Authors: Jane L. D. Currie, Catherine P. James, Jennifer L. Rohn, Anna L. David
Categories: Research, Urinary tract infection, pregnancy, urinary culture, microbiology, cytokines, urothelial cells
Source: BMC Pregnancy and Childbirth
Authors: Jane L. D. Currie, Catherine P. James, Jennifer L. Rohn, Anna L. David
Abdominal pain in pregnancy may be caused by urinary tract infection (UTI), which is associated with preterm birth and pyelonephritis. Standard urine culture is insensitive; alternative tests, previously studied in chronic UTI, may improve UTI diagnosis in pregnancy. We hypothesised that women with abdominal pain in pregnancy may have urinary pathology not detected by standard tests.
This single-centre, prospective case-control observational study compared patients presenting with abdominal pain after 14 weeks’ gestation, using both standard and alternative tests, against gestation-matched patients presenting with another non-abdominal pain acute problem, and asymptomatic patients attending routine antenatal care. Urine samples were a clean-catch midstream void. Standard tests were urinary dipstick, microscopy and culture. Alternative tests (1) objective symptoms inventory; (2) quality of life assessment (EQ-5D-5 L); (3) fresh unspun urine microscopy; (3) urinary ATP; (4) enhanced sediment culture; (5) urinary IL-6, IL-8 and lactoferrin; (6) urothelial cell analysis. Non-parametric statistical methods were used.
Pregnant women whether with abdominal pain (n = 50), other acute hospital presentations (n = 58) or attending routine antenatal care (n = 51), had symptoms of UTI with scores equivalent to those seen in patients with chronic UTI. Women presenting acutely with abdominal pain did have more pain symptoms, nocturia, and lower quality of life scores, with a different distribution of bacteria using enhanced urine culture, but they had equivalent rates of positive standard urine culture compared to those with no pain or those in routine antenatal care. Subsequently they were more likely to be diagnosed with a UTI and prescribed antibiotics. Urinary symptoms and markers of urinary pathology were prevalent in all groups, but different in those clinically diagnosed with UTI.
Our study shows that current standard testing for UTI in pregnancy, and understanding of what is normal and abnormal, is inadequate. The use of alternative tests, well validated in a chronic UTI research programme, raises questions about current urine testing practice in pregnancy and the assumptions that drive them. Further research should examine these tests in different pregnancy contexts, to determine if they can enhance UTI diagnosis, better guide management of urine pathology in pregnancy and improve pregnancy outcomes.
The online version contains supplementary material available at 10.1186/s12884-026-08835-6.
Urinary tract infection (UTI), including in pregnancy, is traditionally diagnosed by urine culture, with a 70-year-old definition upon which nearly all practice and research rests [1]. Recent studies find that there is a urinary microbiome, and that standard culture may miss bona fide infections [2–4]. This shifts our understanding of the diagnosis, pathophysiology and potential harms of UTI in pregnancy, such as pyelonephritis, preterm birth (PTB) and the need for antibiotics.
Abdominal pain in pregnancy is common. Diagnosis is difficult especially where the differential range from mild Braxton Hicks uterine contractions with limited clinical impact, to severe impact for the mother and fetus, such as threatened preterm labour. Some pregnant women with pain may have urinary pathology not detected by standard tests. Abdominal pain may also serve as a non-culture-centric paradigm for exploring infectious urinary pathology in pregnancy, without assumptions about what defines UTI as the presenting symptom of abdominal pain will likely include some women with either frank or subclinical UTI.
In pregnancy, UTI has been associated with pyelonephritis, leading to severe maternal morbidity including sepsis and acute respiratory, renal and cardiac failure and death. UTI is also independently associated with preterm birth (PTB) [5], particularly between 33 and 36 weeks’ gestation [6]. Asymptomatic bacteriuria (ASB) occurs in 2–10% of pregnancies, and increases the risk of pyelonephritis [7], but it remains controversial how effective screening is in preventing pyelonephritis and PTB, especially in low-risk women [8]. Yet accurate UTI diagnosis is imperative because of the implications of antibiotics for the mother (adverse effects and anaphylaxis), the fetus (increased risks of harm following antibiotics for preterm labour with intact membranes [9]; population-level associations between maternal UTI treatment and cerebral palsy [10] and childhood asthma [11]), and the global issue of antimicrobial resistance [12]. Common diagnoses also have economic implications.
There are many causes of abdominal pain in pregnancy [13], some pregnancy-specific (e.g. threatened preterm labour, placental abruption) and some from abdominal or pelvic organ systems. It is not always possible to diagnose the cause [14, 15]. One antenatal study diagnosed 14.2% of abdominal pain patients with UTI [16], using a non-standard UTI definition, but did not report obstetric outcomes. Unexplained acute abdominal pain is also common in emergency presentations outside pregnancy [17]. Little is known about the association of UTI with abdominal pain in pregnancy.
UTI diagnosis is traditionally defined by growth of a single uropathogen at ≥ 10^5^CFU/ml^1^, with mixed growths and epithelial cells on microscopy assumed to represent contamination [18, 19]. However, recent studies using genomics and/or enhanced culture techniques have revealed an extensive normal bladder microbiome in both symptomatic patients and controls [20–24], even in participants with negative standard cultures [2, 25]. Furthermore mixed growths can be associated with infection [26], with polymicrobial infection linked with higher nosocomial mortality infection [27] and more invasive uropathogens [28]. There is also a urinary microbiome in pregnancy [29], though this has been less studied [30]. Unfortunately, it is difficult to distinguish pathogens from commensals [23, 24, 31–33], a fact that will complicate diagnosis until sufficient advances are made in our understanding of the comprehensive virulence determinants in bacterial genomes.
The 10^5^ CFU/ml diagnostic threshold for ASB [34] or UTI is attributed to Kass, but his description addressed the differentiation of contamination from true bacteriuria [35]. Kass accepted there could be significant bacteriuria at lower counts [36]. In acutely dysuric women, 10^2^ CFU/ml may be sufficient to diagnose UTI [37–39]. The threshold of 10^2^ CFU/ml, with acute dysuria, may also be sufficient for UTI diagnosis in pregnancy [40, 41]. Around one third of pyelonephritis in pregnancy cases arise in patients without ASB using current criteria [42].
In pregnancy, Group B Streptococcus (GBS, also known as Streptococcus agalactiae) bacteriuria is a particular concern, owing to the association between maternal GBS colonisation and neonatal infection [43] as well as PTB [44]. GBS bacteriuria at any level has been associated with PTB [45]. Untreated GBS bacteriuria in early pregnancy at lower colony counts was associated with chorioamnionitis, with histological grade of chorioamnionitis correlated with the bacteriuria colony count [46]. The role of low-count bacteriuria in pregnancy needs to be better understood. Outside of pregnancy, although shed urinary epithelial cells traditionally were assumed to represent skin flora contamination [47, 48], the urothelium is now known to undergo physiological shedding which is increased on infection [49–53]. The urothelial origin of these epithelial cells has been confirmed by their presence in catheter urine specimens and by analysing expression of uroplakin-III [49], which is highly specific for superficial urothelial umbrella cells [54]. Some non-pregnant studies have even used them as a proxy for infection [55–57]. One study in pregnant women comparing three sampling methods (midstream, first-morning, and clean-catch) [58], identified epithelial cells present in 50–60% of all sample types.
UK NICE guidelines specify that the urine dipstick should not be used for diagnosis in pregnant women [59], and indeed, studies have shown they perform poorly [20, 60, 61].
