Feed availability and quality remain major constraints to ruminant productivity in West Africa, where livestock systems rely heavily on locally available resources such as natural forages, crop residues and agro-industrial by-products. However, reliable ration formulation requires accurate information on feed chemical composition, while existing data are fragmented and highly variable. This study conducted a systematic review of peer-reviewed literature published between 2000 and 2025 to synthesize available data on the chemical composition of ruminant feeds in West Africa. Following PRISMA guidelines, 44 studies reporting quantitative feed composition data were retained. Feed resources were classified into agro-industrial by-products, agricultural by-products and forages, and descriptive statistics were calculated for key nutritional parameters. The results revealed substantial variability in nutrient composition across feed types and even within the same feed resource. Cottonseed cake emerged as a major protein-rich supplement, legume haulms showed higher nutritional value than cereal residues, and several browse species such as Moringa oleifera and Leucaena leucocephala demonstrated high protein potential. These findings highlight that fixed feed composition values are poorly suited to heterogeneous tropical feeding systems. The reference ranges established in this review provide a more reliable basis for feed evaluation and ration formulation and can support the development of locally adapted feeding strategies and decision-support tools for West African livestock systems.
Accurate estimation of enteric methane emissions is important for improving greenhouse gas inventories and supporting mitigation strategies in tropical cattle production systems. This study aimed to develop empirical equations for predicting daily enteric methane emissions from growing bulls managed under semi-intensive conditions in Benin. The database comprised 507 valid observations obtained from 39 crossbred growing bulls across 13 nutrition trials conducted at the Okpara Breeding Farm. Enteric methane emissions were measured individually using the GreenFeed® system, while dry matter intake (DMI), body weight (BW), and average daily gain (ADG) were obtained from the experimental datasets. Nine candidate linear mixed-effects models incorporating these variables individually, additively, and through selected interactions were evaluated. Model performance was compared using the Akaike information criterion (AIC), Bayesian information criterion (BIC), marginal R2, root mean square error (RMSE), and mean absolute error (MAE). The model combining DMI and BW (M5: CH4 = 170.04 + 4.09 × DMI − 0.22 × BW) provided the best balance between model fit and parsimony, with the lowest AIC (−8143.03) and a marginal R2 of 0.671. DMI was positively associated with daily methane production, whereas BW showed a negative conditional association after adjustment for DMI. More complex models including ADG or interaction terms did not provide sufficient improvement in model fit to justify their additional complexity. The resulting equation provides a preliminary locally derived approach for estimating methane emissions under the production conditions represented in the present dataset. However, because model development and evaluation were based on the same database, independent validation using larger and more diverse cattle populations is required before broader application.