Validity and Reliability
The validity (the agreement between the true value and a measurement value) of body composition is key to determine the precision of a measurement. A multi-compartment model, specifically the four-compartment model (fat mass, total body water, bone mineral mass, residual mass), is regarded as the criterion method in determining body composition, and thus this model is often used in comparison when assessing the validity of DEXA measurements . Despite this, there is a relative lack of validation studies using these comparisons, predominantly due to the time, labour, and financial cost involved .
The validity of DEXA scans when compared to the four-compartment model is currently equivocal. A study by Arngrimsson et al., (2000) in male and female distance runners reported that DEXA underestimated body fat measurements by around 2% . This finding was replicated in a study by Van der Ploeg et al., (2003) in a healthy adult population , and slightly less than the 4% difference in trained men found by Withers et al . Fat-free mass was also reported to be overestimated by approximately 2.5kg .
Conversely, other research has shown an overestimation of body fat by 3-4%, and an underestimation of fat-free mass by up to 3kg [3, 11]. The differences in the reported validity of DEXA scans for body composition may partly be due to differences in software, manufacturer, and the wide range of body composition values across various cohorts. It is worth noting that although current research is conflicting, it is commonly assumed that DEXA is a valid technique for the assessment of body composition.
It is also important to consider the validity of DEXA results not only for cross-sectional data (observational data at one given time point), but also when assessing changes in body composition over time, as this is often of interest to the sport science practitioner. Early research in this area used the novel method of scanning participants, before placing lard samples on the participants to simulate a gain in adipose tissue. One study showed that fat mass and lean body mass measurements were not significantly different from expected change following the placement of 8.8kg of lard on healthy female subjects . Another study also reported measurements in line with expectations when 11.1kg was added, however, there was evidence that fat mass was underestimated by 1.9% when the amount of lard added was increased to 22.2kg .
Data on DEXA use to assess body composition in athletic populations is scarce. Van Marken et al.,  showed no difference in mean body composition change between the four-compartment model and DEXA scans in male bodybuilders. This finding was replicated by Santos et al.,  in male judo competitors. Despite these findings, both studies reported that individual error rates of 4% for body fat measurement, highlighting the large individual differences within measurements.
The reliability (the reproducibility of the observed value when the measurement is repeated) of DEXA scanning is important for determining the precision of a single measurement, as well as the ability to detect change over multiple measurements. Despite the various benefits of DEXA scanning for body composition assessment, it is not without error. Inconsistent results can occur when using scanners from different manufacturers , as well as two scanners produced by the same manufacturer . Software updates may also alter body composition results due to algorithm variation . Differences in the x-ray beam used can also contribute to error; pencil beams and fan beam DEXA systems have been shown to provide significantly different body composition test results, limiting comparison between the two .
Biological variation may also affect reliability in DEXA measurements, particularly in athletic populations. The effects of fluid intake on body composition has been studied previously, showing that fluid intake of 0.8-2.4L of water can significantly increase lean mass estimates in the trunk region . Similar results were reported by Thomsen et al., , showing that estimates of lean body mass were increased by over 1kg per hour following a standard meal weighing 1311g, as well as after drinking 1L of water.
Changes in any of these variables are not uncommon in athletic populations and may mask or give the false impression of changes in skeletal muscle mass. Bone et al.,  reported that glycogen and creatine loading increased lean mass values by 2.1% and 1.3%, respectively. Variation in hydration status of 5% has also been shown to alter DEXA predictions of body fat by 3% . As manipulation of these variables is commonplace with athletes, standardisation is key to accurately detect meaningful change.