An alternative paradigm for considering UTI is the damage-response framework [62] in which the host response to the organism rather than the organism itself is of most importance. Host-response factors include urothelial cell shedding. The most well-known response is pyuria, the presence of white cells in the urine, which is diagnostically useful in chronic UTI where other methods including standard culture are not [63]. Other potentially diagnostically useful host biomarkers include urinary adenosine-5′-triphosphate (ATP) [64–69] and urinary cytokines, chemokines and antimicrobial peptides which form part of the host immune response, especially urinary IL-6 (cytokine) [70–72], urinary IL-8 (chemokine) [73, 74] and urinary lactoferrin [75–77].
Here we address the research do women with abdominal pain in pregnancy have urinary pathology not detected by standard tests? We prospectively compared urinary pathology in pregnant women presenting acutely to hospital with abdominal pain (“pain cases”), using both standard and alternative tests, against women presenting with another acute non-abdominal pain problems (“acute controls”), and women attending routine antenatal care (“normal controls”). We used standard methods, including urinary dipstick, microscopy and culture, alongside alternative point-of-care testing, laboratory and culture methods, and questionnaires to explore our hypothesis.
This prospective case-control study compared urinary pathology in women in the second and third trimester of pregnancy presenting to hospital acutely with abdominal pain in pregnancy (“pain cases”), women presenting to hospital with other acute non-abdominal pain problems (“acute controls”) and women attending for routine antenatal care (“normal controls”) using standard and alternative tests.
Fig. 1Diagram to summarise study design and methods for the Abdominal pain study. MFAU: Maternal Fetal Assessment Unit, an obstetric triage day unit that sees women from 14 weeks of gestation
Recruitment occurred in Elizabeth Garrett Anderson Wing, University College London Hospital (UCLH), UK, from 23/10/2014 until 14/10/2015 inclusive.
Pain cases and acute controls were recruited in the Maternal Fetal Assessment Unit (MFAU) where women can attend from 14 weeks of gestation. Normal controls were recruited in midwife antenatal clinics or obstetric ultrasound department.
Pregnant women (≥ 18 years of age, ≥ 14 weeks gestation) presenting acutely to MFAU or attending a routine antenatal appointment.
Women not booked for their pregnancy at UCLH.
Allocation as pain cases or acute controls was based on the patient’s presenting complaint on arrival to MFAU and not on post-triage categorisations. This was to avoid preconceptions about the nature of pain in urinary pathology and to recruit a broad group of patients.
There was no exclusion for obstetric or medical complications other than no current problem requiring review in MFAU.
Acute controls and normal controls were gestation-matched to pain cases, plus or minus two weeks’ gestation, aiming for one acute control and one normal control for each pain case.
Gestation-matching was chosen as gestation was a possible confounder. Demographic and risk factor data was subsequently compared between the three groups to assess for the risk of recruitment bias.
Ethical approval was obtained by an amendment under the umbrella of the Cervical Length, InflaMmation and Bacterial species and preterm birth (CLIMB) study protocol. This received ethical approval from the Joint UCL/UCLH Committees on the Ethics of Human Research (REC 09/H0714/66). Written informed consent to participate was obtained from all participants. This study adhered to the Declaration of Helsinki.
Outside of use in a specialist chronic LUTS clinic, alternative tests had not been approved for clinical care, and many were analysed much later from frozen samples, hence were not used to direct treatment. Where symptoms or standard clinical urine tests suggested infection, or carriage of Group B Streptococcus (GBS), patients were treated according to standard hospital guidelines.
Demographic, obstetric and medical data was prospectively collected. Prior to review by a clinician (obstetrician or midwife in MFAU), participants were asked if they thought they had a urinary tract infection currently responding either no/unlikely, don’t know/possible, and yes/likely.
Participants completed the Artemis LUTS questionnaire [78] and EQ-5D-5 L quality of life questionnaire [79]. These were pseudonymised with a study ID for later analysis in batches.
A voided urine sample was collected using a clean catch mid-stream technique via a study-specific kit with wide-neck collection vial. Sample handling and processing is outlined in Fig. 2. Samples underwent standard analysis with urine dipstick and were sent to the hospital laboratory for routine urine microscopy and culture.
Urine samples underwent alternative analysis with point-of-care ATP test, fresh unspun microscopy, enhanced sediment culture, ELISA for IL-6, IL-8 and lactoferrin, and epifluorescence microscopy for epithelial cells and, for a subset of samples, for uroplakin-3 staining.
Fig. 2Flowchart of sample handling and processing. Abbreviations: MSU = mid-stream urine specimen, ATP swab = urinary adenosine-5′-triphosphate point of care test, API = Analytical Profile Index, ELISA = Enzyme-linked immunosorbent assay
Repeated freeze-thaw cycles were avoided. Laboratory analysis of frozen samples continued until April 2018.
The urine dipsticks used were Multistix^®^8 SG reagent strips (Siemens, Munich, Germany) and were assessed visually according to manufacturer’s instructions.
Routine laboratory microscopy, culture and sensitivity (MC&S) was undertaken in the local NHS laboratory using the Mast Uri^®^ robotic system.
Standard NHS microscopy cell count results were reported as not seen / very scanty / scanty / moderate / numerous. These were summarised for this study as not seen / present, with all other categories collapsed into present.
The NHS laboratory reported a positive culture as a single urinary pathogen of ≥ 10^5^ CFU/ml colony count. Other culture results screening culture negative; no significant growth; mixed growth of uncertain significance. Antimicrobial sensitivity results were not analysed in this study.
The Artemis LUTS inventory was used [78]. This is a questionnaire assessment of urinary frequency and nocturia, incontinence, and 34 other LUTS, grouped into four symptom domains of stress incontinence (7), urgency (12), voiding (8) and pain (13). Symptoms were recorded as binary variables (present or absent). Patients were asked to record symptoms recalled from the preceding few days. Scores were summed for each of the four symptom groups. Threshold reference values have not been determined. Population estimates of mean and standard deviation are available based on previous research in non-pregnant people [78].
QoL assessment used the EQ-5D-5 L questionnaire [79], performed as per manufacturer’s instructions.
A drop of freshly voided, unspun urine was transferred to a Neubauer haemocytometer counting chamber using a disposable Pasteur pipette (Sigma-Aldrich, Gillingham, UK) according to standard technique, to give a count per microlitre [80]. Trained researchers counted white blood cells, red blood cells and epithelial cells, using an Olympus CX41 light microscope (x400; Olympus).
Fresh unspun urine microscopy is used for diagnosis and surveillance in a chronic LUTS service [81]. In this setting, normal white blood cell count (WBC) is 0 WBC/µl with pyuria defined as ≥ 1 WBC/µl [82], with a further threshold at ≥ 10 WBC/µl corresponding to standard reference values [83].
ATP was measured using the LuciPac pen system (Kikkoman), according to the manufacturer’s instructions. The luminometer produces a digital reading, in relative light units (RLU).
Sediment culture protocol was described previously [2]. Briefly, specimens were centrifuged, supernatant removed, the sediment resuspended in normal saline and serial dilutions performed. The resuspended sediment and dilutions were cultured on chromogenic agar plates at 37 °C for 24 h. Following incubation, organisms were identified using colour, morphology and standard biochemical tests (API strips, bioMérieux) and quantified.
