Meta-analysis blueprint sets 34,975-sample surveillance plan for 14 animal diseases across India

India’s veterinary epidemiologists have produced something the country has never had before: a single, statistically rigorous sampling blueprint for measuring how common 14 dangerous diseases are across the world’s largest livestock population. The plan, described in Frontiers in Veterinary Science, pools hundreds of Indian prevalence studies through a systematic review and meta-analysis and then converts the resulting estimates into a concrete field campaign of 34,975 samples spread across 2,018 villages. The diseases on the list include ten zoonoses that can jump from animals to people—brucellosis, tuberculosis, Q-fever, scrub typhus, Japanese encephalitis, Crimean-Congo haemorrhagic fever, cysticercosis, cryptosporidiosis, listeriosis and salmonellosis—alongside four transboundary animal diseases: swine influenza, porcine reproductive and respiratory syndrome, lumpy skin disease and African swine fever.

The work was carried out by K. P. Suresh and M. Ashwini of the Indian Council for Agricultural Research – National Institute of Veterinary Epidemiology and Disease Informatics in Bengaluru, together with P. Azhahianambi of Tamil Nadu Veterinary and Animal Sciences University and Subeer S. Majumdar and Nagendra R. Hegde of the Biotechnology Research Innovation Council – National Institute of Biotechnology in Hyderabad. It grew out of the DBT One Health Consortium, a network of 28 centres established in 2021 with support from the Department of Biotechnology to connect the veterinary, wildlife and medical sectors. The consortium’s founding goal was precisely the one this study now serves: nationwide baseline prevalence data for significant zoonotic diseases and the transboundary animal diseases that recently affected northeast India, in what the authors describe as the largest One Health-related initiative in the country’s post-COVID era.

The technical core of the study is a PRISMA-compliant systematic review. The team searched Google Scholar, ResearchGate, Krishikosha, PubMed and ScienceDirect for Indian prevalence studies published up to 2021, combining disease names with terms such as prevalence, epidemiology and cross-sectional study. Duplicate records were removed with Zotero, and titles, abstracts and full texts were screened independently. Studies qualified only if they reported original Indian prevalence data, clearly described the study population and diagnostic methods, and supplied enough information for prevalence estimation; reviews, case reports, experimental studies and duplicate publications were excluded. Two junior reviewers rated face validity and two senior reviewers rated content validity on five-point Likert scales, with agreement assessed by quadratic-weighted Cohen’s kappa. The kappa values ranged from 0.63 for porcine reproductive and respiratory syndrome to 0.91 for swine influenza, spanning substantial to almost perfect agreement.

Because the primary studies differed so markedly in geography, populations, sampling strategies and diagnostic assays, the researchers anticipated considerable between-study heterogeneity and therefore pooled the prevalence estimates with a random-effects model, which gives more weight to more precise studies. The heterogeneity statistics confirmed that decision emphatically. The I² values ranged from 91 percent for Q-fever to 100 percent for swine influenza and porcine reproductive and respiratory syndrome, and heterogeneity was statistically significant for every disease evaluated. The corresponding τ² estimates, which quantify true between-study variance, ranged from 0.0168 for Q-fever to 0.1400 for cryptosporidiosis. In other words, the spread in reported prevalence reflected genuine differences among the primary studies rather than sampling noise alone.

The pooled estimates themselves ranged from 4 to 30 percent across the 14 diseases. Cysticercosis sat at the bottom with a pooled prevalence of 4 percent, followed by listeriosis at 9 percent and Q-fever at 10 percent. Brucellosis and tuberculosis each pooled at 12 percent, Crimean-Congo haemorrhagic fever at 11 percent, Japanese encephalitis and salmonellosis at 14 percent, lumpy skin disease at 18 percent and African swine fever at 19 percent. At the top end, porcine reproductive and respiratory syndrome pooled at 29 percent, cryptosporidiosis at 27 percent, scrub typhus at 22 percent and swine influenza at 30 percent. The African swine fever figure deserves a caveat: because the disease was first reported in India only in 2020, no India-specific field prevalence data existed when the review began, so the design prevalence was conservatively deduced from studies conducted outside Asia.

