Background: Accurate estimation of food portion size among free-living persons during dietary assessment is a common challenge. The weighed food intake presently regarded as the gold standard method is time and resource intensive with a high respondent burden. Retrospective methods such as the food frequency and 24-hour dietary recall where portion size estimation aids are employed in lieu of weighing are thus more frequently used. Among the portion size estimation aids, there is accumulated evidence that photographic food atlases are highly convenient and effective. So far, the only one based on the Ghanaian diet covers carbohydrate foods. Aim: To develop and validate a photographic atlas for portion size estimation of commonly consumed animal and plant protein source foods in the Ghanaian diet. Method: Fifty-four commonly consumed animal and plant protein source foods were identified in a cross-sectional survey among 834 adult residents (age £18 years) in Greater Accra Metropolitan Area. Colour photographs of each food in graduated portions of between 1 and 8 were produced and bound. Following a pilot study (n=50), the atlas was validated by the visual perception method in which 127 participants from the cross-sectional study estimated 47 predetermined portions of 12 test foods by comparing them with portions shown in photo series from the atlas. Participants’ estimations were categorized as correct (if the exact/correct picture of portion displayed was selected), adjacent (if picture immediately before or after the correct picture was selected) or distal (any other selections). Correct and adjacent estimations were classified as acceptable. Differences in proportions of accurate (correct) and acceptable (correct and adjacent) estimations across sex, age group, BMI and educational status were examined by Pearson’s chi square test of independence at p£0.05. Results: A total of 5894 references to 54 commonly consumed protein source foods were found in the 3-day 24HDR assessments. Fish and sea foods were most frequently consumed (34.6% of total protein food consumption occasions) followed by poultry (21.3%), legumes (18.8%), meat and meat products (13.6%), milk and milk products (11.7%). The atlas produced comprises 62 sets of serial, guide and range photos. In the validation, participants with tertiary education made a significantly higher proportion of accurate estimations than those with no formal education, primary or secondary school education. The proportion of accurate and acceptable estimations also differed according to the size of portion evaluated. Small portions were generally better estimated than large portions. No evidence of statistically significant differences in the proportions of accurate or acceptable estimations was found across sex, age group or BMI. Participants found amorphous foods challenging to estimate. In this regard, evaporated milk was the most difficult: nearly half (48.4%) of its portion size estimations were distal. The proportion of accurate (exact/correct) estimations observed for the 12 test foods ranged from 19.3% (fried tilapia fingerlings) to 88% (grilled sausage) while acceptable estimations ranged from 51% (evaporated milk) to 95.8% (boiled egg). Conclusion: In this study, a high proportion (i.e. 80%) of the 5969 portion size evaluations made during validation corresponded to either the correct portion or the one directly adjacent to it. Based on these results, the atlas produced is a potentially useful tool for portion size estimation of animal and plant protein source foods in the Ghanaian diet.

show more