Changes in antibiotic usage and ARG prevalence in the context of COVID prevalence in early (2020-21) and 2nd phase of pandemic (2021-22)
For better interpretation of the data, we divided the study period into two phases: the pandemic year during 2020-21 (April 2020 - March 2021) where the COVID-19 restrictions were in place, and the year during 2021-22 (April 2021 - March 2022) as a comparison.
Covid prevalence was higher in the 2nd stage of pandemic with statistically significant (p < 0.01) increase of 449.8% in the catchment area. In addition to higher prevalence, enhanced diagnostic methods, increased availability of test kits, and the implementation of widespread testing campaigns allowed for more comprehensive detection and reporting of COVID-19 cases in 2021-22. For antimicrobials, inconsistent trends for the individual drug usage between 2020-21 and 2021-22 were observed, as expected, with an overall increase in the total amount of community-wide antibiotic usage after the pandemic year. Antibiotic PNDI and PNDP showed an increase of 17.2% and 5.8%, respectively in 2021-22, in comparison to the previous pandemic year (Fig. S2). It should be noted that PNDI is a much more accurate indicator, representing actual (both primary and secondary care) antibiotic usage in any given catchment, as compared to PNDP, which primarily reflects primary care prescriptions (i.e. not accounting for lack of patient compliance or usage/excretion of antibiotics sourced outside catchment). However, PNDP uniquely obtained from England National Health Service prescription database provides an important reference point and another level of confidence for WBE PNDI outputs in the context of its future utilisation in regions and countries without freely available prescription/sales data. Of the 17 antibiotics targeted, clindamycin and flucloxacillin usage remained stable throughout the two years, while amoxicillin and clarithromycin were clearly affected by COVID-19 restrictions during the year 2020-21 with an average of 31.5% (p < 0.01) and 13.5% (p < 0.05) lower usage, respectively, followed by an increase in 2021-22.
Higher ARG abundance in 2021-22 (196% increase on average, Fig. S3) than the previous year was observed (excluding city D). Statistically significant correlations (0.48 ≤ r ≤ 0.79, p < 0.05) were found between the two towns (A and B), suggesting comparable community health risks posed by AMR in the same geographical region with smaller community size (< 30k). This also emphasises the applicability of WBE approach in monitoring regional AMR.
PNDI vs PNDP
Detailed information on the comparison of PNDI and PNDP for individual antibiotics, averaged across all four towns/cities studied, is shown in Fig. 1. Information on individual sites (PNDI vs PNDP) is provided in SI Fig. S4. It should be mentioned that for amoxicillin, clindamycin and sulfamethoxazole, the PNDImetabolite level aligned better with PNDP in comparison to the parent compound and therefore was used in the following analysis. In general, PNDI and PNDP datasets displayed comparable levels, with amoxicillin and clarithromycin showing the best concordance at a PNDI/PNDP ratio of 1.28 and 0.99, respectively (Fig. 2 and Fig S4). In addition, significant positive correlations (p < 0.01) between PNDI and PNDP were observed for the two drugs (Table S2). These results clearly indicate the applicability of WBE to estimate clinical prescription using amoxicillin and clarithromycin as indicators of community’s infection prevalence. While clarithromycin has been suggested as a suitable biomarker in previous studies,15,21 amoxicillin (the acid metabolite) should be considered as a representative β-lactam biomarker. In particular, amoxicillin and clarithromycin are among the most prescribed antibiotics for positive COVID-19 patients in primary care in England.8 Prescription of macrolides is often related to seasonality with an enhanced usage in winter.15,22 In agreement with this, statistical seasonal increases (p < 0.05) in clarithromycin and erythromycin usage in winter were observed in this study at all sites. Similarly, it is not surprising to see β-lactam antibiotic showing clear seasonal patterns with doubled usage in cold seasons. This could be attributed to both higher usages in winter and lower stability in summer for penicillins and cephalosporin classes.23
Clindamycin and erythromycin represented the antibiotics which showed greater discrepancies for PNDI and PNDP (Fig. 1&2). The variability may be attributed to their frequent topical application in comparison to oral prescription. For instance, clindamycin creams are applied topically, as a first-line acne treatment and its topical administration constituted approximately 50% of the total prescription on average, and similar prescribing volume was even reported in winter season 2019.23
Tetracyclines represented a complex drug family. From the environmental aspect, larger site like City D with longer wastewater residence time, oxytetracycline and tetracycline were detected in only 6.3% and 28.2% of samples, respectively, across the monitoring period, while these drugs have been prescribed frequently in clinical settings at City D. By contrast, for smaller sites, prescription rates for tetracycline were 4.2% and 50.0% at site A and B, respectively, while this drug has been detected at a 100% rate in wastewaters. This clearly suggests that the medicine is dispensed and consumed within different zones, or throughout different months, or is linked with other sources such as food-producing animals and pets. This emphasises, yet again, the unique value of WBE as a tool providing more accurate data on antibiotics usage (a unique spatiotemporal fingerprint) in tested communities. Comparable PNDI and PNDP levels for tetracyclines with less than 26% difference within the large catchment zone suggested that WBE and clinical data served as complementary tools.
