Effect of delivery season on subsequent birth interval in early 20th century in Japan

Questionnaires of birth dates of family members (13,404 families in total) were analyzed in order to examine the effects of delivery season of a baby on the subsequent birth interval. Deliveries at maternal age of 20-34 years were used. In 1921-1935, the mothers who had been delivered of a baby in August-October showed the shortest (30.62 months geometric mean) and those in February-April the longest (34.05 months) non-last intervals, with a highly significant difference among the four delivery seasons (P less than 0.001, Kruskal-Wallis test, n = 5678). Although the intervals were abruptly prolonged just before the last birth, the above difference was also consistent in the last intervals. When seasonal distributions of last and non-last births were compared, last births tended to be concentrated in the summer half of a year (P less than 0.05) in 1921-1935. In 1951-1965, overall geometric mean of the interval shortened to 28.44 months, and the length of intervals did not differ appreciably according to the season of preceding delivery. Deliveries in late summer (August-October) in 1921-1935, therefore, were associated with increased risk of termination of reproduction, on one hand, but a lowered chance of prolongation of the subsequent interval, on the other hand. Possible environmental factors are discussed to explain this apparently paradoxical phenomenon.