Design: systematic review and Bayesian meta-analysis, PROSPERO registered, 20 databases searched
Population: 77 studies of people presenting with epileptic seizures in endemic regions; 35 with clear case definitions in the pooled estimate
Primary outcome: pooled proportion with neurocysticercosis diagnosed by imaging, biopsy or autopsy
Effect: 25.2% (95% Bayesian credible interval 19.5-32.2), with high between-study heterogeneity and a decline after 2008

A Bayesian meta-analysis searched twenty databases and screened 16,202 records to estimate how often neurocysticercosis lies behind epileptic seizures in Taenia solium endemic regions. Seventy-seven studies met inclusion; the pooled estimate rests on the 35 that used clear case definitions for both the infection and the seizures, and that diagnosed by neuroimaging, biopsy or autopsy.

Among those, 25.2% of people presenting with epileptic seizures had neurocysticercosis (95% Bayesian credible interval 19.5 to 32.2). Between-study heterogeneity was substantial. Studies without clear definitions reported considerably lower proportions, which the authors read as misclassification rather than as genuine variation. The proportion did not differ by sex or age, and it declined between the 1990-2008 period and 2008-2023 across Africa, the Americas and southeast Asia.

For Indian practice this quantifies something already assumed. It supports imaging a first unprovoked seizure rather than treating empirically, and it makes the ring-enhancing lesion a probability rather than a curiosity. The decline is the more interesting finding and the least secure: the authors are explicit that it may reflect changing case definitions, better imaging or different study populations rather than a real fall in transmission.

Image a first unprovoked seizure in anyone from or resident in an endemic area
Record the case definition you are using - the review shows how much loose definitions distort the estimate
Consider the pork exposure and household history in the same consultation as the imaging request
Do not read the falling proportion as proven declining transmission
Remember the pooled figure is not transportable: heterogeneity between settings was high

Why it matters

It puts a defensible number on a diagnosis that in endemic practice is usually treated as an impression.

The statistics, in plain English

A credible interval of 19.5 to 32.2% is the Bayesian equivalent of a confidence interval, and its width here is honest rather than reassuring: the studies genuinely disagreed, with a heterogeneity parameter the authors describe as fairly high. That means the pooled 25.2% describes the body of studies, not any particular clinic - a hospital in a high-transmission district will see more, an urban tertiary centre fewer. Reading it as a local prior rather than a local rate is the correct use.