Sulfamethoxazole and trimethoprim are often prescribed in combination (co-trimoxazole), typically at a ratio of 5:1. However, the corresponding prescription records suggested a much lower drug ratio (0.76:1), which is comparable to the calculated consumption ratio at 1.05:1. This indicated that human consumption/excretion of co-trimoxazole is not the main origin for the presence of sulfamethoxazole and trimethoprim in wastewater, and the consumption/excretion of trimethoprim only is present within the catchment. Trimethoprim is one of the commonly prescribed antibiotics in primary care and used to treat and prevent urinary tract infections (UTIs).
Another interesting finding is that better PNDI and PNDP alignment was evident during the pandemic year 2020-21 than the subsequent year. This phenomenon could be attributed to two primary factors. First, there is the aspect of patient compliance, i.e. patient’s willingness to adhere to prescribed antibiotics within the recommended treatment course. The second factor is related to movements within the catchment area. The lockdown measures implemented during the pandemic year led to reduced commuting to neighbouring cities, resulting in both the dispensing and consumption of the drug occurring within the same WWTP catchment.
Data triangulation - COVID-19, antibiotic usage and ARGs
A community’s infectious disease (expressed as COVID-19 infection rate); antibiotic usage (expressed as both prescription and consumption); and resistance levels (expressed as ARGs), were used for further data triangulation analysis. Trends of amoxicillin, cefalexin, clarithromycin and ciprofloxacin in both clinical and environmental settings aligned with the COVID-19 cases data (Fig. 3). In general, PNDI showed better alignment with COVID-19 trends compared to PNDP. This is expected as PNDIs show true community antibiotic usage. For COVID-19 positive patients, 23.4% of them were prescribed at least one antibiotic in primary care in England, among which amoxicillin was the most frequently prescribed antibiotic, followed by doxycycline, clarithromycin, nitrofurantoin, phenoxymethylpenicillin and co-amoxiclav.8,24 This has been reflected in the present study, for amoxicillin and its metabolite AMXa, and clarithromycin, both consumption and prescription showed statistically significant correlations with COVID-19 trends in the catchment area (rs ≤0.81, p < 0.05) (Fig. S5, 6). In fact, stronger correlations were seen with metabolites as metabolites reflect intake vs cumulative usage, including direct disposal as in the case of the parent antibiotic (Fig. S5). Cefalexin and ciprofloxacin were also among the antibiotics which showed positive correlations with COVID-19 infection rates (rs ≤0.85, p < 0.05) (Fig. S5, 6). Ciprofloxacin is a commonly used broad-spectrum antibiotic for the treatment of respiratory tract infection (RTI) and the prescriptions only showed slight increases without statistical difference during early stage of COVID-19 pandemic in England.6
COVID-19 infections tended to drive RTI-associated antibiotic usage during the early stages of the pandemic year, during which antibiotic prescribing occurred for suspected or proven COVID-19. By contrast, both prescription and consumption patterns started to rise earlier than the peak reported in COVID-19 cases. For example, amoxicillin showed increases in June and December 2021, while clarithromycin exhibited a rise in December 2020 and December 2021. This suggests a potential lag time between the reported outbreak of COVID-19 cases and the actual outbreak time point. These findings support the hypothesis that, when appropriate biomarkers are applied, WBE serves as an early-warning system for community's infectious diseases.
As observed in our previous study, 16 ARGs did not correlate as well as antibiotics with COVID cases (Fig. S7) and indeed correlations of ARGs with antibiotics were variable between different groups (Tab. S3). We reported similar observations in our longitudinal pre-pandemic 2018-19 study. 22 However, clear trends of lockdown driven decrease of antibiotics and ARGs due to lack of social interactions were apparent.