Enzyme-linked immunosorbent assay (ELISA) was used to measure urinary IL-6, IL-8 and lactoferrin concentrations [77, 84, 85]. Commercial kits were used, following manufacturers’ instructions. Protein concentrations were measured using Thermo Fisher Pierce™ Coomassie (Bradford) Protein Assay Kit (23200), following manufacturers’ instructions. It has been suggested that urine biomarker studies should report both adjusted and unadjusted biomarker concentrations to facilitate detection of appropriate biomarkers and enable comparison between studies [86]. Accordingly, results were reported with and without normalisation.
The protocol for epifluorescence microscopy for bacteria-associated epithelial cells and uroplakin-3-associated epithelial cells was described previously [49]. Briefly, 80µl of specimen was centrifuged creating a visible disc of urinary particulate on the slide, fixed and washed for incubation at room temperature. Wheat germ agglutinin (WGA) conjugated to Alexa Fluor-488 (Invitrogen) was used to label the cell membrane to aid cellular identification and demarcation, and DAPI (4’’,6-diamidino-2-phenylindole) (Sigma-Aldrich) was used for fluorescent counterstaining of host and pathogen DNA.
For uroplakin-3 staining cells were permeabilised, primed with goat serum, and incubated with the primary antibody - a 10 dilution of primary antiuroplakin-III mouse monoclonal antibody (Progen Bioteknik) in 1 goat serum. After washing, the slide was incubated with the secondary antibody with counterstain – a 250 dilution of goat anti-mouse secondary antibody conjugated to Alexa Fluor-555 (Invitrogen), DAPI and goat serum. Cells were counted using 100x magnification on a fluorescent microscope using appropriate fluorescent filters. Alexa fluor 488 (WGA) excites at a wavelength of 495 mm and emits at 519 mm, giving a green appearance to cell membranes under fluorescence. Alexa fluor 555 (Uroplakin-3) excites at a wavelength of 555 nm and emits at 565 nm, giving a red appearance to urothelial cell membranes. DAPI excites at 360 nm and emits at 460 nm giving a blue appearance to nuclei and bacteria.
Pseudonymised data was stored in an Excel spreadsheet and decanted into SPSS. Data analysis used Excel, SPSS version 26, with graphs created in Prism version 8.
Online calculators were used to calculate confidence intervals of proportions (http://vassarstats.net/prop1.html) according to the Newcombe method, using continuity correction [87]. Online calculators were used for ELISA calculations (myassay.com). Venn diagrams were constructed using a web-based tool (http://www.interactivenn.net) [88].
Normality was assessed using the Kolmogorov-Smirnov test. From previous work it was anticipated that most variables would be best described by non-parametric distributions, and this was the case for all variables except age. Thus, non-parametric statistical methods were used throughout. All statistical tests were two-sided and at the 95% level of confidence.
For all studies, continuous variables were described using median, 95% confidence intervals of the median, and sometimes range. Categorical variables were described using proportion and 95% confidence intervals (CI) of the proportion. Some continuous variables were converted into dichotomous categories, such as white cell count; 0/mm^3^ / > 0/mm^3^. Some categorical variables were similarly treated. Log transformation was used for cell and microbial counts. Before log transformation zeros were changed to 1. Between-group differences were compared using the Kruskal-Wallis test and the 95% CI about the median were used. Categorical variables were compared using Chi Square (χ^2^), or Fisher’s Exact (FE) where assumptions were not met.
When commencing this work, pyuria had been shown to be the most useful currently available marker in the field of chronic UTI, in that it correlates with symptoms and symptom response to treatment [2, 49, 83]. Because of concerns about the utility and validity of standard culture, and the different paradigm of UTI for this research, we used white cell count as the primary outcome for the sample size calculation, as with other studies from this group at the time [75, 89].
Using white cell count as the primary outcome, we calculated a sample size of 40 in each group which would have more than 80% power to detect a clinically significant effect, based on log10 WBC counts. In anticipation of using non-parametric methods, and loss to follow-up, we aimed to recruit 50 women presenting with abdominal pain, 50 presenting acutely without pain, and 50 low risk routine antenatal clinic attendees.
159 patients were 50 pain cases (patients presenting with abdominal pain after 14 weeks gestation), 58 acute controls (gestation-matched patients presenting with another non-abdominal pain acute problem), and 51 normal controls (gestation-matched asymptomatic patients attending for routine antenatal care).
Median gestation at recruitment was 191 days (range 98–287 days). Median age at recruitment was 33.5 years (range 19.5–43.6 years). 49/147 (33.3%) of participants were of non-white ethnicity. There was no difference between groups for these parameters. 41/48 (85.4%) of pain cases lived in the most deprived top five deciles, compared with 38/57 (66.7%) of acute controls and 32/51 (62.7%) of normal controls (p = 0.029, χ^2^).
89/159 (56.0%) of participants were nulliparous. Median BMI was 23.5 (95%CI 22.8–24.3). There were 7/157 (4.5%) current smokers, and 5/159 (3.1%) twin pregnancies. There were no between groups differences in these parameters.
PTB risk factors (excluding multiple pregnancy) were present in 14/159 (8.8%) with no difference between groups (p = 0.247). These included previous PTB/late miscarriage (n = 8), Large Loop Excision of the Transformation Zone (LLETZ)/cone biopsy (n = 5), and uterine anomaly (n = 1).
A previous (patient-defined) history of UTI was present in 17/159 (10.7%), with no difference between groups (p = 0.139). Six patients reported recurrent UTI outside pregnancy and one patient reported this during pregnancy.
Supplementary Table 1 compares obstetric and neonatal outcomes for the three groups. Delivery details were missing for 3 participants (2 pain cases, 1 normal control). In the following data, p-values refer to between-group comparisons.
Median gestational age at delivery was 276 days (39 + 3 weeks and days, 95% CI 274–280 or 39 + 1 to 40 + 0, p = 0.067). Median number of days from recruitment to delivery was 86 days (95% CI 69–97, p = 0.220). There were 5/48 (10.4%) PTBs in the pain group, compared with 3/58 (5.2%) acute controls and 2/50 (4%) normal controls (p = 0.462). Spontaneous PTBs occurred in 2/47 (4.3%) of pain cases (one PTB could not be categorised as spontaneous or iatrogenic), 2/58 (3.4%) acute controls and 1/50 (2.0%) normal controls (p = 0.864).
There were no differences in rates of induction of labour, mode of birth, infant sex, birthweight, or Apgar scores. Modes of birth were comparable to the institutional rates at that time [90].
There were no admissions for antenatal sepsis or pyelonephritis. Intrapartum sepsis, or suspicion of sepsis, occurred in 10/154 (6.5%) (p = 0.402), postpartum sepsis, or suspicion of sepsis in 15/154 (9.7%) (p = 0.715) and neonatal sepsis, or suspicion of sepsis in 16/150 (10.7%) (p = 0.094). Babies were admitted to neonatal care in 17/154 cases (11.0%, p = 0.054). This includes babies in transitional care, such as those having intravenous antibiotics.
Figure 3(A) compares presenting problems for pain cases and acute controls. All pain cases had pain as a presenting problem. In addition, 25/48 (52.1%) reported additional symptoms, the most common being uterine contractions (6/48, 12.5%), vaginal bleeding (6/48, 12.5%), urinary symptoms (4/48, 8.3%) and back pain (4/48, 8.3%). The commonest presenting symptoms for acute controls were reduced fetal movements (17/54, 31/5%), suspected ruptured fetal membranes (11/54, 20.4%) and vaginal bleeding (11/54, 20.4%). Urinary symptoms accounted for 2/54 (3.7%).
