Potential Predictive Factors for Postpartum Endometritis and the Microbiology in Cervical Cultures.

WRITTEN REPORT Medicine programme, degree project (30.0 c) Potential Predictive Factors for Postpartum Endometritis and the Microbiology in Cervical Cultures. Written by: Ellinor Eriksson Supervisor: Maria Jonsson Date: 2018-08-27 Uppsala University Department of Women s and Children s Health

Potential predictive factors for postpartum endometritis and the microbiology in cervical cultures. Written by: Ellinor Eriksson Illustration on front page: Ellinor Eriksson, 2018. Table of contents Populärvetenskaplig sammanfattning (Summary in Swedish)... 3 Abstract... 4 List of abbreviations... 5 Definitions... 5 Background... 6 Description of the condition... 6 Pathogenesis and microbiology... 7 Prognosis... 7 Prevention... 8 Treatment... 9 Risk factors... 9 Importance of this study... 10 Objectives... 10 Material and methods... 11 Study design and settings... 11 Study population... 11 Data collection... 11 Ethical approval... 11 Variables... 12 Statistical analysis... 12 Results... 14 Discussion... 18 Potential predictive factors... 18 Microorganisms in cervical cultures... 20 Strengths and limitations... 21 Conclusion... 23 Acknowledgement... 23 References... 24 Appendix 1 List of variables... 31 2

Populärvetenskaplig sammanfattning (Summary in Swedish) Postpartumendometritit (PPE) är en infektion i livmoderslemhinnan som drabbar ungefär 1 av 50 kvinnor efter förlossning. Diagnosen ställs vanligen vid feber i frånvaro av annan infektionsorsak. Nedre buksmärta, illaluktande flytningar och förhöjda infektionsprover är kliniska fynd som stärker misstanken, men vilka inte är specifika för PPE. Infektionen är vanligtvis lindrig men de flesta patienter kräver antibiotikabehandling. Med adekvat behandling är prognosen god men komplikationer kan tillstöta såsom abscesser, sepsis, septiska embolier, toxiskt chocksyndrom och i värsta fall död. Globalt är infektion den tredje största orsaken till mödradödlighet, motsvarande drygt tio procent. Trots att PPE är en relativt vanlig och i vissa fall allvarlig diagnos är det fortfarande inte helt känt vilka bakterier som orsakar infektionen. Det finns flera föreslagna riskfaktorer för PPE. Kejsarsnitt är den viktigaste och mest studerade. Antibiotikaprofylax rekommenderas till kvinnor förlösta med kejsarsnitt, vid intrapartal feber samt på vissa kliniker vid långvarig vattenavgång (LVA), trots att evidensen för antibiotikaprofylax vid LVA är omdiskuterat. Det finns fler faktorer som är associerade till PPE men där evidensen är lägre och studieresultaten skiljer sig. Vår studie visar att manuell lösning av moderkakan är oberoende associerat till PPE vid vaginal förlossning. Andra faktorer som tycks vara associerade, men som kan vara beroende av varandra, är överburenhet, igångsättning av förlossning, värkstimulerande behandling, förlängt utdrivningsskede och kejsarsnitt trots antibiotikabehandling. Det är en liten studie och mer forskning behövs för att bättre kunna identifiera riskfaktorer. Nyttan med antibiotikaprofylax till riskgrupper vid vaginal förlossning bör även studeras i randomiserade kontrollerade studier. 3

Abstract Objective. To evaluate predictive factors for postpartum endometritis and to describe the microbiology. Design. Case-control study. Setting. Uppsala University hospital, Sweden in 2017. Methods. Cases (n =57) were women who had a vaginal onset of labour and who, within 14 days postpartum, were diagnosed with endometritis. Controls (n = 114) were the first two women giving birth after each case, matched for parity. Multiple gestation, elective caesareans, preterm births (<34 weeks) and intrapartum fever ( 38.0 C) were excluded. Data were collected from medical records, partograms and results of the cervical cultures. Crude and adjusted odds ratios (OR) with 95% confidence intervals (CI) were calculated by logistic regression. Results. Post-term delivery ( 42 weeks), induction of labour, prolonged duration of second stage, oxytocin treatment, caesarean section and manual placental removal were associated with postpartum endometritis. Only manual placental removal (OR 11.3, 95% CI 1.2-102.4) persisted in the multivariate analysis. Cervical cultures were obtained from the majority of the cases but only 47.1% were positive and showed diversity. Conclusion. Manual placental removal is a predictive factor for postpartum endometritis. Other potential predictive factors found in this material need to be evaluated in larger studies. The microbiology of the infection is still not clearly defined. 4

List of abbreviations American College of Obstetricians and Gynecologists ACOG Body Mass Index BMI Confidence interval CI Interquartile range IQR Meconium-stained amniotic fluid MSAF Odds ratio OR Postpartum endometritis PPE Prolonged rupture of membrane PROM Toxic shock syndrome TSS World Health Organization WHO Microorganisms Escherichia coli E coli Gardnerella vaginalis G vaginalis Group A streptococcus GAS Group B streptococcus GBS Group G streptococcus GGS Staphylococcus aureus S aureus Definitions See appendix 1 for list of definitions. 5

