Wastewater-Based Antimicrobial Resistance Surveillance: A Systematic Review of Metagenomic and Complementary Molecular Evidence from Wastewater and Wastewater-Impacted Environments

Two reviewers independently screened the title and abstract and, based on the predetermined inclusion criteria, determined whether the articles met the inclusion criteria. Reports were then evaluated in full text at the same time for potentially eligible reports. At both meetings, differences were resolved by discussion and consensus. Records from the bibliographic databases were merged before screening, and duplicate records were eliminated before the assessment of the titles and abstracts. The reasoning for including or excluding articles at the screening stage was recorded to help with PRISMA reporting purposes.

The scope of this review was refined during the revision process to include wastewater and wastewater-impacted environments, including municipal and hospital wastewater, wastewater treatment plant influents and effluents, sludge/biosolids, and receiving environments affected by wastewater discharge. This amendment was introduced to better capture the environmental pathways through which antimicrobial resistance genes (ARGs) occur and persist.

2.5. Data Extraction and Analysis

Data from the 15 included studies were extracted independently by two reviewers using a structured template. Extracted variables included author/year, study aim, setting/location, number and type of sites where reported, environmental matrix, analytical method, treatment stage or comparison, principal ARG findings, and study/appraisal limitations. Discrepancies were resolved by discussion and consensus. Because the included studies differed in sampling design, sequencing depth, ARG databases, normalization, matrix and analytical platform, formal meta-analysis was not appropriate. A narrative synthesis was organized around ARG diversity and prevalence, genomic context/potential mobility, treatment-stage persistence or removal, and environmental/public-health interpretation. No formal sensitivity analysis, reporting-bias analysis, or GRADE-style certainty assessment was undertaken.

2.6. Quality Assessment

JBI critical appraisal tools were selected according to the way each included study was reported. Studies structured as cross-sectional comparisons of settings, sources or analytical approaches were appraised with the JBI Analytical Cross-Sectional Studies Checklist [18], whereas descriptive occurrence/prevalence surveys were appraised with the JBI Prevalence Studies Checklist [19, 20]. Ten studies were assigned to the cross-sectional checklist and five to the prevalence checklist, with each included study appearing once. Responses were coded as Yes, No, Unclear or Not applicable according to whether the published report clearly satisfied the item, failed it, provided insufficient information, or made the item inapplicable. The JBI tools were used to characterize reporting and methodological limitations, not to validate the review’s scientific conclusions (Supplementary Appendix C).

A methodological appraisal of studies using metagenomic approaches was performed in addition to the general appraisal of the observational-study tools, because there are some technical sources of bias which are not fully captured by the general tools used for observational study appraisal. The appraisal took into account sampling representativeness and replication, sampling and concentration, principle of DNA extraction, depth of sequencing and QC procedures, negative controls, antimicrobial-resistance gene database and versions, identification thresholds, read-based versus assembly-based analyses, abundance normalisation, host assignment, linkage of mobile genetic elements, and availability of sequencing data and analytical code, as indicated in Supplementary Appendix D, E.

2.7. Ethical Consideration

Ethical standards were observed in line with international and institutional guidelines relevant to the study designs included in this review. This review did not involve direct experimentation with humans or animals. Ethical approval was not required and informed consent was not obtained since this was a secondary analysis of previously made studies. However, every effort was made to ensure that there was strict reporting, transparency and that guidance in conducting and reporting reviews was followed.

3. RESULTS

3.1. ARG Diversity and Prevalence Across Wastewater Matrices

The 15 included studies comprised predominantly shotgun-metagenomic investigations together with selected complementary molecular studies. The study by Hendriksen et al. [21] showed a link between national antibiotic use prevalence, sanitation infrastructure and prevalence of ARGs, highlighting the role of sociodemographic factors. Similarly, Becsei et al. [22] noted differences among five cities in Europe, where the outcomes of bioburden testing were affected by Pseudomonas blooms that inhibited the detection of some ARGs. This highlights that the resistome may suffer from microbial community composition bias.