Fig. 3A Comparison of presenting symptoms for pain cases and acute controls. B Comparison of final clinician diagnosis for pain cases and acute controls
The only between-groups symptom differences were suspected ruptured fetal membranes, reduced fetal movements (both more common in acute controls) and back pain (more common in pain cases). Following Bonferroni correction, only reduced fetal movements remained significant (p < 0.005).
Final clinician diagnosis was derived from review of the case-notes. Diagnoses were missing from medical records for 8/50 pain cases (16%) and 6/58 (10.3%) acute controls.
Diagnoses were coded thematically, and cases counted. In some cases, multiple diagnoses were given; these have been coded separately. Some diagnoses were expressed as a possibility (such as “?UTI”) and no distinction is given here between possible, probable and certain diagnoses.
Figure 3(B) compares the final diagnoses for pain cases and acute controls. Among pain cases, the commonest diagnoses were UTI (18/42, 42.9%), musculoskeletal or ligamentous pain (15/42, 35.7%) and ‘nil abnormality / reassure’ (15/42, 35.7%). Other diagnoses included heartburn (2/42, 4.8%), and labour, vaginal bleeding (no cause identified), Braxton Hicks (non-labour uterine tightenings), ruptured fetal membranes and a “show” (vaginal passage of the endocervical mucus plus), all in 1/42 cases (2.4%).
Among acute controls, the commonest diagnoses were ‘nil abnormality / reassure’ in 32/52 (61.5%) and vaginal bleeding (cause identified) in 5/52 (9.6%): bleeding from cervical ectropion in three, low lying placenta and subchorionic haematoma. Other diagnoses included obstetric cholestasis and ruptured fetal membranes in 3/52 (5.8%) cases each; UTI, vaginal bleeding (no cause identified) and thrush in 2/52 (3.8%) cases each; and migraine, dehydration, labour, anaemia and Bell’s palsy in 1/52 (1.9%) cases each.
There were significant differences between the two groups in terms of musculoskeletal / ligamentous pain (p < 0.001) and UTI (p < 0.001). All other diagnostic labels were not statistically different between the groups after Bonferroni correction (p < 0.0029).
Table 1 compares intent to prescribe antibiotics for pain cases and acute controls at this presentation. Of those for whom data was available, 14/43 (32.6%) pain cases were prescribed oral antibiotics, versus 1/52 (1.9%) acute controls. A further 4/95 (4.2%) were advised for antibiotics if MSU culture was positive. 5/95 (5.2%) received other antimicrobial treatment, which included oral erythromycin for Preterm Prelabour Rupture of the Membranes (PPROM), two patients with antifungal treatment for vaginal candidiasis, antibiotics for term ruptured membranes with GBS perineal carriage, and one patient who was already on antibiotics for suspected UTI but had been advised to increase the dose. The prescription of antimicrobials at time of review did differ between groups (p < 0.001).
Table 1Intent to prescribe antimicrobials at time of presentationPrescribed oral antibiotics at time of reviewPain cases (n = 43)Acute controls (n = 52)14/43 (32.6%)1/52 (1.9%)Plan for antibiotics if culture positive2/43 (4.7%)2/52 (3.8%)No antibiotics prescribed26/43 (60.5%)45/52 (86.5%)Other1/43 (2.3%)4/52 (7.7%)Fisher’s Exact p < 0.001
Prior to clinician review, participants were asked if they thought they had a UTI (Table 2). Of those for whom data are available, 12/36 (33.3%) pain cases thought they might have, or did have, a UTI, compared with 2/50 (4%) acute controls (p < 0.001).
Table 2Participant belief about whether they think they have a UTIPatient belief about UTIPain cases (n = 36)Acute controls (n = 50)Unlikely UTI (n = 72)24/36 (66.7%)48/50 (96%)Possible UTI (n = 10)9/36 (25%)1/50 (2%)Likely UTI (n = 4)3/36 (8.3%)1/50 (2%)Fisher’s Exact p < 0.01
Table 3 shows that of the 4 patients who thought they had a UTI, 3/4 (75%) were diagnosed with UTI by the clinician; of the 67 who did not think they had a UTI, 63/67 (94.0%) were not diagnosed with UTI.
Table 3Comparison of clinician diagnosis of UTI versus patient diagnosis of UTIPatient diagnosis of UTI UTI ( n ** = 4)**
Possible UTI ( n ** = 9)**
No UTI ( n ** = 67)** Clinician diagnosis of UTIUTI (n = 14)374No UTI (n = 66)1263Fisher’s Exact p < 0.001
To aid in understanding clinical reasoning regarding LUTS, clinical notes were reviewed, and documentation of UTI symptoms was thematically analysed (See Supplementary Table 2).
UTI symptoms were documented in 20/34 (58.8%) of pain cases compared with 11/49 (22.4%) of acute controls for whom data were available (p = 0.001, χ^2^). Where symptoms were documented, LUTS were present in 7/20 (35%) pain cases, and in 3/11 (27.3%) acute controls (p = 0.66). Documentation was mostly brief, particularly for negative findings (for example, “no urinary symptoms”).
UTI symptoms were documented for 11/16 (68.8%) of those who were diagnosed with UTI compared with 18/67 (26.9%) who were not diagnosed with UTI (p = 0.003, χ^2^, see Supplementary Table 3).
There were no between-group differences in booking urine cultures, or urine cultures throughout pregnancy.
Comparing booking urine cultures, there were 76/156 (48.7%) screening culture negative, 32/156 (20.5%) non-significant growth, 25/156 (16.0%) mixed growth, and 20/156 (12.8%) positive, with no difference between groups in their results (p = 0.517).
554 standard cultures were reported throughout the 159 pregnancies, including booking cultures. Of these, 271/554 (48.9%) were screening culture negative, 130/554 (23.5%) showed non-significant growth, 110/554 (19.9%) showed mixed growth of uncertain significance, and 43/554 (7.8%) were positive. This was not different between groups (p = 0.261).
The median number of standard urine cultures recorded per patient was 4 (95% CI 3–4) for pain cases, 3 (95% CI 2–3) for acute controls, and 3 (95% CI 2–3) for normal controls (between-group difference, p = 0.009). 33/159 (20.8%) participants had a positive urine culture at some point during the pregnancy, which did not differ between groups (p = 0.402).
Standard microscopy and culture results are described in Table 4.
Table 4Standard microscopy and culture resultMeasurePain cases (n=50)Acute controls (n=58)Normal controls (n=51)SignificanceProportion with microscopy performed (%, n=97)(n=50)(n=58)(n=51)χ²=36.089 p<0.00136/5018/5843/51(72%)(31.0%)(84.3%)White cells (any seen) (%, n=68)(n=36)(n=18)(n=43)χ²=5.097 p=0.07823/3610/1835/43(63.9%)(55.6%)(81.4%)Red blood cells (any seen) (%, n=16)(n=36)(n=18)(n=43)χ²=11.8635 p=0.0039/366/181/43(25.0%)(33.3%)(2.3%)Epithelial cells (any seen) (%, n=94)(n=36)(n=18)(n=43)χ²=0.4606 p=0.79435/3617/1842/43(97.2%)(94.4%)(97.7%)Culture positive >10^5^CFU/ml (%, n=10)(n=50)(n=56)(n=51)χ²=1.5757 p=0.45482/503/565/51(4%)(5.4%)(9.8%)
Microscopy was performed for 97/159 (61.0%) participants, which differed between groups (p < 0.001), with 36/50 (72%) pain cases, 18/58 (31.0%) acute controls and 43/51 (84.3%) normal controls.
