Letter to the editor |
Limitations of a case–control study on bisphenol A (BPA) serum levels and recurrent miscarriage
Department of Community and Preventive Medicine, Box 1043, Mount Sinai School of Medicine, One Gustave L.Levy Place, New York, NY 10029, USA
E-mail: trudy.berkowitz{at}mssm.edu
Sir,
Recently, Sugiura-Ogasawara et al. (2005)
described the results of their case–control study and stated that it ‘provided the first concrete evidence that high exposure to BPA [bisphenol A] may be associated with recurrent miscarriage, especially with antinuclear antibody (ANA)-positive patients.’ As discussed in detail below, the findings from this study do not support an association.
The authors state that serum BPA levels were collected serially in patients (n = 45) and controls (n = 32). However, no data are provided on the number or timing of serial samples, or the time of sample collection in relation to miscarriage or subsequent pregnancy for patients. It is unclear if the reported BPA levels are for a single sample per individual or for all samples combined, and whether an analysis of differences in concentrations over time was conducted. The absence of these data prevents any understanding of the temporal relationship between BPA levels and pregnancy or miscarriage. BPA has a very short half-life in the body of 
5 h, so the timing of the measurement is critical (Völkel et al., 2002). It seems unlikely that serum samples were collected in a time-frame that would be relevant to any of the serial miscarriages since the patients were selected after their history of miscarriage had been established.
A statistically significant difference in mean BPA serum levels is reported for patients with recurrent miscarriages (three or more consecutive first trimester miscarriages) compared to controls. However, the median concentration is virtually identical for these two groups (0.71 ng/ml in patients and 0.705 ng/ml in controls), which indicates that there are no actual differences between these two groups. The apparent difference is likely attributable to the high BPA levels of a few individuals as shown in Figure 1. The statistical test, Welch’s test, used to measure the difference between patients and controls is inappropriate; a log-transformation of the data or a non-parametric test should have been used. Either of these approaches would more appropriately account for the potential influence of a few high sample levels.
Because of differences between the patient and control populations, they cannot be considered comparable since the patients were housewives and the controls were medical workers. Most importantly, the controls had no history of live births, infertility and miscarriage; therefore, the fertility of these women is unknown. Additionally, in the comparison of BPA levels between patients and controls, other factors known to be associated with increased miscarriage do not appear to have been controlled for as confounding factors. For example, hypothyroidism and systemic lupus erythematosus (SLE) are associated with an increased risk of miscarriage (Crombleholme, 2004; Hellmann and Stone, 2004). We know, based on a separate analysis of the patients, that eight of the 45 patients were classified as hypothyroid (assuming the table legend is correct) and 10 of the 45 patients were positive for ANA (a very sensitive marker for SLE). However, no comparable data are provided for the control population. Therefore, these conditions may have been over-represented in the patient population and be the cause of the reported difference.
The article also states that 35 of the 45 patients subsequently became pregnant and 17 (48.6%) miscarried again. The authors highlight the fact that the mean BPA levels in the women who miscarried are non-significantly higher than those who had a successful pregnancy. However, the median BPA concentration was actually higher in women with successful pregnancies, which means that at least half of the women who finally had successful pregnancies had higher BPA levels than those who miscarried again. This is inconsistent with their hypothesis that BPA levels are associated with an increased risk for miscarriage.
Finally, the enzyme-linked immunosorbent assay analytical method used to detect BPA in this study has not been validated for measurement in serum and is subject to extensive cross-reactivity. Therefore, it is not clear whether the reported concentrations reflect actual BPA levels or the levels of one or more cross-reacting compounds (Kodaira et al., 2000).
Sugiura-Ogasawara et al. (2005) note that this is a preliminary study and that the sample size was small. These facts and the lack of information on the temporality of the sampling, the lack of a difference between the patients and controls, and the failure to control for confounding factors, particularly conditions seen in the patients and associated with recurrent miscarriage, indicate that the reported association between BPA and miscarriage is not supported by the evidence provided.
References
Crombleholme WR (2004) Obstetrics. In Tierney TM, McPhee SJ and Papadakis MA (eds) Current Medical Diagnosis and Treatment, Chap 18.
Hellmann DB and Stone JH (2004) Arthritis and musculoskeletal disorders. In Tierney TM, McPhee SJ and Papadakis MA (eds) Current Medical Diagnosis and Treatment, Chap 20.
Kodaira T, Kato I, Li J, Mochizuki T, Hoshino M, Usuki Y, Oguri H and Yanaihara N (2000) Novel ELISA for the measurement of immunoreactive bisphenol A. Biomed Res 21,117–121.
Sugiura-Ogasawara M, Ozaki Y, Sonta SI, Makino T and Suzumori K (2005) Exposure to bisphenol A is associated with recurrent miscarriage. Hum Reprod 20,2325–2329.
Völkel W, Colnot T, Csanady GA, Filser JG and Dekant W (2002) Metabolism and kinetics of bisphenol A in humans at low doses following oral administration. Chem Res Toxicol 15,1281–1287.[CrossRef][Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||