Hospital-wastewater studies reported higher ARG abundance or diversity than municipal/community wastewater in some settings, including clinically relevant blaNDM-1, blaOXA and mcr variants (Table 1). These observations identify settings with elevated resistance determinants but do not by themselves demonstrate selection, amplification, dissemination or transmission. Jantharadej et al. [23] reported 611 ARG subtypes in Bangkok canal-water samples, including abundant β-lactam resistance determinants and pathogenic genera such as Arcobacter and Burkholderia. The canal findings are consistent with wastewater influence on the receiving environment, but do not establish direct transmission pathways.

Interestingly, culture-based and targeted qPCR approaches, such as those conducted by Tsvetanova and Boshnakov [24] and Magnano San Lio et al. [25] detected fewer ARGs but a higher relative abundance of clinical relevance resistance genes including erm(B) or blaOXA. This illustrates the trade-off between detection breadth and quantitative sensitivity, indicating that methodological approaches directly influence the diversity and composition of the resistome observed. Although these studies have all demonstrated the ubiquitous presence of ARGs across wastewater matrices, the diversity and abundance reported is systematic, depending on sampling design, sequencing depth and analytical methods (Supplementary Appendix D, E).

3.2. ARG Co-occurrence, Genomic Context, and Potential Mobility

The included studies provided different levels of evidence regarding the genomic context and potential mobility of ARGs. Detection of ARGs in metagenomic reads confirmed the presence of ARGs in a sample. In contrast, taxonomic co-occurrence only confirmed the presence of ARGs and a specific microbial group within a microbial community. However, these levels of evidence should not be interpreted as equivalent to demonstrated transmission, because genomic association or potential mobility does not confirm successful horizontal gene transfer, survival of resistant organisms, or human exposure. More evidence was available for ARGs localized to assembled contigs, metagenome-assembled genomes, plasmids, integrons, or other mobile genetic elements. Genomic linkage does not necessarily imply horizontal gene transfer, but can imply potential for mobility. Likewise, the presence of ARGs in taxa that contain opportunistic pathogens does not mean that the resistance gene was on a viable pathogenic strain unless a host assignment was done with high enough genomic specificity. The integron (intI1) was enriched in the influent samples and often found to be associated with the carbapenemase genes such as blaNDM and colistin resistance genes such as mcr-1 as reported by Sekizuka et al. [26]. In the same way, plasmid associated ARGs were also identified in E. coli, Klebsiella and Acinetobacter by Prieto Riquelme et al. [27] indicating that the distribution of resistomes is related to the bacterial host and the potential for genetic mobility.

Variability across time was observed in some investigations. In Rotterdam, the microbial community of the sewer changed significantly across the seasons [26]. In Copenhagen, the microbial community was different. However, measurements of selected ARGs showed seasonal variation in effluents from the Copenhagen WWTP [26], with some resistance determinants being more abundant in summer than winter effluents. While longitudinal surveillance of selected ARGs was provided in three Sicilian wastewater-treatment plants by Magnano San Lio et al. [25], the authors did not provide supporting analysis to conclude that the plants experienced seasonal changes due to prescription. A comparison of genomic findings across studies revealed significant variability in ARG types and prevalence due to differences in sampling context, sequencing depth, and analysis methods. For example, Hendriksen et al. [21] detected a wide range of ARGs globally, while Silvester et al. [28] found a higher prevalence of carbapenem resistance in hospital wastewater (Table 2). Studies with higher sequencing depth captured rare ARGs, while those using 16S rRNA sequencing missed broader resistome diversity [15, 25]. These genomic discrepancies emphasize the need for standardized methodologies to enable more consistent and comparable findings across studies, highlighting the critical role of hospital effluents in the environmental occurrence or dissemination potential of AMR determinants. The evidence was therefore interpreted hierarchically as follows: ARG detection in unassembled reads; ARG–taxon co-occurrence; ARG localization within assembled contigs or metagenome-assembled genomes; ARG linkage to plasmids, integrons or other mobile elements; and, at the highest level, direct evidence of environmental or human transmission of viable phenotypically resistant organisms. Most included studies provided evidence within the first four levels of this hierarchy; however, none demonstrated direct environmental or human transmission of viable resistant organisms.

