Estimating the prevalence of clinical manganism using a cascaded screening process in a South African manganese smelter

Objectives: A diagnostic algorithm for clinical manganism was developed to screen all employees at a South African manganese smelter. Methods: The study design was for all 754 smelter employees in 2006/7 to be screened by an occupational health nurse using nine questions and nine brief neurological examination procedures. More than one symptom, any neurological sign, or blood manganese exceeding 40 μg/l triggered referral for neurological examination by an Occupational Medical Practioner (OMP). Abnormal findings by the OMP triggered referral to a movement disorders specialist neurologist and to a neuropsychologist. Features of parkinsonism and a clinical picture consistent with the scientific literature were used to diagnose manganism. Results: Total manganese dust was mostly within (<5 mg/m3) or near (<9 mg/m3) the South African Occupational Exposure Limit, with one outlier near 20 mg/m3. Occupational Health Service problems and uncertainty about the nature of manganism before the full diagnostic algorithm was developed, resulted in 10 referrals who were certified as manganism cases by the state compensation authorities. They were only assessed in the early stages of this screening programme, and never examined by the above specialists. Of 744 employees screened with the full diagnostic algorithm, the nurse referred 152 (20.3%) and the OMP 27 (3.5%) of all those screened respectively. No definite manganism cases were diagnosed, while one (0.13%) employee was found to have possible manganism and another had an indeterminate neurological diagnosis. A sensitivity analysis assuming that all 10 compensated cases were either normal, or alternatively had definite manganism, yielded a prevalence range for definite manganism from 0% to 1.3%. Conclusion: Acknowledging possible downward bias when excluding the 10 employees who did not receive the full workup, the true prevalence of definite manganism was likely to be either zero or close to zero. © 2009 Elsevier Inc. All rights reserved.