A hierarchical Bayesian approach for risk assessment of melamine in infant formula based on cases of related nephrolithiasis in children

Abstract

Although the 2008 outbreak of nephrolithiasis in children due to melamine-contaminated infant formula has subsided, it remains uncertain whether the present tolerable daily intake (TDI) of melamine provides sufficient protection for young children. To conduct a safety assessment for melamine in infant formula, we established a dose–response relationship based on 13 nephrolithiasis cases selected from 932 children, all of whom were under 5 years of age and had potentially been exposed to contaminated milk in China or Taiwan. According to the children's exposure history, distributions of individual daily melamine intake (mg/kg BW/day) were reconstructed using Monte Carlo simulations to account for uncertainties in exposure duration and melamine concentrations in the contaminated milk. Based on the simulated individual average daily intake (AVDI) of melamine, subjects were further classified into four separate AVDI groups: high, medium, low and a reference group. A statistical logistic model was then fitted for the dose–response relationship between nephrolithiasis incidence and daily melamine intakes using Markov chain Monte Carlo (MCMC) simulations. Based on the background exposure, spontaneous rate, and mode of action (MOA) of nephrolithiasis in children, the simulated lower bounds of the 95% CIs daily melamine intake ranged from 0.008 to 0.03 mg/kg BW/day corresponding to an additional risks of 0.1% is proposed as a plausible TDI, which is approximately an order lower than the current WHO-suggested TDI level of 0.2 mg/kg BW/day. More stringent regulations on melamine levels in infant formula should be considered to protect young children fully.

Keywords:

• risk assessment
• modeling
• additives
• baby food
• milk

Acknowledgements

The authors would like to thank Professors Trong-Neng Wu, Chow-Feng Chiang, and Hung-Lung Chiang for providing suggestions for this paper; Dr. Chih-Jung Chen for collecting the data; and Chia-Yen Yu for statistical programming and plots. This work was supported by the National Science Council of Taiwan (NSC 97-2314-B-192-001-MY2). The authors declare that they have no conflicts of interest.

Notes

Both authors contributed equally to this study.

Additional information

Notes on contributors

C.-C. Chen

Both authors contributed equally to this study.