Two other diseases forced unconventional sampling units. For scrub typhus and listeriosis, the Indian literature contained insufficient animal-reservoir prevalence data to derive a livestock-based design prevalence. The pooled estimate for scrub typhus therefore came from human case data, and the plan targets rodent and vector collection sites within selected villages rather than live animals. For listeriosis, the estimate was derived from food-matrix studies, and the sampling plan directs its efforts at food product batches from processing and retail premises serving the same villages. The overall totals of 34,975 samples and 2,018 villages include these non-livestock components, which are itemised by disease in the study.

Turning pooled estimates into a field plan required a two-stage stratified random sampling design, generated with an in-house epi-calculator originally developed under the National Animal Disease Referral Expert System version 2 at ICAR-NIVEDI. Villages served as the primary sampling units and animals within villages as the secondary units, with the village acting as the stratification variable for the targeted animal species in each state. The calculations followed the two-stage cluster-sampling formulae of Cameron and Baldock for detecting disease at a specified design prevalence, adjusted for imperfect diagnostic sensitivity and specificity and corrected for finite populations at both the village and animal levels. The design targeted a 95 percent surveillance confidence level, meaning the probability of detecting at least one positive sample if the disease is present at or above the specified design prevalence.

The resulting disease-specific allocations show how the burden is distributed. Brucellosis, which can infect cattle, buffalo, sheep, goats and pigs, requires 160 villages and 3,228 samples, with 18 to 21 samples per village. Q-fever demands 185 villages and 4,517 samples from cattle and buffalo, while listeriosis, despite its lower pooled prevalence, requires 156 villages and 4,183 samples because of the food-batch sampling. Tuberculosis calls for 158 villages and 3,230 samples, Crimean-Congo haemorrhagic fever 154 villages and 3,695 samples, and Japanese encephalitis 134 villages and 2,389 samples focused on pigs. Swine influenza, the highest-prevalence disease, needs only 105 villages and 1,075 samples, with 9 to 10 samples per village, because higher design prevalence reduces the sample size needed for detection.

The authors are careful about what their numbers do and do not mean. Because the pooled estimates were derived predominantly from convenience and purposive samples in the primary literature rather than from a nationally representative baseline survey, they should be interpreted as design-stage estimates for sample-size planning rather than as definitive national prevalence figures. States contributing a disproportionate number of primary studies—for example for brucellosis and cysticercosis—may not be representative of India as a whole, and prevalence is likely to vary by agro-climatic zone, livestock management system and species. The team therefore recommends that state-level implementing agencies treat the calculated sample sizes as minimum requirements and consider stratifying further by known risk factors where local epidemiological data permit. Publication bias, particularly for diseases represented by few studies, also cannot be entirely excluded and may have skewed some estimates.

Even with those limitations, the study stands as the first scientifically rigorous meta-analysis-based sampling plan assembled under one platform for many diseases in India, and it arrives at a moment when the country is thought to be one of the global hotspots for zoonotic emergence, given dense human and livestock populations living in proximity and rapid socio-ecological change. Timely detection reduces the risk of spread to new areas, minimises infections and mitigates financial impact, and active surveillance—periodic sampling and testing by veterinary officials—tends to outperform passive reporting, especially in resource-limited systems. If the consortium’s field teams execute the plan, the resulting baseline prevalence data could form the foundation of future control programmes and give India’s One Health ambitions the quantitative backbone they have so far lacked.

Subject of Research: A meta-analysis-guided sampling plan for estimating the prevalence of 10 zoonotic and 4 transboundary animal diseases across India

Article Title: A meta-analysis-guided sampling plan for estimating the prevalence of 10 zoonotic and 4 transboundary animal diseases across India

Article References: Suresh, K. P., Ashwini, M., Azhahianambi, P., Majumdar, S. S., & Hegde, N. R. (2026). A meta-analysis-guided sampling plan for estimating the prevalence of 10 zoonotic and 4 transboundary animal diseases across India. Frontiers in Veterinary Science, 13, Article 1919244. https://doi.org/10.3389/fvets.2026.1919244

Image Credits: AI Generated

DOI: 10.3389/fvets.2026.1919244

Keywords: meta-analysis-guided, sampling, plan, estimating, prevalence, zoonotic, transboundary, animal, diseases, across, India, scientific research