White blood cells were reported in 68/97 (61.0%) participants with no difference between groups (p = 0.078). Red blood cells were reported in 16/97 (16.5%), which ranged from 1/43 (2.3%) normal controls to 9/36 (25%) pain cases and 6/18 (33.3%) acute controls (p = 0.003). Epithelial cells were reported in 94/97 (96.9%) participants, which did not differ between groups (p = 0.794).
Standard urine culture was performed in 157/159 (98.7%) participants and was positive (with > 10^5^ CFU/ml) in 10/157 (6.4%) participants, with no between-group difference (p = 0.455).
Detailed standard culture results are described in Supplementary Table 4, for illustrative purposes only, given the small numbers. Of the 10 positive results, 5/157 (3.2%) were Escherichia coli, 2/157 (1.3%) were Enterococcus spp., 2/157 (1.3%) were Streptococcus agalactiae, and 1/157 (0.6%) were Citrobacter koseri. Of the non-positive results, 98/157 (62.4%) were reported as screening culture negative, 26/157 (16.6%) as no significant growth, and 12/157 (14.6%) as mixed growth of uncertain significance. Rates of these categories did not differ between groups (p = 0.191, Fisher’s Exact test).
Standard culture results were compared with the antibiotic prescription at the time of the visit. 3/4 (75%) of those with positive cultures were not prescribed antibiotics whereas 7/60 (11.7%) of those with negative cultures were prescribed antibiotics.
Urinary dipstick results are described in Supplementary Table 5. In line with typical antenatal practice, dipstick test was recorded positive if there was any colour change reaction to leucocytes, protein or nitrite as these may all trigger sending a urine sample for standard culture. Thus 109/159 (68.6%) participants showed dipstick positive with no difference between groups (p = 0.110); 1/159 (0.6%) was dipstick positive for nitrite.
Table 5 compares overall dipstick result with clinician diagnosis of UTI. There were no statistically significant differences between groups with respect to leucocyte esterase, blood, nitrite, protein or glucose or a combination using dipstick positive for protein / nitrite / leucocyte esterase. There was an association between dipstick and clinician diagnosis of UTI.
Table 5Comparison of overall dipstick result versus clinician diagnosis of UTIClinician diagnosis of UTIUTINo UTITotalPositive (any of protein / nitrite / leucocyte)194261Negative13233Total207494χ^2^ = 10.108 p = 0.001
Participants were asked to complete the Artemis LUTS inventory [78]. All participants filled in the questionnaire although not all completed all questions.
In summary, there were small between-group differences in 24-hour frequency, and differences in pain score appropriate to a group complaining of pain.
151/159 (95.0%) participants reported at least one urinary symptom with no between-group differences (p = 0.282). Figure 4 shows the distribution of total LUTS scores, which was wide. There was no significant difference between groups (p = 0.074).
Fig. 4Total LUTS scores according to study group (showing median and 95% confidence intervals)
Median daytime frequency was 7.5 (95% CI 6.5–7.5, range 2.5–19.5), with a small significant difference between groups (Kruskal Wallis, p = 0.024). Median frequency was higher for pain cases (7.5, 95% CI 6.5–10.5) and normal controls (7.5, 95% CI 6.5–9.5) than in acute controls (6.5, 95% CI 5.5–7.5).
Median nocturia was 1.5 episodes (IQR 2.0, range 0.5–6.5), with a significant difference between groups (Kruskal Wallis, p = 0.015). For pain cases median nocturia was higher (2.5, 95% CI 2.5–3.5) than in acute controls (1.5, 95% CI 1.5–2.5) and normal controls (1.5, 95% CI 1.5–2.5).
Daytime incontinence was reported in 58/154 (37.7%) participants, with no between-group difference (p = 0.998). Night incontinence was noted in 21/154 (13.6%) participants, with between-group difference (p = 0.328).
LUTS scores were sub-divided into four symptom groups. 68/159 (42.8%) reported any stress incontinence symptoms; the median score overall was 0/7 (95% CI 0–1, range 0–5). 110/159 (69.2%) reported any overactive bladder symptoms; the median score overall was 2/12 (95% CI 1–2, range 0–8). 126/159 (79.2%) reported voiding symptoms; the median score overall was 2/8 (95% CI 2–2, range 0–8). There was no difference between groups for these.
35/50 (70%) of pain cases reported any LUTS pain symptoms versus 28/58 (48.3%) acute controls and 22/51 (43.1%) normal controls (p = 0.016). The median score for pain symptoms was significantly different between groups (p = 0.003). Median pain score for pain cases was 3/13 (95% CI 1–4, range 0–8) which was significantly different to acute controls (median 0, 95% CI 0–2, range 0–9) and normal controls (median 0, 95% CI 0–2, range 0–9).
Figure 5 compares symptoms reported by participants in each group. LUTS were common in all groups of women, not just those presenting with abdominal pain or acutely without abdominal pain. The commonest symptoms reported, across all groups, were urinary urgency (75/158, 47.5%), incomplete bladder emptying (72/155, 46.5%), bladder pain on filling (70/157, 44.6%), bladder pain relieved by voiding (61/144, 42.4%), double voiding (65/156, 41.7%), waking rising urgency (61/156, 39.1%), and cough-sneeze incontinence (60/159, 37.7%).
Fig. 5Detailed LUTS compared across study groups (asterisks signify significant difference with p < 0.05 comparing three groups for this symptom; after Bonferroni correction these were not significant)
The only differences between groups were symptoms of bladder pain on filling (p = 0.046), bladder or suprapubic pain (p = 0.039), left or right iliac fossa pain (p = 0.005), and pain radiating to legs (p = 0.037), all of which were more frequent in pain cases. Using a Bonferroni correction, none of these remained significant.
Participants presenting with abdominal pain reported lower quality of life. Median EQ-5D-5 L index value differed between groups (p = 0.032). Median index value for pain cases was 0.74 (95% CI 0.71–0.84) compared with 0.84 (95% CI 0.84-1.00) for normal controls, with acute controls in between. Median VAS score also differed between groups (p < 0.001). Median pain cases score was 70% (95% CI 60–80) compared with 88.5% (95% CI 80–90) for normal controls, with acute controls in between.
Using fresh unspun microscopy (see Supplementary Table 6), white blood cells (WBC) were noted in 139/159 (87.4%) samples, with median white cell count of 14 WBC/µl (95% CI 10–24) and no difference between groups (p = 0.054, p = 0.102 respectively). Cell counts ranged from 0 to 1190 WBC/µl.
Red blood cells (RBC) were noted in 77/159 (48.4%), with median red cell count of 0 RBC/µl (95% CI 0–2), and no difference between groups (p = 0.754, p = 0.634 respectively). Cell counts ranged from 0 to 7520 RBC/µl.
Epithelial cells were noted in 140/159 (88.1%), with median epithelial cell count (ECC) of 14 ECC/µl (95% CI 8–18), with no difference between groups (p = 0.863, p = 0.233 respectively). Cell counts ranged from 0 to 184 ECC/µl.
Urinary ATP was measured for all participants (see Table 6). Median ATP was 5947 RLU (95% CI 4653–8064) for pain cases, 8203 RLU (95% CI 6967–9149) for acute controls, and 5157 RLU (95% CI 3649–7186) for normal controls. There was a distribution difference between groups (p = 0.005). The range of ATP values was wide across all three groups.