For a more informative interpretation of multi-correlations between consumption, prescription, COVID-19 infections and resistance genes, a total of 93 data points (monthly average) derived from raw data points of the 4 sampling sites were used to generate PCA plots. As shown in Fig. 4, the variation of ARGs could be partially explained by the variation in the prescription/consumption of associated antibiotics and Covid-19 infection rates with 57–77% of the variance explained in the dataset. Community antibiotic consumption correlated more closely to the resistance genes, as shown in Table S3, suggesting, yet again, that the WBE data could better reflect the environmental selective pressure on background microorganisms. As mentioned in previous section, community consumption was not necessarily linked to the clinical prescription, while the latter was usually impacted by the community infectious diseases level. To sum up, consumption contributed more to ARG residues in wastewater, while antibiotic prescription was more correlated with COVID-19 cases. Heatmaps showing correlations between Covid-19 infection rates and PNDI, PNDP, and ARGs in each sampling site are provided in Fig. S5-7 to further explore the potential spatial and demographic influence. In general, more correlations were seen among infection rates and antibiotic usage (PNDI & PNDP) in larger communities (> 100k inh, city C/D); while infection rates tended to correlate better with ARGs in smaller communities (< 30k inh, town A/B), especially for qnrS and blaCTX-M. This is possibly linked with the presence of hospitals treating COVID-19 cases in C/D catchments. However, the varying socioeconomic and demographic factors among the communities (e.g. town A/B having higher percentage of > 60 years old population vs cities C/D with large student population) albeit located in neighbouring catchment areas, should not be overlooked.
The impact of the COVID-19 pandemic on daily life has waned, but the valuable insights we have gained from tackling this global infectious disease collaboratively remain significant. The preventive and control measures as well as surveillance tools (lockdowns, social distancing, vaccination, rapid diagnostics, wastewater monitoring, etc.) implemented to address the global pandemic have provided an unparalleled opportunity to understand how to address the on-going ‘pandemic’ - antimicrobial resistance, which remains overshadowed globally. Wastewater surveillance will serve as an essential tool at the disposal of public health authorities that bridges the gap between public and environmental health beyond the context of COVID-19.25 In low-income countries in particular, where electronic medical records and detailed contextual data are not available to fulfil the requirement of big data to inform public health.26 WBE provides anonymous information on antibiotic consumption at population level. In the event that there is a delay in the release of official prescription and infectious disease data, information derived from community wastewater could be used near real-time to inform public health.
This is the first longitudinal study incorporating the four components: community incidence of COVID-19 infections, antibiotic prescription, antibiotic consumption, and wastewater antibiotic resistance genes as a whole. Nevertheless, it is important to acknowledge that each dataset, on its own, is subject to limitations. An online cross-sectional survey conducted in England and Scotland suggested a 64% lateral flow device (LFD) self-reporting rate, which is consistent with the Official figures estimating that 63% of the participants registered their LFT results with an official government agency.27,28 While self-report behaviour may vary considerably across regions, these data suggested that the official Covid-19 cases associated with LFT results could be underestimated. The incorporation of the metabolite dimension is crucial for a reliable WBE scenario. However, achieving accurate quantification of trace-level compounds in wastewater poses challenges due to factors such as stability (e.g., hydrolysis and biodegradation) and matrix effects, making the development of a robust analytical method essential. In the case of amoxicillin, the acid metabolite served as a better biomarker for the estimation of public exposure. More representative metabolite biomarkers should be included for future studies. Understanding the antibiotic resistome catalogue in wastewater, yet new to WBE domain, is crucial for linking with clinical resistance determinants and uncovering novel resistance genes.29 The utilisation of metagenomic sequencing will offer in-depth insights into multi-dimensional correlations and will be integrated into the next phase of our research.
In conclusion, changes in community behaviours associated with COVID-19 interventions have indirectly contributed to the reduction of community-wide AMR. In the prescription-consumption model established via WBE approach, amoxicillin and clarithromycin exhibited great potential as appropriate biomarkers complementary to clinical data for the surveillance of community’s infectious disease, antibiotic stewardship and patient compliance. This study strongly supported the One Health concept for addressing AMR potentially at regional and national scale. Future research should extend to larger catchment areas to establish more WBE pipelines to inform public health with an ultimate goal to help address AMR.