Acosta et al. [29] showed that meropenem enrichment increased detection of rare carbapenemases, demonstrating that analytical sensitivity can affect whether low-abundance ARGs are detected. Low abundance alone should not be interpreted as disproportionate public-health risk without information on host, viability, mobility, exposure and clinical relevance. Becsei et al. [22] and Jantharadej et al. [23] also reported associations between ARG signals and microbial taxa. Co-occurrence or association with MGEs can indicate possible genomic context or mobility potential, but does not demonstrate horizontal gene transfer or subsequent transmission to humans or animals.

3.3. Stage-Wise ARG Removal or Persistence

The changes in the indicators of antimicrobial resistance in each of the stages of the wastewater-treatment process were different depending on both the treatment process and the method of measurement. Reported outcomes were absolute abundance, relative abundance, detection frequency, prevalence of resistant organisms, and log reduction, and were not directly interchangeable and were interpreted separately. If an overall biomass of bacteria is decreased during treatment, a decrease in absolute ARG concentration can occur with an increase in relative abundance.

Tsvetanova and Boshnakov [24] reported decreases in culturable counts of AMR bacterial species in treated effluents in all cases, except for Enterobacteriaceae and Enterococcus, which were detected. Sekizuka et al. [26] observed about a 1-log reduction in some signals associated with the ARGs from the influent to effluent. Majeed et al. [30] reported a total ARG abundance approximately 50% less in secondary effluent than in influent, and that approximately 90% of the ARG types found in secondary effluent also appeared in the influent. The findings suggest that the ARG diversity continues to exist despite a decline in some abundance measures. Quantitative comparison of the performance of the different treatments was not appropriate due to differences in the data analysis platform employed, microbial denominators, and sequencing depths used in the included studies. Treatment results were therefore evaluated and compared within each study, based on the results reported in each and not across technologies by incompatible measures.

3.4. Metagenomic Methodological Appraisal

The metagenomic appraisal revealed differences in the reported technical and analytical parameters among metagenomic studies (Table 2). Some discrepancies were noted regarding the reporting of sampling replication, sequencing depth, quality control procedures, ARG database selection, normalisation methods, assembly-based analyses, host assignment, and the availability of sequencing data/analytical code [28, 29, 31]. These differences are sources of methodological heterogeneity and should be taken into account when comparing metagenomic findings from studies.

3.5. Implications for Environmental and Public Health Risk

Repeated detection of clinically relevant ARGs in wastewater and wastewater-impacted waters supports their use for environmental hazard identification and surveillance. Detection does not establish gene expression, viable pathogenic hosts, horizontal transfer, direct exposure or transmission. Likewise, co-occurrence of ARGs with microbial taxa or MGEs in the same sample should not be interpreted as genomic colocalization without contig-, plasmid- or MAG-level evidence. Knight et al. [16] and Hendriksen et al. [21] illustrate population-level surveillance applications, while ARGs such as blaNDM-1 and mcr variants detected in hospital and municipal wastewater indicate clinically relevant resistance determinants in environmental matrices [23, 28], as indicated in Table 3. Rout et al. [31] provides contextual evidence from Ganga River sediments rather than evidence of wastewater-treatment efficacy.

Across the included studies, ARG detection patterns varied with sampling context, treatment stage, microbial ecology, sequencing depth, database coverage and analytical method [32, 33]. Recurrent detection of β-lactamase, sulfonamide and aminoglycoside resistance determinants supports environmental surveillance, but does not establish dissemination or quantitative public-health risk [34, 35]. The quality appraisals identified both strengths and limitations, including variation in sampling design, sequencing depth, normalization, host assignment and reporting thresholds. These methodological limitations should be considered when interpreting consistency and generalisability across studies.

