Protocol - Child Head Circumference

Several overarching, critical issues for high-quality data collection of anthropometric measures that optimize the data in gene-environment etiologic research include (1) the need for training (and retraining) of study staff in anthropometric data collection; (2) duplicate collection of measurements, especially under field conditions; (3) use of more than one person for proper collection of measurements where required; (4) accurate recording of the protocols and the measurement units of data collection; and (5) use of required and properly calibrated equipment. The PhenX Anthropometrics Working Group recommends that this protocol also be used on participants from birth to 36 months of age (see Centers for Disease Control and Prevention growth charts).

The Expert Review Panel notes that this protocol could be used for children up to 60 months of age if there are environmental concerns (e.g., malnutrition; see World Health Organization growth charts).

Notes from the Expert Review Panel: The notion of recommending replicate measurements comes from the reduction in random errors of measurement and accompanying improved measurement reliability when the mean of multiple measurements is used rather than a single measurement. This improvement in measurement reliability, however, depends on the reliability of a single measurement in the hands of the data collectors in a particular study (Himes, 1989). For example, if a measure such as recumbent length in a given study has a measurement reliability of 0.95 (expressed as an intraclass correlation coefficient), taking a second measurement and using the mean of the two measurements in analyses will improve the reliability to only 0.97, yielding only a 2% reduction in error variance for the additional effort. If, in the same study, the reliability of a single triceps skinfold measurement was 0.85, using the mean, including a replicate measurement, would raise the reliability to 0.92 and yield a 7% reduction in error variance, more than a three-fold improvement compared with recumbent length. The intraclass correlation coefficient is specifically recommended here for assessing reliability because it takes account of random and systematic errors of measurement, whereas the interclass correlation (e.g., Pearson correlation) takes account of only random errors of measurement.

Because the benefits of taking replicate measurements are so closely linked with the existing measurement reliability, it is recommended that as a part of the training of those who will be collecting anthropometry data, a reliability study be conducted that will yield measurement reliability estimates for the data collectors, protocols, settings, and participants involved in that particular study (Himes, 1989). If the measurement reliability for a single measurement is greater than or equal to 0.95, the recommendation is that replicate measurements are not necessary and will yield little practical benefit. If the measurement reliability is less than 0.95, the recommendation is to include replicate measurements as prescribed.

If replicate measurements are indicated because of relatively low reliability, a second measurement should be taken, including repositioning the participant. A third measurement should be taken if the first two measurements differ by more than 0.5 cm. If it is necessary to take a third measurement, the two closest measurements are averaged. Should the third measurement fall equally between the first two measurements, all three should be averaged.