The study was conducted in 151 freshwater fish aquacultures located in 63 local government areas of the Southern Guinea Savannah ecological zone of Nigeria. The zone has three sub-aqua-ecological zones characterised by variable climatic conditions (alternating warm and cold temperatures, favourable humidity and long duration of rainfall suitable for freshwater fish farming), many water bodies, rivers, streams, fadamas and dams (www.nigerstate.gov.ng).

A mixed study design of a cross-sectional study and participatory epidemiology (PE) approach was carried out between 2019 and 2020, and targeted freshwater fish aquaculture farmers and their ponds domiciled in rural and urban areas. For this research, participants were eligible for selection if each fulfilled the condition of being an artificial or natural pond fish farmer culturing fish for sustainable economic livelihood. Each participant has to engage in either an intensive management system, in which fishes are fed with an external food supply, or an extensive system that involves feeding fishes with phytoplankton and zooplankton naturally available, with no supplementation diets.

Sample size determination was performed with the OpenEpi version 3.1 software (Dean, Sullivan & Soe 2013). The power was set to 50% at 95% confidence level and 8% degree of precision for the study area (Southern Guinea Savannah zone). A total of 151 fish farms were obtained as a sample size and enrolled in the study. Samples were collected from 151 fish farms across the study area. Also, the sample size for biological samples was obtained using 50% power and an 8% margin of error set at 95% confidence level. The 151 biological samples were made up of fish tissues (fillets and livers, which are bioaccumulate residues), and pond water was collected and used for antibiotics screening.

For the participatory survey, preliminary information on both intensive and extensive fish farming communities was obtained from the Department of Fisheries in the region. These communities, domiciled in the three sub-aqua ecological zones, were identified and 9 out of the total 15 were conveniently considered. However, the number of communities considered per ecological zone for the PE exercises depended on the numbers and sizes of the available clusters of fish farms (concentration of two or more fish farming households in a settlement). Nine fishing communities were purposively selected. Two key informants were conveniently considered in each of the chosen fish farming communities, and a total of 18 key informants were considered and led by other fish farmers who participated in the PE exercises. However, the number of other participants was unrestricted.

A multi-stage sampling procedure was conducted to select the aquaculture ponds. In the first stage, the whole Southern Guinea Ecological zone was conveniently considered for sampling. For the second stage, the three sub-aqua-ecological zones were purposively considered, and fish pond registers of the three sub-zones were obtained from the Zonal Fisheries Offices as sampling frames. In the third and final stage, a systematic random sampling procedure, with a sampling interval of one, was conducted for pond selection. A sampling interval was obtained by dividing the calculated sample size (n = 151) by the number of biological samples (fillet, liver and pond water) expected to be collected (n = 151). A minimum of 50 freshwater farms were selected in each aquaculture sub-ecological zone, and biological samples were taken from them for quantitative antibiotic residue screening.

For the PE survey, key criteria for the selection of the fish farming communities were based on the number of available fish farm clusters in each zone as mentioned earlier. A purposive sampling method was used to select the communities and key informants for PE exercises

We designed a structured questionnaire that contained close-ended questions to ease response precision and used it to collect quantitative data (Thrusfield 2009). Questions were based mostly on literature and experts’ opinions. The tool had four sections: (1) socio-demographic characteristics of respondents (ponds owners or managers); (2) aquaculture ponds information; (3) knowledge about antibiotic usage in fish ponds, pathways for antibiotic residues and resistance spread to humans, and the socio-economic variables that drive residues and resistance occurrence; and (iv) attitudes towards antibiotic usage in freshwater fish aquaculture ponds. Six animal health technologists were recruited, trained and performed the interviewer-administered questionnaire exercise. The questionnaire was pre-tested on five farmers in the southern aquaculture-ecological zone and all gaps in the variables were identified and corrected to ease subsequent adequate data delivery. For quality control, we supervised questionnaire administration daily.

We verbally obtained the informed consent of participants before the commencement of questionnaire administering exercises as well as verbally informed of the survey aim. Respondents were informed of responses’ confidentiality, voluntary participation and the opportunity to withdraw at will without prejudice.