4. DISCUSSION

4.1. Principal Findings

A key difference throughout the body of evidence is that wastewater metagenomics shows the presence, relative abundance, and diversity of ARGs, and at times, the context of the ARGs in the genome. Read-based detection cannot tell if the resistance genes are expressed or if the microbial host is alive. Taxonomic co-occurrence can be used to infer ecological association but not necessarily host assignment, while contig-, MAG-, plasmid-, or integron-level linkage can be used to infer genomic context and potential mobility. Even these approaches do not demonstrate actual horizontal transfer, environmental exposure, or transmission to humans. Wastewater ARG detection, therefore, should be considered a measure of possible hazard and surveillance value, not necessarily as public health importance. Accordingly, the synthesis presented here describes recurring patterns across heterogeneous studies rather than estimating a common effect size across wastewater settings.

This review is not intended to prove that ARGs are present in wastewater, which is already proven, but to contrast the different ways in which shotgun-metagenomic approaches can characterize resistome diversity, quantify treatment stage change, elucidate ARG hosts and mobile genetic context, and provide cross-study surveillance. The review distinguishes metagenomic evidence from complementary targeted methods. It identifies key considerations where sequencing depth, annotation databases, normalization, and assembly strategy have an impact on interpretation and comparability of wastewater AMR results.

The most substantial contributions to the evidence base derive from global-scale investigations. Hendriksen et al. [21] analysed sewage from 79 sites in 60 countries, identifying substantial geographic variability in ARG abundance correlated with sanitation infrastructure and socio-economic indices. Similarly, Prieto Riquelme et al. [27] employed harmonized sampling to show cross-national differences in resistome composition. These studies exemplify the scalability of metagenomic wastewater surveillance and align with international frameworks such as the World Health Organisation’s Global Antimicrobial Resistance and Use Surveillance System (GLASS), which has begun developing environmental modules [32]. The United Nations Environment Programme’s 2023 AMR update also emphasized wastewater as a priority surveillance domain within a One Health strategy [33]. Both Hendriksen et al. [21] and Prieto Riquelme et al. [27] did not directly connect resistome data with patient-level clinical phenotypes, thus restricting their direct translational value.

4.2. Diversity and Distribution of ARGs

Regional and national studies added context on spatial, temporal and source-related variation. Hospital effluents were reported to have higher ARG abundance or diversity than community wastewater in some included studies, including Talat et al. [34]. Seasonal associations were reported in some datasets, but these should be interpreted cautiously unless supported by repeated longitudinal sampling and appropriate analysis. Environmental conditions such as rainfall, runoff and temperature may influence wastewater-associated resistome patterns, but detection patterns alone do not demonstrate transmission. Wastewater surveillance may therefore contribute complementary population-level information within a One Health framework, while proposed early-warning or regulatory applications still require validated thresholds, exposure assessment and clinical correlation [33].

Innovation was evident at the methodological level in a study reported by Davis et al. [15], where shotgun metagenomics was benchmarked against droplet digital PCR, using spike-in standards to quantify detection limits, and concluded that metagenomic data remain semi-quantitative unless calibrated. This finding aligns with broader evaluations that sequencing-based resistome profiling requires internal standards to avoid bias [31]. Authors compare high-throughput qPCR with metagenomics, revealing strong convergence for standard ARGs but divergence for rare ones, a discrepancy also reported in other comparative platforms [25]. Findings further demonstrated that culture enrichment improved detection sensitivity for low-abundance ARGs, though it distorted community structure and relative abundances. Together, these studies underline the methodological fragmentation of the field, an issue also highlighted by the OECD. It has called for harmonised pipelines and inter-laboratory calibration before wastewater AMR data can be used for policymaking.

Smaller-scale studies provided contextual evidence from receiving environments. Rout et al. [31] examined Ganga River sediments and reported spatial variation in ARGs and virulence factors; these findings do not measure wastewater-treatment efficacy. Fresia et al. [35] reported ARGs in Montevideo sewage and coastal waters, supporting environmental co-occurrence across connected settings but not demonstrating direct transmission, treatment-stage removal, or human exposure. Such receiving-environment studies should therefore be interpreted as contextual environmental evidence rather than as direct measures of WWTP performance or quantitative public-health risk.

