During the past decade, a number of large drug trials suggested that the initiation of levodopa therapy should be delayed to reduce the risk of motor complications in patients with Parkinson’s disease. However, the relative contribution of the cumulative exposure to levodopa and of disease progression to the pathophysiology of motor fluctuations and dyskinesias is still poorly understood. In this 4-year multicentre study, we investigated a large cohort of patients with Parkinson’s disease in a sub-Saharan African country (Ghana), where access to medication is limited and the initiation of levodopa therapy often occurs many years after onset. The primary objective was to investigate whether the occurrence of motor complications is primarily related to the duration of levodopa therapy or to disease-related factors. Study design included a cross-sectional case-control analysis of data collected between December 2008 and November 2012, and a prospective study of patients followed-up for at least 6 months after the initiation of levodopa therapy. Ninety-one patients fulfilled criteria for clinical diagnosis of idiopathic Parkinson’s disease (58 males, mean age at onset 60.6 ± 11.3 years). Demographic data were compared to those of 2282 consecutive Italian patients recruited during the same period, whereas nested matched subgroups were used to compare clinical variables. Demographic features, frequency and severity of motor and non-motor symptoms were comparable between the two populations, with the only exception of more frequent tremor-dominant presentation in Ghana. At baseline, the proportion of Ghanaian patients with motor fluctuations and dyskinesias was 56% and 14%, respectively. Although levodopa therapy was introduced later in Ghana (mean disease duration 4.2 ± 2.8 versus 2.4 ± 2.1 years, P < 0.001), disease duration at the occurrence of motor fluctuations and dyskinesias was similar in the two populations. In multivariate analysis, disease duration and levodopa daily dose (mg/kg of body weight) were associated with motor complications, while the disease duration at the initiation of levodopa was not. Prospective follow-up for a mean of 2.6 ± 1.3 years of a subgroup of 21 patients who were drug-naïve at baseline [median disease duration 4.5 (interquartile range, 2.3–5) years] revealed that the median time to development of motor fluctuations and dyskinesias after initiation of levodopa therapy was 6 months. We conclude that motor fluctuations and dyskinesias are not associated with the duration of levodopa therapy, but rather with longer disease duration and higher levodopa daily dose. Hence, the practice to withhold levodopa therapy with the objective of delaying the occurrence of motor complications is not justified.