We considered the fact that most antibiotics administered to fishes are excreted into the pond water column unchanged and pass to water bodies as aquaculture discharges and run-offs, with major residues and resistance potentially impacting the environment. Therefore, fish, column and wastewater samples (15 fillets, 15 livers and 121 water samples) were collected from the 151 ponds and a few surrounding water bodies. Approximately, each tissue (fillet or liver) weighed 1 g and the water sample weighed 4 mL. We aseptically collected the samples in sterile sample bottles, packed them in an icebox at the temperature of 4 °C and took them to the laboratory for antimicrobial residue screening.

Each fish’s fillet or liver sample was triturated with a mixer, a 2 g homogenised sample was weighed in a 50 mL centrifuge tube and 4 mL 1% solution of trichloroacetic acid (Liquor 1) was added. This was oscillated for 2 min and centrifuged at 4000 rpm (revolutions per minute) at 37 °C for 10 min. About 250 μL supernatant was wiped out and 750 μL redissolving solution (Liquor 2) was added to dissolve the dried residue, and 50 μL of each sample was used for the assay. (Dilution factor: 8; detection limits: 0.4 parts per billion [ppb]). Also, about 1900 μL redissolving solution was added to each pond water sample to dilute and was mixed for 30 s; 50 μL of each sample was used for the assay. (Dilution factor: 40; detection limits: 2 ppb).

A tetracyclines (TCs) kit (Solarbio^® Life Sciences) was used for the determination of the TCs residues in samples. The ELISA kit was used for quantitative analysis of TC in tissues and conducted following the manufacturer’s guidelines. A calibration curve was plotted between the absorbance percentages of standards on the y-axis and the corresponding log of standards concentration (ppb) on the x-axis. A standard semi-log curve was drawn; the absorbance percentages were substituted and the corresponding concentrations were obtained from the standard curve. This was multiplied by the corresponding dilution factor to obtain the actual concentration of TCs in the samples.

The PE exercises were conducted using the participatory tools as described (Catley, Alders & Wood 2012; Alhaji & Babalobi 2016). We used participatory rural appraisal (PRA) tools of key informants, semi-structured interviews (SSI), checklists, proportional piling and triangulation to collect qualitative and semi-qualitative data.

A key informant was an elderly farmer with equal to or more than 10 years of fish farming management experience and commands much respect from members of the same occupation in the fishing community. In the participatory appraisal, only a SSI was used. The study team used a ‘checklist’ of important points covered during SSI, which were basically on antibiotics used and reasons for usage on ponds. It made the interview flexible and allowed respondents to express their words within their conceptual thoughts. The introduction of the appraisal team, as well as the aim, were done to the participants before commencing each PE session.

We conducted an SSI session for about an hour in each community. To facilitate discussion, the appraisal team firstly asked questions about fish ponds’ general management, secondly followed by specific questions on important antibiotics used in the management of bacterial infections, and thirdly other antibiotic uses in ponds within the preceding 1–5-year period of the interview. A checklist of open-ended questions was used to guide and standardise discussions, while the key informants’ responses were probed.

Counters (pebbles), flipcharts and permanent markers were used for this exercise. Farmers and the appraisal team compiled a list of commonly used antimicrobials mentioned during SSI. Farmers drew circles on the flip charts, each representing mentioned most commonly used antimicrobial. Some 100 pebbles were given to them and told to be piled in the respective named antibiotic circles proportionally to the frequency of perceived usefulness on the ponds. After consensus was reached among them, pebbles in each circle were counted to give a proportion that represented the relative frequency of each commonly used antimicrobial on the ponds in that particular fish farming community.

At the end of each proportional piling exercise in a community, participants cross-checked and debated on the piles until an agreement was reached on the obtained data. At the end of all the PE exercises in the nine (n = 9) fish farming communities, the appraisal team further triangulated (compared) obtained results and average scores or piles were computed. They represented mean final semi-quantitative (participatory) outcomes of the perceived relative frequency of antibiotics used in the ponds. Also, the appraisal team used the results of antibiotic screening (key biological samples) to triangulate or validate the semi-quantitative outcomes obtained during piling exercises (Mariner & Paskin 2000).

The One Health conceptual framework was designed to assess interactions of the various biophysical factors, environmental exigencies and the sociocultural activities that interface to drive antibiotic residues and resistance emergence in aquaculture. The model illustrated the interactions of the domains that converged to drive the emergence and points for One Health interventions (Figure 1).

