Background The aim of the present study is to evaluate the utility of extraction versus non-extraction-based commercial melatonin ELISA kits for determining the melatonin concentration in amniotic fluid obtained in early and late pregnancy. Methods Pregnancy duration less than 28 weeks was defined as early and from 28 weeks until delivery as late gestation. Nine samples were obtained in early and 18 in late pregnancy. Two commercially available melatonin ELISA kits (melatonin ELISA RE54021, including methanol-based extraction and direct saliva melatonin ELISA RE 54041, not including an extraction step, both from IBL-International, Germany) were used to determine melatonin concentrations in amniotic fluid. Results The mean melatonin concentration in ELISAs assayed by the non-extraction was significantly lower than those assayed after extraction. Subgroup analysis showed that there was no significant difference between melatonin concentration measured by non-extraction versus extraction ELISA in early pregnancy (11.2 ± 7.4 vs. 12.2 ± 7.7, respectively, P = 0.463) but that the mean melatonin concentration in late pregnancy was significantly lower when assayed by non-extraction ELISA than when assayed by extraction ELISA (14.8 ± 9.3 vs. 145.1 ± 179.3, respectively; P < 0.001). Agreement between both measurements in late pregnancy was rather poor (r2 = 0.271, P = 0.022), as opposed to the good correlation found in early pregnancy (r2 = 0.929, P < 0.001). Conclusions The present study revealed that a melatonin assay without an extraction step, such as direct saliva ELISA, does not seem to be a valid method to determine the melatonin concentration of amniotic fluid, especially in late gestation.