Background Description of the condition The diagnosis of postpartum endometritis (PPE) is clinical, and mainly based upon the presence of postpartum fever in absence of any other infectious aetiology. The American Committee on Maternal Welfare defines puerperal febrile morbidity as an oral temperature of 38.0 or higher on any two of the first ten days postpartum, excluding the first 24 hours. 1 Low grade fever during the first day is not uncommon and often passes spontaneously, and is therefore excluded in the definition. 2 Midline abdominal pain, purulent or foul-smelling lochia, uterine tenderness and leucocytosis are common clinical findings supporting the diagnosis of endometritis but are variable and nonspecific for the condition. 3 Abrupt occurrence of fever exceeding 38.9 C and severe abdominal pain is more often associated with PPE caused by group A streptococcus (GAS). However, it is not possible to clinically determine the aetiology of the infection 4. Symptoms can manifest hours to several days after childbirth contributing to difficulties in identifying cases especially after hospital discharge. 5 Failure to identify signs of infection delays the diagnosis and treatment which increases the risk of morbidity and mortality. 6 PPE is diagnosed after 0.2% to 3.6% of vaginal deliveries and it is up to 21 times more common after caesarean deliveries. 7,8 The incidence of PPE in Sweden in 2015 was 1.9% after vaginal births and 3.6% after caesarean deliveries including both elective and emergency caesareans. The total incidence of PPE in Sweden was between 1.3% to 3.6% altogether. 9 In a cross-sectional study from a hospital in Denmark, investigating women giving birth during May 2007 to April 2008, the incidences within four weeks postpartum were 2.0% after vaginal delivery, 2.0% after elective caesarean delivery and 5.0% after emergency caesarean delivery. The total rate after all deliveries was 2.0%. 10 In 1995 to 2000 at a hospital in the United States, 1.6% of all women giving birth were diagnosed with PPE. The rate was 0.5% among women with vaginal delivery and 7.0% among women with caesarean delivery, both elective and emergency caeasareans. 11 In a trial from Israel from 2000, the rates were 0.2% after vaginal delivery and 2.6% after caesarean delivery. 7 A study in Uganda presented a total incidence of PPE of 1.8%, of which 86% had had caesarean deliveries. 12 6

Pathogenesis and microbiology PPE is an infection of the decidua, the part of the endometrium that changes before and during pregnancy. The infection often occurs during labour and delivery due to contamination of the uterine cavity and invasion of the endometrium by cervicovaginal organisms. 3 The occurrence of infection depends on the interaction among the quantity of bacterial inoculum, the virulence of the present bacteria and the host s mechanisms of defeating infection. 13 The microbiological aetiology is rather difficult to identify and is not clearly defined. Unless blood cultures are positive or the genital culture is obtained from the upper genital tract without contamination from the vagina, the infection is rarely laboratory confirmed. 3 The typical infection is polymicrobial and more than one organism is usually found in the cultures, often a mixed aerobic and anaerobic flora. 14 There are many different microorganisms associated to PPE, suggested as potential pathogens. The largest bacterial groups are Ureaplasma urealyticum, Mykoplasma hominis, organisms associated with bacterial vaginosis, such as Gardnerella Vaginalis (G vaginalis), Bacteroides species, Peptostreptococcus species and Bacteroides bivius 15, and other miscellaneous facultative and anaerobic bacteria, including Escherichia coli (E coli) and Streptococci. 14 Streptococci is the most common gram-positive facultative microorganism including for example group B streptococcus (GBS), enterococci, such as Enterococcus faecalis, and other streptococci. In some patients Staphylococcus aureus (S aureus) can be found in the endometrial cultures without any other pathogen present. 14 Some suggests Streptococcus agalactiae, a type of GBS, as the primary pathogen implicated. 16 PPE caused by GAS are rare, however, endometritis is a common presentation of GAS infection in the puerperium. 4 In addition to those bacteria mentioned, Peptococcus species and Clostridium species are possible microorganisms, causing particularly post-caesarean endometritis. 17 Prognosis Nowadays, most cases of PPE are rather mild and cured by proper administration of antibiotic therapy. However, puerperal fever has been a common cause of maternal mortality before the antibiotic era, and is still a common cause for complications in low-income countries if treated inadequately. Globally, infection is the third largest direct reason to maternal mortality, corresponding for just over ten percent. 18 Since the use of antibiotics has increased, 7