Table 6ATP measurement for Abdominal pain studyMeasure (units)Pain cases (n = 50)Acute controls (n = 58)Normal controls (n = 51)Overall(n = 159)ATP - median(relative light units, RLU)594782035157671795% CI (RLU)4653–80646967–91493649–71865513–7900Range (RLU)718–48,8001095–21,889679–16,347679–48,800p = 0.005 Kruskal-Wallis
Sediment culture was carried out for all participants (Supplementary Table 7). Median total colony counts were 1220 CFU/ml (95% CI 720–1945) with no difference between groups (p = 0.239). The overall range was 0–1,390,319 CFU/ml. The median colony counts for the dominant species were 976 CFU/ml (95% CI 603–692, range 0–1280000) with no difference between groups (p = 0.246). The number of isolates identified was a median of 3.0 (95% CI 3–3), with no difference between groups (p = 0.875). The proportion with any bacteriuria was 157/159 (98.7%) with no difference between groups (p = 0.534).
Isolates identified by aerobic sediment culture, categorised by genus, are shown in Fig. 6(A) and Supplementary Table 8. No phenotypic identification by Analytical Profile Index (API) was performed in 7/471 cases (1.5%). The commonest three genera were Staphylococcus spp (present in 169/471 (35.9%) samples), Enterococcus spp (74/471, 15.7%), and Lactobacillus spp (58/471, 12.3%). Other genera identified were Escherichia spp (44/471, 9.3%), Corynebacterium spp (34/471, 7.2%), Streptococcus spp (30/471, 6.4%), Candida spp (19/471, 4.0%), Citrobacter spp (8/471, 8%), and Aerococcus spp (6/471, 1.3%). Other genera were identified in 22/471 (4.7%) cases. These distributions are significantly different (p = 0.023, Chi square).
Fig. 6A - The ten most common isolates obtained using sediment culture, categorised by Genus, arranged in descending order of frequency by pain cases; B - The ten most common isolates obtained using sediment culture, categorised by species, arranged in descending order of frequency by pain cases
Isolates identified by sediment culture, categorised by species, are shown in Fig. 6(B) and Supplementary Table 9. No API was performed in 7/471 cases (1.5%).
The commonest three isolates overall, categorised to species level (except Lactobacillus spp), were Staphylococus haemolyticus (72/471 isolates, 15.3%), Enterococcus faecalis (69/471, 14.7%) and Lactobacillus spp (58/471, 12.3%). Other common species included Staphylococcus epidermidis (53/471, 11.3%), Escherichia coli 1 (44/471 (9.3%), Streptococcus agalactiae (26/471, 5.5%), Candida albicans (15/471, 3.2%), Corynebacterium striatum / amycolatum (15/471, 3.2%), Corynebacterium Group G (15/471, 3.2%) and Staphylococcus aureus (14/471, 3.0%).
These distributions were significantly different (p = 0.0075, χ^2^ = 38.61). The ten commonest species occurred in all three groups. The most striking difference between the groups was that E. coli occurred in 26/149 (17.5%, 95% CI 11.9–24.7) isolates in normal controls compared with 6/149 (4.0%, 95% CI 1.7-9.0) pain cases and 12/173 (6.9%, 95% CI 3.8–12.1) acute controls.
All species identified according to study group are illustrated in a Venn diagram in Fig. 7. This shows that 15 species were identified in all study groups, while many others were only identified in one study group.
Fig. 7Venn diagram of species identified by enhanced culture
Urinary cytokine results are demonstrated in Supplementary Tables 10 and Fig. 8.
Fig. 8Dot plots of urinary markers versus study group showing median and 95% confidence intervals; A: IL-6 (logarithmic scale); B: IL-8 (logarithmic scale); C: Lactoferrin (logarithmic scale)
Median urinary IL-6 (Fig. 8(A)) for pain cases was 1.71pg/ml (95% CI 1.0-3.2), for acute controls was 1.90pg/ml (95% CI 1.3-5.0) and for normal controls was 1.22pg/ml (95% CI 0.5–2.3). There was a difference between groups (p = 0.046). The mean proportion with IL-6 below the lower limit of detection (LLOD) was 24/159 (15.1%), with no difference between groups (p = 0.121). The range was 0.14pg/ml to 281pg/ml.
Median urinary IL-8 (Fig. 8(B)) for pain cases was 60.6pg/ml (95% CI 32.9-117.1), for acute controls was 38.7pg/ml (95% CI 17.1–80.2) and for normal controls was 21.9pg/ml (10.4–52.2). There was no difference between groups (p = 0.063). The mean proportion with IL-8 below the LLOD was 10/159 (6.3%) with no difference between groups (p = 0.848). The range was 0.71pg/ml to 2782pg/ml.
Median urinary lactoferrin (Fig. 8(C)) for pain cases was 6.0ng/ml (95% CI 3.5–19.2), for acute controls was 5.2ng/ml (95% CI 2.2–8.3) and for normal controls was 2.2ng/ml (2.2–3.3). Note the LLOD was 2.2ng/ml. There was a difference between groups (p = 0.006) with the median for pain cases higher than normal controls. The proportion with lactoferrin below the LLOD was 71/159 (44.7%) which was different between groups (p = 0.016). The range was 2.2ng/ml to 895.5ng/ml.
Median protein concentration for pain cases was 86.5 µg/ml (95% CI 57.6-103.5), for acute controls was 89.2 µg/ml (95% CI 76.4-105.5), and for normal controls was 60.1 µg/ml (95% CI 44.9–77.7). There was a different distribution between groups (p = 0.008).
When normalised for protein, there was no difference between groups for any of urinary IL-6, IL-8 or lactoferrin (respectively p = 0.134, p = 0.213, p = 0.251).
Figure 9. Epi-fluorescence microscopy images of shed urinary cells. DAPI and WGA stain for host / pathogen DNA and cell membrane respectively. Each channel for WGA (1) and DAPI (2) shown in monochrome. Composite (3) shows DAPI in green and WGA in magenta. Each image viewed at 100x magnification. Scale bar shows 10 μm. A Epithelial cell with associated bacteria, from a pain case. B Epithelial cell without associated bacteria, from an acute control.
Fig. 9shows examples of epithelial cells with and without associated bacteria The top line (A) shows a cell associated with bacteria from a pain case, the bottom line (B) shows a cell without bacterial association, from an acute control
The median proportion of epithelial cells associated with bacteria was 21.1% (95% CI 14.0-27.8) with no difference between groups (p = 0.464) (Supplementary Table 11). This distribution is shown in Fig. 10.
Fig. 10Percentage of epithelial cells with associated bacteria
The proportion of slides with any bacterially-associated cells was 111/157 (70.7%, 95% CI 62.8–77.6) with no difference between groups (p = 0.482).
Figure 11 shows uroplakin-3 stained epithelial cells (urothelial cells). These are without associated bacteria (C) and with associated bacteria (D and E).
Fig. 11Immunofluorescence under epi-fluorescent microscopy. DAPI and UP3 (uroplakin-3) stain for host / pathogen DNA and urothelial surface membrane glycoprotein respectively. Each channel for UP3 (1) and DAPI (2) shown in monochrome. Composite (3) shows DAPI in green and UP3 in magenta. Each image viewed at 100x magnification. Scale bar shows 10μm. C Urothelial cell without associated bacteria, from an acute control. D Urothelial cell with associated bacteria, from a normal control. E Two overlapping urothelial cells with associated bacteria, from a normal control
Supplementary Table 12 shows results of uroplakin-stained epifluorescent analysis of epithelial cells. The median proportion of epithelial cells that were uroplakin-3-positive was 100% (95% CI 94–100), with no difference between groups (p = 0.864). The median proportion of uroplakin-3-positive epithelial cells that had associated bacteria was 1.3% (95% CI 0–31). 9/18 (50%) slides had any uroplakin-3-positive cells with associated bacteria.
