Comparison with other studies
In the present study, ASB was defined as growth of one species of Enterobacterales in counts of 105 CFU/ml or more and was present in 2.1% of women. For antimicrobial susceptibility testing we used the quantitative cut-off of ≥ 104 CFU/ml to define ‘significant growth’, as previously done in an international survey of antimicrobial susceptibility in uncomplicated urinary tract infections (21).
The ASB proportion in the present study was lower compared to some earlier studies from sub-Saharan Africa, citing proportions of 7% − 40% (22–25), but it was comparable to those found in other studies (26, 27). The presently lower proportions of ASB may be related to the stringent definition, i.e. including only Enterobacterales as significant organisms, whereas in most other studies with higher ASB proportions, Staphylococci represented a substantial number of cases (22–25).
AMR rates among the E. coli isolates obtained from urine of pregnant women were significantly lower compared to AMR rates of clinical isolates. Among the individual antibiotics, this difference was most apparent for ciprofloxacin, i.e. 16.2% for the urine isolates in pregnant women versus over 60% among clinical isolates. Likewise, proportions of ESBL producing and MDR isolates among E. coli from urine in healthy pregnant women were 3.2% and 5.2% versus 35.4%, and 60.4% respectively among clinical isolates. It is tempting to speculate that these differences reflect the use of antibiotics such as ciprofloxacin and third generation cephalosporins in the community setting.
The proportion of ESBL producers among the clinical E. coli isolates of the comparator studies (35.4%) was slightly lower compared to the 45% reported for sub-Saharan Africa in recent meta-analyses (28–30). Carriage rates of ESBL producing E. coli from stool samples ranged from 38% in Chad to 58% in the Central African Republic (31, 32). For the urine isolates obtained in pregnant women, AMR rates were lower compared to those found in other cross-sectional studies assessing ASB among pregnant women in sub-Saharan Africa. A study from Ghana from 2018 reported high resistance rates among E. coli to nitrofurantoin (35.4%), ciprofloxacin (48.8%), gentamicin (41.5%) and cefuroxime (32.9%) (33). Two studies from Nigeria (2007 and 2010) reported resistance rates among E. coli of approximately 20% against second generation cephalosporins, 40% against gentamicin and 20–70% against ciprofloxacin (23, 34), which was similar to results from a study performed in Uganda in 2010 (35). A possible explanation for the observed difference to our results is the fact that we strictly excluded participants with symptoms and signs of urinary tract infection.
E. coli from urine in healthy pregnant women as an indicator of AMR in the community
In the present cohort of over 6000 pregnant women attending ANC, only 1.5% declared symptoms suggestive of an urinary tract infection and only 1.7% had evidence of antibiotic use as demonstrated by urine analysis. The latter proportion is very low compared to 30% − 40% antibiotic use (based on parents’ declaration) among children suspected of invasive bacterial infection in three of the comparator studies (13–15). As such, the presently observed low AMR rates among E. coli isolates from the urine of healthy pregnant women tends to confirm our pre-study assumption, i.e. there may be a risk of overestimation of AMR rates when performing surveillance on selected clinical samples (36). However, other factors must be taken into account when comparing resistance rates between both groups. First, the species E. coli has distinct pathotypes displaying different degrees in pathogenicity and AMR (37); further genetic studies are planned to assess the pathotypes of the isolates from pregnant women versus those of the clinical samples. Further, in view of low Infection Prevention & Control in healthcare facilities, it is not excluded that part of the clinical isolates were belonging to a particular hospital-associated cluster.
Notwithstanding these considerations, it is tempting to forward E. coli in urine of pregnant women as a potential indicator for benchmarking, comparing and monitoring community AMR rates across communities over different countries and regions. Such community AMR data generate valuable information about the empiric choice of antibiotics in the local context (38, 39) but may also reflect the effect of AMR control measures. As shown at least in this study setting (and to be confirmed in other settings as well), pregnant women have limited illness and antibiotic use and are accessible through ANC clinics. As part of ANC clinics, urine is routinely sampled for dipstick analysis of glucose and protein and WHO recommends midstream urine culture for the diagnosis of ASB (40). The dipslide devices presently used were affordable (cost approximately 1 €/device) and user-friendly; they have a long shelf-life (6–9 months at room temperature) and allow for reliable inoculation on-site and subsequent transport to the laboratory. At the downside, there are the challenges of midstream-urine sampling (including contamination) and the reading of the colony counts on the dipstick devices as discussed above. Moreover, the proportion of significant growth is low. Leukocyte esterase and nitrite analysis (incorporated in most urine dipsticks) can be used as a screening tool to select samples for culture (40) but in the present study they were not very accurate to predict growth; further research for a reliable biomarker predicting growth is recommended.
Limitations and strengths
As noted above, reading of colony counts on dipslide devices tended to be subject to interpretation and this may have impacted the classification of non-significant growth, significant growth and ASB. However, our results showed that pathogen and AMR profiles were similar between E. coli from the latter two groups. Second, despite well-designed instructions and training, 8.3% of the samples were contaminated, probably related to the less stringent urine sampling in the context of a ANC compared to clinical care. Third, in retrospect, we realized that the GPA system of parity had not been fully understood by all study nurses, leading to possible too low reported numbers of nullipara compared to primipara. Additionally, there were missing data from 36 consecutive participants (0.6% of all included patients) at one of the healthcare centers. Strengths included the systematic methods used to perform this study, with high numbers of participants included and a consistent work-up by a small team of nurses and laboratory staff. The definitions for ASB and contamination were stringent, adding to the robustness of data.