Epidemiologic data on chronic inflammatory demyelinating polyneuropathy (CIDP) is limited, and previous studies have shown variable results. The frequencies of CIDP subtypes remain unknown. Variations due to use of different diagnostic criteria have not been studied. We examined the prevalence and incidence of CIDP in Leicestershire and Rutland, UK (population 963,600). Prevalence day was 1 May 2008. The prevalence of CIDP fulfilling the 2006 clinical and electrophysiologic European Federation of Neurological Societies/Peripheral Nerve Society (EFNS/PNS) criteria was 4.77 per 100,000 (95% confidence interval [CI] 3.49-6.37). Using the 1991 American Academy of Neurology (AAN) criteria, the prevalence was 1.97 per 100,000 in this population (95% CI 1.19-3.08). Lewis-Sumner syndrome was diagnosed in 15.2% of patients, and 23.9% had pure sensory onset. Over 40% required no immunotherapy, and 84.6% of those treated responded. More than 80% of the AAN criteria-negative but EFNS/PNS criteria-positive patients were responsive to treatment. Both sets of criteria were equally likely to identify patients who required therapy. The mean annual incidence rate over the 3 years preceding the prevalence day was 0.70 per 100,000/year using EFNS/PNS criteria (95% CI 0.43-1.08), and 0.35 per 100,000/year using AAN criteria (95% CI 0.17-0.64). We conclude that the AAN criteria may underestimate prevalence and incidence of the disease. The EFNS/PNS criteria provide higher diagnostic sensitivity and are of greater clinical relevance, and they also offer a useful breakdown of the epidemiologic data for CIDP subtypes.