The proportion of uroplakin-3-positive urothelial cells with associated bacteria was 1.3% (95% CI 0–31), compared with 21.1% (95% CI 14–28) of all epithelial cells. The proportion of samples with any uroplakin-3-positive urothelial cells with associated bacteria was 9/18 (50%, 95% CI 27–73), compared with 15/18 (83.3%, 95% CI 58–96) of all epithelial cells for the same samples. There was no difference in distribution, however (Fisher’s Exact, p = 0.206).
We performed a secondary analysis of pain cases and acute controls, comparing women who were diagnosed with UTI on the day of recruitment against those who were not diagnosed with UTI (Table 7). This showed that IL-8, lactoferrin and fresh microscopy white cell counts were higher in those diagnosed with UTI. This applied for cytokines normalised for protein as well. Applying a Bonferroni correction (p < 0.004) the difference in lactoferrin remained significant.
Table 7Comparison of alternative tests according to clinical diagnosis of UTIMedian (95% CI of median)Diagnosed with UTI (n = 20)Not diagnosed with UTI (n = 74)SignificanceTotal LUTS8 (6–11)5.5 (4–8)MWU = 887p = 0.17324-hour frequency11 (8–14)9 (8–11)MWU = 859p = 0.110Quality of life %VAS75 (60–84)75 (70–80)MWU = 674p = 0.922Quality of life index value0.77 (0.71–0.84)0.84 (0.80–0.91)MWU = 575p = 0.217White cell count28 (14–78)8 (6–16)MWU = 1038p = 0.006Epithelial cell count24 (10–50)14 (7–20)MWU = 885p = 0.180ATP6600 (4340–16980)8030 (6920–8850)MWUp = 0.868Total colony count23,330 (815-126970)1250 (620–2900)MWU = 939p = 0.066Number of isolates3.5 (3–4)3 (3–3)MWU = 891p = 0.150IL-61.2 (0.5-5.0)1.9 (1.3–3.2)MWU = 592p = 0.171IL-8191 (53–471)34 (17–70)MWU = 999p = 0.017Lactoferrin21.7 (8.7–36.7)4.1 (2.2–6.1)MWU = 1098p = 0.001**p* < 0.05 Bonferroni correction significance = 0.05/12 = 0.004. MWU=Mann Whitney U test
Our primary hypothesis was that pregnant women with abdominal pain may have urinary pathology not detected by standard tests. We compared standard and alternative urinary tests in women with abdominal pain with acute controls who had other presenting symptoms, and normal controls attending routine antenatal care. Importantly, all groups had findings suggestive of urinary pathology not detected by standard tests, implying either that all pregnant women have urinary pathology or that alternative tests are not reliable at UTI diagnosis.
Symptoms were prevalent in all groups, with scores equivalent to those seen in patients with chronic LUTS. Pain cases reported more pain symptoms and nocturia, and lower quality of life scores, with a different distribution of genera and species using enhanced culture, but there were no other differences between groups. Despite this, and despite equivalent rates of positive standard culture, pain cases were more likely to be diagnosed with UTI and prescribed antibiotics than acute controls. Study group demographics were comparable across groups, other than pain cases being more likely to reside in areas with a higher index of deprivation. An association between lower socio-economic status and UTI has been previously reported [91, 92].
Women diagnosed with UTI had higher urinary IL-8, lactoferrin and fresh microscopy white cell counts than those not diagnosed with UTI, despite there being no difference in the proportion that were positive by standard culture. This may validate the clinical diagnosis of UTI, may reflect the influence of urine dipstick leucocyte esterase on diagnosis, or may reflect clinical gestalt, i.e. holistic clinical judgement based on knowledge and experience and the risk/benefit balance favouring treatment to avoid obstetric complications such as PTB [93].
The other noticeable finding from these data are the ranges of results seen. There was a marked range in all groups in symptom scores and inflammatory markers such as cell counts, ATP and urinary cytokines, indicating that our study may be underpowered. This demands further exploration to understand the pathophysiology and clinical relevance.
We hypothesised that epithelial cells in the urine represented urothelial cells and in the subset of cases that underwent uroplakin-3 testing, a urothelial origin was confirmed. This is in keeping with another study of urinary epithelial cells in pregnancy using the same method [89], and questions the assumption that epithelial cells must represent contamination.
There was heterogeneity of presenting symptoms. A small number in both pain cases and acute controls mentioned urinary symptoms. Acute controls, in particular, represent a diverse range of potential pathology. Similarly, there was heterogeneity of diagnoses, with a proportion in both groups found not to have pathology, and a predominance of musculoskeletal/ligamentous pain and UTI in the pain group. This would be consistent with the original hypothesis that women with abdominal pain in pregnancy may have UTI. The range of presentations is similar to those cited in a recent UK obstetric triage study [94], while the rate of non-diagnosed pain was similar to a recent Finnish study [95].
There was reasonable agreement between patients and clinicians regarding likelihood of UTI. Respect for patient experience of UTI symptoms, despite negative cultures, has led to advances in understanding chronic UTI [96]. Clinician gestalt is valuable, and where a diagnostic test disagrees, both the clinical diagnosis and test should be critically examined [93]. The recent ALTAR study, assessing the role of methenamine hippurate as an alternative to antibiotics for recurrent UTI in non-pregnant patients, did not rely on culture positivity for defining UTI, which acknowledges the flaws in this test and respects patient experience [97].
Documentation of LUTS was more common in those presenting with pain, and in those diagnosed with UTI. Fewer than half of the documented medical history mentioned LUTS. Data from non-pregnant hospital patients show a low correlation between self-reported urinary symptoms and medical record documentation in patients with a positive culture UTI [98, 99]. Urinary symptoms in pregnant women may be under-valued in diagnostic reasoning.
Polymicrobial enhanced cultures were almost universal, although standard cultures showed no growth in more than half, supporting the concept of the non-sterile bladder microbiome in pregnancy. This result is similar to what is seen outside pregnancy, where a urobiome is essentially ubiquitous, and polymicrobial growth on sediment culture is common in healthy individuals [3, 29, 100, 101]. A cohort study of patients screened for ASB at first pregnancy appointment found 2942/6095 (48%) had non-significant growth (bacterial growth < 10^5^CFU/ml threshold) in standard urine culture; PTB was more common in this group than those with no growth (adjusted OR 1.91, 95% CI 1.13–3.26) [102].
Regarding urinary ATP testing, there was a distribution difference between groups, although clinical interpretation is uncertain. A recent pregnancy study demonstrated voided urinary ATP values at term of mean 2504 attamoles (95% CI 1415–5779); the same patients giving a subsequent catheter sample showed mean 12,323 attamoles (95% CI 8250–20980), suggesting a mechanical aspect such as bladder irritation or contraction contributing to ATP levels [89].
24 genera were identified on enhanced sediment culture. This order of magnitude is similar to a recent study of 29 first trimester urine samples, where 35 genera were identified by expanded quantitative urine culture methods (EQUC) with MALDI-TOF for identification [103].
Distributions of the genera and isolates differed between groups with a mixture of contaminants or commensals (such as S. haemolyticus or S. epidermidis) and traditional uropathogens (such as E. coli and E. faecalis). It cannot be assumed that all traditional non-uropathogens are contaminants. Other methods are needed to establish the origin, or indeed virulence status, of these organisms. Interestingly, although there are similarities between vaginal and urinary microbiomes in pregnancy, they are not completely concordant and do not predict each other more than 75%^103^.
