The article does not state the variance of this IQ test and does not state how many subgroups they analyzed and whether they corrected for that or if they a-priory were planning on analyzing gender.
Picking subgroups until you've found one that is statistically significant is a well known technique in P-hacking.
Other problems:
- The IQ score for girls is 5 points higher on average than for boys in the study group
- Fluoridation has the opposite effect for girls, the average IQ was 1.5 points higher. It is really strange for a chemical to have a completely different effect on young girls and boys. Statistical variance is much more likely
- The average IQ score in the sample was 107, well above the average 100
- Parent IQ was not a variable, only the very roughly correlated high school diploma or not and household income.
"Extraordinary claim require extraordinary evidence", and this study seems to at best be an indication that further research is needed rather than iron-clad proof of this effect.