We entered generated data into Microsoft Excel 7 (Microsoft Corporation, Redmond, Washington, United States [US]) spreadsheets, and analysed with OpenEpi version 3.1 (Dean et al. 2013). Descriptive statistics was used for response variables on knowledge and practices on antimicrobial use and resistance was presented as frequencies and proportions.

To assess associations of social and cultural factors (independent or explanatory variables) with antibiotic residue and resistance occurrence in the fish ponds, we created dependent (outcome) variables from the participant’s responses during interviewer-questionnaire administering. A 0–9 rating score scale, which represented response stringency to questions, was designed. We considered response score values within the range of 0–4 points as ‘poor’, and those within 5–9 points as ‘satisfactory’. Univariate analysis was first performed to assess associations between the explanatory and outcome variables using Chi-square tests or Fischer’s exact tests, where necessary (Dohoo, Martin & Studahl 2012). Variables found to be significant were then modelled using a likelihood backward multivariable logistic regression to control for confounders and effect modifiers. The difference between antimicrobial residue concentrations in the samples was determined with analyses of variance (ANOVA). A p-value set at 0.05 was considered significant in the analyses.

We used Kendall’s Coefficient of Concordance W statistic, a non-parametric statistic (Kendall & Smith 1939; Legendre 2010; Siegel & Castellan 1988), to analyse the semi-quantitative data obtained from the PE exercises for the level of agreements among the farmers in the fish farming communities. W values vary from 0 to 1. W = 0 indicates there was no agreement and W = 1 means there was very strong agreement at 95% confidence level. A p < 0.05 was considered statistically significant for the agreement among them. Furthermore, we used descriptive statistics, mean and proportions (%) to analyse data on different frequently used antibiotics in the fishing communities. An online software programme (https://www.statstodo.com/KendallW.php) was used.

The study proposal was submitted and cleared by the Niger State Ministry of Livestock and Fisheries Research Ethics Committee Ref. MLF/NGS/REC/03/2017. Clarias gariepinus, also known as African catfish, were sampled and humanely sacrificed (Blessing, Marshall & Balcombe 2010).

All the selected 151 aquaculture pond farmers responded and 78.2% (n = 118/151) indicated antibiotic use in fish farms. About 88.7% (n = 134/151) of the participants were males and 85.4% (n = 129/151) of them were married. Furthermore, 58.9% (n = 89/151) of the participants possessed tertiary education, whereas 9.9% (n = 15/151) of them had no formal education (Figure 2).

About 78.1% (n = 118/151) of the participants reported possessing knowledge about antibiotic usage in freshwater fish aquaculture, while 21.9% (n = 33/151) did not. Those with the knowledge obtained index information from friends (51.7%, n = 61/118), relations (40.7%, n = 48/118) and animal health personnel (7.6%, n = 9/118). Few farmers (22.9%, n = 29/118) knew antibiotics for being used in the treatment of infections in freshwater aquaculture and 49.2% (n = 58/118) reported prevention. Furthermore, 22.0% (n = 26/118) of the farmers knew the drugs to be used for metaphylaxis. Also, 11.0% (n = 13/118) responded knowing misuse of antibiotics in freshwater fish aquaculture when given without prescription, 7.6% (n = 9/118) knew it to be when administered for viral and fungal infections, and 77.1% (n = 91/118) did not know about antibiotic misuse. However, 26.3% (n = 31/118) agreed that antibiotic misuse and overuse in freshwater aquaculture predisposed to residue and resistance emergence, while 73.3% (n = 87/118) of them disagreed. In addition, 41.5% (n = 49/118) did not know about the effects of residue and resistance in aquaculture fish (Table 1).

About 78.0% (n = 118/151) of the farmers used antibiotics in freshwater fish aquaculture. Meanwhile, 94.1% (n = 111/118) of farmers practised self-prescription of antibiotics used, while 5.9% (n = 7/118) patronised veterinarians’ services. About 62.7% (n = 74/118) of the farmers acquired antibiotics used in ponds from veterinary pharmacies, 26.3% (n = 31/118) from human pharmacies and 11.0% (n = 13/118) patronised drug hawkers. Also, 56.8% (n = 67/118) of the farmers carried out arbitrary application of antibiotics in ponds, and 43.2% (n = 51/118) used accompanied label instructions. Furthermore, 94.9% (n = 112/118) of the farmers conducted self-administration of antibiotics used in fish aquaculture, while 5.1% (n = 6/118) invited the services of veterinary authorities.

