Little is known about predictors of decline in vitamin D status (vitamin D decline) over time. We aimed to determine demographic and lifestyle variables associated with vitamin D decline by sufficiently controlling for seasonal effects of vitamin D uptake in a middle-aged to elderly population. Using a longitudinal study design within the larger framework of the Murakami Cohort Study, we examined 1044 individuals aged between 40 and 74 years, who provided blood samples at baseline and at 5-year follow-up, the latter of which were taken on a date near the baseline examination (±14 d). Blood 25-hydroxyvitamin D (25(OH)D) concentrations were determined with the Liaison® 25OH Vitamin D Total Assay. A self-administered questionnaire collected demographic, body size and lifestyle information. Vitamin D decline was defined as the lowest tertile of 5-year changes in blood 25(OH)D (Δ25(OH)D) concentration (<6·7 nmol/l). Proportions of those with vitamin D decline were 182/438 (41·6 %) in men and 166/606 (27·4 %) in women (P < 0·0001). In men, risk of vitamin D decline was significantly lower in those with an outdoor occupation (P = 0·0099) and those with the highest quartile of metabolic equivalent score (OR 0·34; 95 % CI 0·14, 0·83), and higher in those with 'university or higher' levels of education (OR 2·92; 95 % CI 1·04, 8·19). In women, risk of vitamin D decline tended to be lower with higher levels of vitamin D intake (Pfor trend = 0·0651) and green tea consumption (Pfor trend = 0·0025). Predictors of vitamin D decline differ by sex, suggesting that a sex-dependent intervention may help to maintain long-term vitamin D levels.