Short term increase in low birth weight babies after Fukushima
- PDF / 398,979 Bytes
- 2 Pages / 595.276 x 790.866 pts Page_size
- 86 Downloads / 199 Views
LETTER TO THE EDITOR
Open Access
Short term increase in low birth weight babies after Fukushima Alfred Körblein Abstract An analysis of low birth weight (LBW) births in ten contaminated prefectures of Japan, 1995–2018, finds a statistically significant increase in the LBW proportion in 2012–2013, but no increase after 2013. In the rest of Japan (37 prefectures), no increase in LBW births was found after the Fukushima accident. Keywords: Fukushima, Low birth weight, Logistic regression, Trend analysis In their recent article, Scherb et al. [1] investigated annual data of live births with low birth weight (LBW), 1995–2018, in selected regions of Japan before and after the Fukushima nuclear accident in March 2011. The trend of LBWproportion was analyzed employing logistic regression with a fourth-degree polynomial for the temporal trend, and allowing for a level shift from 2012 onward (variable cp, defined as cp = 1 in 2012–2018 and cp = 0 otherwise). Results were presented for Japan as a whole and three sub-areas: (1) five highly contaminated prefectures (Fukushima, Miyagi, Ibaraki, Tochigi, and Iwate), (2) five moderately contaminated prefectures (Yamagata, Saitama, Tokyo, Kanagawa, and Chiba), and (3) the rest of Japan (37 prefectures). In Japan as a whole, the level shift was statistically significant (OR = 1.020, p-value 0.025). The effect was greater in area 1 (OR = 1.055, p-value 0.010) than in area 2 (OR = 0.021, p-value 0.011), and not statistically significant in area 3 (OR = 1.015, p-value 0.105). Since the numbers of live births and LBW births for all three areas were provided in Table 1 of [1], I was able to check the results. In my analysis I used logistic regression with program R (https://www.r-project.org/); the option family = quasibinomial was applied to adjust the variances for overdispersion. This means that F-tests were used instead of chi-square tests.
Correspondence: [email protected] Nürnberg, Germany
With a fourth-degree polynomial for the time trend as applied in [1], the estimate for the odds ratio (OR) of the level shift in the data from Japan as a whole was OR = 1.018, p-value 0.074. A fifth-degree polynomial for the temporal trend, however, fitted the data better (deviance = 74.1, df = 17) than a fourth-degree polynomial (deviance = 90.8, df = 18). Now, the odds ratio for the level shift decreased to OR = 1.009 (95% CI: 0.991, 1.028), p-value 0.347. Similarly, the regressions of the data for the three sub-areas yielded non-significant level shifts when a fifthdegree polynomial was applied: area 1: OR = 1.039 (0.988, 1.092), p-value 0.152; area 2: OR = 1.013 (0.992, 1.035), pvalue 0.244; area 3: OR = 1.004 (0.982, 1.026), p-value 0.730). For all three data sets, a fifth-degree polynomial yielded a better fit than a fourth-degree polynomial. To increase the statistical power, I pooled the data from areas 1 and 2 and termed them the study region; area 3 was used as the control region. The odds ratio of the level shift in the study region was 1.019 (0.994, 1.044), p-value 0.152. No
Data Loading...