On the rate of antibiotics use on the farms, 20.3% (n = 24/118) of the farmers used prescribed leaflet instructions, a single high dose of antibiotics was daily administered by 26.3% (n = 31/118) of the farmers, while a single high dose of antibiotics was weekly administered by 53.3% (n = 63/118) of them. However, only 5.1% (n = 6/118) of them observed antibiotic withdrawal periods in fish aquaculture after usage while 94.9% (n = 116/118) did not observe it. While 7.6% (n = 9/118) of the farmers applied antibiotics for growth promotion, 34.7% (n = 41/118) of them used antibiotics for metaphylaxis (Table 2).

Identified significant pathways of antibiotic residue and resistance spread to humans through freshwater fish aquaculture were antibiotic residue contaminated fish and fish products ingestion (p = 0.001) and contact with antibiotic residue contaminated fish and fomites (p = 0.001) spread. Furthermore, column water and aerosols contaminated with antibiotic residues are pathways for the dissemination of residues to humans from freshwater fish aquaculture environments (p < 0.001) (Table 3).

Modelled multivariable logistic regression significantly indicated that antibiotic misuse and overuse were more likely to drive antibiotic residue and resistance occurrence in freshwater fish aquaculture (Odds Ratio [OR] = 3.8; 95% confidence interval [CI]: 1.62, 8.74; p = 0.002). Also, poor education and expertise of the farmers were more likely to drive antibiotic residues and resistance (OR = 2.9; 95% CI: 1.24, 6.94; p = 0.010). The farmers’ poor financial position was five times more likely (OR = 5.1; 95% CI: 1.85, 14.00; p = 0.002) to drive residue and resistance occurrence; and the same with the intensive management system that was eight times more likely (OR = 8.5; 95% CI: 3.17, 22.58; p = 0.001) to drive the occurrence. Also, poor biosecurity practices and sanitation within and around the farm environment (OR = 5.1; 95% CI: 1.88, 9.31; p = 0.001) were more likely to influence antibiotic residues and resistance occurrence in freshwater fish aquaculture (Table 4).

The concentrations of TCs residues in fillets ranged from 27.9 μg/kg⁻¹ to 31.4 μg/kg⁻¹ with a mean value of 29.6 μg/kg^−1; in livers, it ranged from 25.6 μg/kg⁻¹to 27.9 μg/kg⁻¹ with a mean of 26.4 μg/kg⁻¹. The TCs residue concentrations in the water column and discharged wastewater ranged from 108.3 μg/kg⁻¹ to 133.4 μg/kg⁻¹ with 123.6 μg/kg⁻¹ mean. The mean values of TCs residue concentrations in fillets, livers and water column and discharged wastewater were: 29.6 ± 9.1 μg/kg⁻¹, 26.4 ± 8.6 μg/kg⁻¹, and 123.6 ± 18.2 μg/kg⁻¹, respectively. However, the mean TCs residue level (123.6 ± 18.2 μg/kg⁻¹) above 100 μg/kg⁻¹ maximum residue limits (MRL) were observed in the water column and discharged wastewater. The residue concentrations in fillets, livers and column and/or waste were significantly different with mean square = 37597.6 and p = 0.001 (Table 5).

Proportional piling was a useful tool extensively used to confirm verbal descriptions in the SSIs. In this study, farmers exhibited adequate existing knowledge about health challenges management in fishing communities. Furthermore, participants averagely ranked commonly applied antibiotics in the fish farms in this order: TCs, penicillin, sulphonamides and streptomycin, among others (Figure 3). The mean proportional pile (relative usage) of TCs among other antibiotics frequently used was 15.8%, which was judged by the farmers as the first most important antimicrobial. There was strong (W = 0.979) and significantly concorded statistically (p < 0.001) among the fish farmers on the piled antibiotics according to the relative frequencies of use in freshwater fish farms.