Across the included studies, recurrent ARG classes and clinically relevant determinants were detected in wastewater-associated matrices. Their repeated detection supports the value of environmental surveillance, but does not establish expression, viability, host pathogenicity, horizontal transfer, or human exposure. Studies of the mobilome may identify genomic contexts compatible with potential mobility [36], but potential mobility should not be described as demonstrated dissemination without direct linkage evidence. Reproducibility is additionally affected by sequencing depth, annotation databases, primer sets, and normalization choices.

4.3. Influence of Wastewater-Treatment Stages

Wastewater surveillance may provide complementary information on population-level resistance patterns and environmental AMR dynamics, although its relationship with clinical AMR outcomes requires further validation. Several global initiatives, including the Joint Programming Initiative on Antimicrobial Resistance (JPIAMR) [37], have advocated the importance of developing standardized environmental AMR surveillance strategies. Despite this, many governments consider monitoring wastewater an experimental practice. Feasibility of wastewater-based AMR surveillance has been explored in different settings; however, evidence from low-resource environments remains limited and requires further validation [38, 39]. However, broader implementation requires evidence linking wastewater resistome measurements with clinical AMR indicators, together with validated thresholds and standardized interpretation frameworks.

4.4. Strengths and Limitations

This review has notable strengths. It systematically synthesized 15 studies across global, national, and local scales, applying JBI cross-sectional and prevalence tools for structured appraisal. By incorporating hospital-specific and community-level data and methodological comparisons, the review captures a broad spectrum of the resistome landscape. However, limitations are equally important. Studies included are very variable in design, sequencing depth, and analysis pipelines and cannot be meta-analyzed, forcing narrative integration. Geographic representation is still biased towards Europe and Asia, and Africa and Latin America are still under-represented. Furthermore, none of the studies directly linked wastewater ARG profiles with clinical AMR outcomes, limiting translational potential.

Methodological weaknesses further constrain interpretation. Short-read sequencing cannot always resolve plasmid contexts or gene hosts, reducing insight into ARG mobility. Variation in annotation thresholds and bioinformatics pipelines undermines cross-study comparability. Few studies accounted for hydraulic or sewer-network variables such as dilution or retention time, which may confound apparent ARG concentrations. These issues echo cautions raised in methodological reviews of metagenomic AMR surveillance [40]. A limitation of this review is the lack of a predetermined protocol and prospective registration. The eligibility criteria were applied uniformly at the screening phase; however, as the scope was revised to incorporate complementary molecular evidence and wastewater-impacted environmental matrices, there was potential for post-hoc methodological decisions to be made.

There are also limitations to the quality appraisal. JBI tools offer a framework for the design and reporting of an observational study, but they are not specifically aimed at environmental metagenomic studies. This may lead them to miss aspects of technical bias such as sequencing depth, extraction method, bioinformatic thresholds, normalisation, negative controls, and genomic host assignment. The supplementary metagenomic-specific appraisal was therefore used, not as an alternative, but to complement the JBI appraisal. The review scope was broadened during revision to include complementary molecular approaches and wastewater-impacted environmental matrices alongside shotgun-metagenomic wastewater studies. This enhances the representation of information from wastewater-associated AMR data, but added methodological variation and restrained the ability to make direct comparisons across studies based on different platforms. This review has a limitation in that there was no prospective, registered protocol. Eligibility criteria were applied uniformly during screening, but the scope was refined during the revisions, which could have led to potential post hoc methodological decisions.

Future work should prioritize standardized metagenomic protocols, including transparent sequencing-depth thresholds, annotation pipelines and quantitative calibration [37, 39]. Integration with clinical surveillance requires prospective correlation with antibiograms, validated thresholds and exposure-based interpretation. Implementation in low- and middle-income settings also depends on broader health-system and resource considerations [40]. Portable sequencing and near-real-time analysis are promising methodological developments, but wastewater metagenomics should currently be characterized as a potential complementary surveillance approach requiring further validation rather than as an established predictor of clinical AMR or individual risk.