Background Varenicline (Champix®), launched in the UK in December 2006, is indicated for the treatment of smoking cessation in adults (≥18 years of age). In 2008, the Medicines and Healthcare products Regulatory Agency (MHRA) in the UK issued a warning suggesting that varenicline was associated with disparate neuropsychiatric symptoms, including depression, suicidal thoughts and behaviour. In response to this regulatory warning, the Drug Safety Research Unit conducted a modified prescription-event monitoring (M-PEM) study to monitor the safety of varenicline. Objective The aim of this study was to estimate the incidence and examine the pattern of neuropsychiatric events reported to general practitioners (GPs) in England during the immediate postmarketing period for varenicline. Methods A postmarketing surveillance study was conducted using the observational cohort technique of M-PEM. Patients were identified from dispensed prescriptions issued by primary care physicians between December 2006 and March 2007. Data on exposure, previous history of psychiatric illness and events reported during and after treatment were collected from questionnaires. In order to determine whether hazards for neuropsychiatric events of interest (depression, anxiety, aggression, suicidal ideation, non-fatal self-harm) were non-constant over time (which could indicate a possible association with the drug), the pattern of events was examined by plotting the smoothed hazard function estimate and then fitting a Weibull model. The Weibull model shape parameter (β) and 95 % confidence interval were used as a test for a non-constant hazard function (where a value of 1 indicates a constant hazard over time). In addition to this analysis, the difference in incidence densities (IDs) between month 1 and months 2–3 were calculated and compared. Results The cohort comprised of 12,159 patients (median age 47 years [interquartile range 19]; 56.9 % [n = 6924 female]). The number of events reported during treatment, reason for stopping, adverse drug reactions (ADRs), and the p-value for the Weibull shape parameter were as follows: depression (n = 94; 42; 19; p = 0.144); anxiety (n = 94; 49; 9; p = 0.009); aggression (n = 7; 4; 2; p = 0.465); suicidal ideation (n = 8; 4; 1; p = 0.989) and non-fatal self-harm (n = 5; 1; 0; p = 0.771). No differences in the IDs between months 1 and months 2–3 were found for any of the events. Conclusion Whilst between 7 and 17 % of neuropsychiatric events were attributed to the drug by GPs and approximately 20–50 % were given as reasons for stopping, no signal was raised using the ID differences approach, and only anxiety was flagged as a potential signal for an ADR using the Weibull model. The signal for anxiety requires further evaluation to determine whether the drug plays a part in the development of anxiety or whether it is a withdrawal symptom caused by smoking cessation. Analysis methods will lack power when the numbers of events are low even when a large number of participants are included in the study.