measurement of each examiner was included in the calculation of inter-observer variability.
The measurements of examiner 2 were corrected statistically by regressing the re-measured values (considered to be the "true" value) for the AP and ML diameters on the original measurements using linear regression for the sub-sample of available images (n = 272). The corrected values were then predicted with the regression equations and used to re-calculate Tvol (retaining the original CC diameter, which could not be re-measured) for the 418 individuals measured by examiner 2. A regression equation for Tvol was also used to directly predict the corrected value from the original value. The agreement between the original and re-measured Tvol values and between the re-measured and statistically corrected Tvol values was summarized using ICC. SPSS version 14.0 (SPSS Inc., Chicago) was used for the analyses.