The median number of bacterial isolates per sample using enhanced culture was three. In a study of enhanced cultures using the same sediment culture technique, comparing patients with chronic LUTS with asymptomatic controls, a similar pattern was found [100]. We should question the assumed superiority of a single positive standard culture, which may not favour growth of some species under the particular clinical laboratory conditions used.
S. agalactiae (Group B Streptococcus or GBS) was isolated in 26/159 (16.3%) of enhanced culture samples, but only two standard culture samples (1.3% of patients). This has implications for pregnancy as opportunities for antibiotic prophylaxis may be missed. While GBS is part of vaginal flora, it is also known that its detection in the urine is associated with poor neonatal outcomes as described previously [43]. In addition, it may cause maternal bloodstream infection [104] and can progress to pyelonephritis, albeit at lower rates than *E. coli *[105].
In our study, median IL-6 levels were comparable in range with a study comparing cranberry supplement with placebo in pregnant women without UTI [70], and a study of patients undergoing amniocentesis to detect intra-amniotic infection, which included patients with and without infection or PTB and no control for UTI [106]. Similarly, median IL-8 levels were comparable with the previously-mentioned studies of pregnant women with ASB [73, 74].
Lactoferrin concentrations ranged from 2.2 to 895ng/ml, with a median of 4.2 (95% CI 2.2–5.8). Mean lactoferrin was 26.4ng/ml (95% CI 11.2–41.6). This is comparable to the values reported in non-pregnant patients, where mean urinary lactoferrin was 30.4+/-2.7ng/ml (average +/- standard error of the mean) in healthy controls and 60.3+/14.9ng/ml in patients without UTI [77]. In patients with UTI, levels were 3300+/-646.3ng/ml. These authors cited a cut-off of 200ng/ml lactoferrin concentration for diagnosis of UTI. Ours is the first report of urinary lactoferrin values in pregnancy, to the best of our knowledge.
Lactoferrin values were higher for pain cases, although there was no difference between groups in normalised cytokine values. The decision about whether and how to normalise urinary cytokines is complex [107]. Normalising for protein allows a direct comparison with two other urinary lactoferrin studies in non-pregnant patients [108, 109]. Higher urinary lactoferrin was found in urogynaecology patients who were culture-positive for UTI compared with those who were culture negative [109]. There were no differences in urinary lactoferrin between controls and patients who had suffered cutaneous burns [108]. The values in those studies of urinary lactoferrin reported are markedly different from each other, with those from this study in between. Nearly half of the current study results had urinary lactoferrin below lower limit of detection (LLOD). Comparing proportions below LLOD did show a significant difference (p = 0.016) with normal controls having the highest proportion.
This study takes a multifaceted approach to UTI in pregnancy, incorporating clinical outcome data, questionnaire based detailed symptomatology and patient quality of life data, as well as a broad range of microbiological and immunological tests some of which, for example urinary lactoferrin, have not been previously studied in pregnancy. Having questioned the gold standard for diagnosing UTI, we explored alternative diagnostic tests. In this prospective study, by using abdominal pain in pregnancy as a paradigm for exploring UTI in pregnancy, we were able to avoid making assumptions about what defines UTI. Importantly, such a group would be expected to include some women with infective urinary pathology. The study was conducted in a real-life clinical environment with patients recruited within the same hospital facility to reduce inclusion bias, either within an obstetric triage unit where they presented acutely with or without abdominal pain, or in the routine hospital-located antenatal clinic.
Broad recruitment criteria were chosen to maximise the likelihood of identifying urinary pathology, however this strategy made it less likely that any between-group differences or lack of differences in tests could be explained. In addition, the heterogeneity of presenting symptoms within the groups, and potential for overlap, may introduce bias. For example, there was a trend towards increased chance of ruptured membranes in the acute controls, although this was not significant when corrected for multiple comparisons.
Given the small sample size, gross differences in pregnancy outcomes would also not be expected. In the Dutch study of ASB [8], the primary outcomes of pyelonephritis with or without PTB were rare, and it was calculated that large sample sizes would have been required to detect a difference. Antimicrobial prescribing differed markedly between Pain cases and Acute controls. This adds complexity to the use of end-of-pregnancy outcomes to compare the groups. The study was powered to detect a difference in pyuria only.
As a cross-sectional study, strength of conclusions is likely to be limited, and although comparison was made with clinical outcomes, in a study of this size it is not possible to reach firm conclusions about to what extent this prevalence of urinary inflammatory and microbiological signals represents pathological processes.
Clinician and patient assessments of likelihood of UTI were not made for normal controls, which would have acted as a baseline. In addition, the proportion of participants with missing data affects data reliability. Epithelial cell analysis was limited by the number of analyses performed and should be assessed in a larger sample.
The use of chromogenic agar and aerobic conditions with the enhanced culture approach favours traditional uropathogens, which will influence the species isolated compared with a sequencing approach. It is unfortunate that Lactobacilli could not be classified at species level, given associations with healthy and symptomatic vaginal and urinary microbiomes of different Lactobacillus species.
The use of API testing for species identification is a potential weakness [110]. At the time of designing the study, this phenotypic approach was being used by our research laboratory as it is rapid and cost-effective. However, sequencing methods would also be able to identify non-culturable organisms. Urinary microbiome studies are starting to unravel how this changes in pregnancy [111], and becoming increasingly important in the field of PTB research [112].
Study particpants were recruited between 2014 and 2015 with laboratory analysis ongoing until 2019. While pregnancy urinary microbiome research has advanced in this time [113], diagnostic methods for UTI in pregnancy remain a challenge [114] and this study design remains unique in a pregnancy cohort to the best of our knowledge. Pyuria by fresh unspun urine microscopy remains the most useful marker in chronic UTI despite considerable ongoing research in this field [115].
This explored alternatives to the long-established but questionable gold standard of UTI diagnosis. Should a viable alternative diagnostic approach be identified, pregnancy studies would need to be large enough to detect clinical differences in outcomes and re-examine the associations between UTI in pregnancy, preterm birth and pyelonephritis in particular. It is difficult to demonstrate clinical evidence in pregnancy that the sensitivity of standard culture is inadequate for diagnosis, for example in pyelonephritis, where definitions have historically required significant bacteriuria [116, 117]. In questioning the gold standard for diagnosis for UTI, we have a dilemma; how to assess any alternative diagnostic approaches? Broader definitions take into account symptoms and pyuria and do not require bacteriuria [8, 118] but further cross-disciplinary innovative work is required to challenge existing dogma [4].
In this prospective study, pregnant women whether with abdominal pain or other acute hospital presentations or simply attending routine antenatal care, had symptoms of UTI with scores equivalent to those seen in patients with chronic LUTS. Women presenting acutely with abdominal pain did have more pain symptoms, nocturia, and lower quality of life scores, with a different distribution of bacteria using enhanced urine culture, but they had equivalent rates of positive standard urine culture compared to those with no pain or those in routine antenatal care. Subsequently they were more likely to be diagnosed with a UTI and prescribed antibiotics. Our study shows that current standard testing for UTI in pregnancy, and understanding of what is normal and abnormal, is inadequate. The use of alternative tests, well validated in a chronic UTI research programme, raises questions about current urine testing practice in pregnancy and the assumptions that drive them. Further research should examine these tests in a different pregnancy context, to determine if they can enhance UTI diagnosis, better guide management of urine pathology in pregnancy and improve pregnancy outcomes.
Supplementary Material 1.