There is significant discussion of this issue in Fluoride in Drinking Water: Scientific Review of EPA's Standards (2006), particularly in chapter 8, Effects on the Endocrine System. This book is a publication of the US National Research Council.
In humans, changes in melatonin are associated with the status of the reproductive system—onset of puberty, stage of puberty, menstrual cyclicity, menopause (Reiter 1998; Salti et al. 2000)—but the functional relationships are not fully understood. The elevated melatonin concentrations characteristic of prepubertal age suggest an inhibitory effect on pubertal development (Aleandri et al. 1997; Salti et al. 2000); sexual maturation begins when serum melatonin starts to decrease (Aleandri et al. 1997; Reiter 1998).
....
Sexual maturation in females occurred earlier in the high-fluoride animals (Luke 1997); males had increases in melatonin production relative to body weight between 11.5 and 16 weeks (when a decrease normally would occur), and testicular weight at 16 weeks (but not at 9 or 28 weeks) was significantly lower in high-fluoride than in low-fluoride animals. The circadian rhythm of melatonin production was altered in the high-fluoride animals at 11.5 weeks but not at 16 weeks. In high-fluoride females at 11.5 weeks, the nocturnal peak (relative to body weight) occurred earlier than in the low-fluoride animals; also, the peak value was lower (but not significantly lower) in the high-fluoride animals. In males, a substantial reduction (P < 0.00001) in the nocturnal peak (relative to body weight) was observed in the high-fluoride animals.
Although no studies are available that specifically address the effect of fluoride exposure on pineal function or melatonin production in humans, two studies have examined the age of onset of menstruation (age of menarche) in girls in fluoridated areas (Schlesinger et al. 1956; Farkas et al. 1983; for details, see Appendix E, Table E-15); [footnote 12] the earlier study was discussed by Luke (1997) as part of the basis for her research. No comparable information on sexual maturation in boys is available.
In girls examined approximately 10 years after the onset of fluoridation (1.2 mg/L, in 1945) in Newburgh, New York, the average age [footnote 13] at menarche was 12 years, versus 12 years 5 months among girls in unfluoridated Kingston (Schlesinger et al. 1956).[footnote 14] The authors stated that this difference was not statistically significant. Note that those girls who reached menarche during the time period of the study had not been exposed to fluoride over their entire lives, and some had been exposed perhaps for only a few years before menarche (they would have been 8-9 years old at the time fluoridation was started). Those girls in Newburgh who had been exposed to fluoridated water since birth (or before birth) had not yet reached menarche by the time of the study.
A later study in Hungary (Farkas et al. 1983) reported no difference in the menarcheal age of girls in a town with “optimal” fluoride concentration (1.09 mg/L in Kunszentmárton, median menarcheal age 12.779 years) and a similar control town (0.17 mg/L in Kiskunmajsa; median menarcheal age 12.79 years). This study shows postmenarcheal girls present at younger ages in the higher fluoride town than in the low-fluoride town, although the reported median ages were the same (Farkas et al. 1983).
Text of footnote 12:
Both Schlesinger et al. (1956) and Farkas et al. (1983) referred to tables of the distribution of ages at the time of first menstruation, but, in fact, both studies provided only frequencies by age (presumably at the time of study, in either 1-year or 0.5-year increments) of girls having achieved menarche by the stated age. Farkas et al. (1983) specifically indicated use of the probit method for ascertainment of the median age at menarche; the data provided by Schlesinger et al. (1956) appear to correspond to that method, but they do not specifically mention it. The probit (or status quo) method appears to be routinely used to estimate the median (or other percentiles of) age at menarche, sometimes in conjunction with an estimated mean age at menarche based on recall data (e.g., Wu et al. 2002; Anderson et al. 2003; Chumlea et al. 2003; Padez and Rocha 2003). According to Grumbach and Styne (2002), “The method of ascertainment of the age of menarche is of importance. Contemporaneous recordings are performed with the probit method of asking, ‘yes’ or ‘no,’ are you menstruating? These may be incorrect because of social pressures of the culture and socioeconomic group considered. Recalled ages of menarche are used in other studies and considered to be accurate within 1 year (in 90% of cases) during the teenage years and in older women, too.”
text of footnote 13:
Probably the median age, although the text simply says “average.” Similar studies appear to use the term “average age at menarche” to refer to the “estimated median age at menarche” (Anderson et al. 2003).
text of footnote 14:
For comparison purposes, estimates of mean or median age at menarche for the white population in the United States include 12.80 years for 1963-1970 (Anderson et al. 2003) and 12.55-12.7 years for 1988-1994 (Wu et al. 2002; Anderson et al. 2003; Chumlea et al. 2003).