Weight Talk: Fitness, Health and Society

To expand on this, the original statistical basis for "normal" BMI is deeply flawed:

Along with BMI’s fundamental limitation - its inability to accurately measure body fatness - other limitations existed within Keys’ study. The ages of the cohorts of men were not uniform. The “Bantu” cohort consisted of men aged 31-60 years, the “University of Minnesota students” cohort consisted of men aged 18-24 years, the “Minnesota executives” cohort consisted of men aged 49-59 years, and all other cohorts consisted of men aged 40-59 years. No women were included in the studies and these studies were not properly representative of different ethnicities, races, and backgrounds. Of the 7,426 men included, just 1.56% of them were of South African descent, from the “Bantu” cohort; 13.9% were of Asian descent from the “Japanese farmers” and “Japanese fishermen” cohorts; 43.7% of men were of European descent from “East Finland,” “West Finland,” “Crevalcore, Italy,” and “Montegiorgio, Italy”; 40.78% of the participants were men from the United States from the “Rome Railroad,” “University of Minnesota students,” “Minnesota executives,” “U.S. Ry. sedentary,” and “U.S. Ry., switchmen” cohorts [2]. While there was some representation of different ethnic groups in the 1972 Keys' study, this representation was somewhat random and was greatly outweighed by that of Caucasian men of European descent.

The History and Faults of the Body Mass Index and Where to Look Next: A Literature Review

The statistical averages derived from the foundational study of BMI really only have any meaning if you are a male with European ancestry - no other group was properly represented.