a sharp decrease in maternal morbidity of PPE has been observed. Therefore, antibiotic therapy is generally accepted as the golden standard of treating PPE. Even though the prognosis of PPE is generally good there are complications of the condition. These includes formation of intra-abdominal and pelvic abscesses, extension of the infection to peritoneal cavity triggering peritonitis, sepsis, septic pelvic thrombophlebitis that might cause septic pulmonary embolism and even death. 3 In presence of toxin-producing GAS, necrosis of tissue and local clotting of blood vessels are possible even after administration of antibiotics. The produced toxins do not respond to antibiotic treatment and can continue to leak into the vasculature, producing a widespread organ failure and toxic shock syndrome (TSS). If TSS derives after antibiotic treatment, it is evidence of the infection being aggressive and untreatable by antibiotics and that the release of toxins will be large enough to cause septic shock. In such cases, prompt surgery and hysterectomy are required to stop organ damage and to reverse spreading of infection. 4 Prevention The basis of preventing infection in women undergoing uncomplicated vaginal births is basic hygiene and infection control measures such as hand washing and disinfection, proper use of disposable gloves, minimizing vaginal examinations and sterilization of equipment. 5,19 In addition, prophylactic antibiotics is now routine in the prevention of PPE in elective and emergency caesarean deliveries and in the event of intrapartum fever (temperature of 38.0 C or higher). 20 23 In some clinics antibiotic prophylaxis is recommended in case of prolonged rupture of membranes (PROM, i.e. longer than 18 hours prior to childbirth) 24. However, there are no convincing evidence of benefits in either maternal or neonatal risk of infection from use of routine antibiotics for PROM (longer than 16-24 hours) near or at term. 25 Administration of intrapartum prophylactic antibiotics is also routinely given to women colonized with GBS or at preterm labour (less than 37 gestational weeks) although mainly to prevent neonatal infection. 26,27 The significance of prophylactic antibiotic treatment after uncomplicated vaginal births has been studied, and the results indicate that routinely antibiotics administrated intrapartum or within 1 hour after delivery reduces the incidence of PPE by up to 70%. However, the studies are few and of rather low quality which limits the interpretation of evidence. 28 The usage is therefore not recommended or praxis in Sweden. Furthermore, the incidence of PPE is rather low which indicates that a large number of women need to be treated in order to 8

avoid a small number of cases. 28 The risks of antibiotic-related side effects and the emerging antibiotic resistance also need to be considered. 29 Treatment Choice of treatment of PPE depends on timing and clinical severity. Women with mild endometritis in a good general condition can usually be treated outside of the hospital using broad-spectrum oral antibiotics such as amoxicillin/clavulanic acid, possibly with addition of metronidazole. 30,31 For treating severe PPE, there are a few different empiric antibiotic regimens. The most effective treatment is the combination of intravenous clindamycin and once-daily dosing of gentamicin which has a small number of treatment failures and wound infections. Successful treatment is usually defined as absence of fever for 24 to 48 hours. 3 Risk factors Several factors may contribute to the risk of developing PPE, both pre-existing maternal conditions and complications that may occur during labour and childbirth. The most important risk factor for developing PPE is caesarean section, especially when performed after vaginal onset of labour. The procedure increases the risk 6 to 21 times compared to the risk at vaginal delivery. 7,11 Operative vaginal delivery (forceps and vacuum delivery) is suggested a risk factor as well. 13 Maternal age less than 17 years 32 and presence of certain genital tract bacteria are pre-labour characteristics associated with PPE. Bacterial vaginosis during pregnancy increases the total risk of PPE two to three times compared with women without bacterial vaginosis. 33,34 At caesarean delivery, studies have shown an up to sixfold increased risk. 35 The association to time-related events during delivery has been studied extensively and some events are suggested as risk factors for PPE. In one study, five time-related peripartum events were investigated including duration of labour, duration of ruptured membranes, the number of vaginal examinations, the time from first vaginal examination to delivery and duration of internal monitoring. The results from that study indicate that duration of labour is the primary determinant of postpartum morbidity and suggest that other time-related events may only be associated to an increased risk of PPE due to correlation to a prolonged duration of labour. 36 Other more recent studies imply that PROM (longer than 16-24 hours) significantly increases the risk for PPE isolated from duration of labour. 13,32,37 There are more studies to support that 9

prolonged labours is a risk factor 38, whereas other studies 33,39 found no increased risk of PPE. Despite the possible association between endometritis postpartum and long labours, there are no randomized controlled studies investigating possible benefits of prophylactic antibiotics or guidelines of its usage. There are more intrapartum variables associated to PPE such as gestational age. Beyond 40 gestational weeks, the risk of PPE increases with every week up to 42 weeks. 40 Gestational age of less than 37 weeks has also been shown to increase the risk for PPE. 7 In the event of meconium-stained amniotic fluid (MSAF), the risk of developing clinical PPE is doubled. 41 There are studies investigating the benefits of prophylactic antibiotics at MSAF showing no statistically significant reduction in incidence of PPE. 41 Other not uncommon complications to delivery, increasing the risk of PPE, are manual removal of placenta after vaginal delivery and postpartum anaemia after both vaginal and caesarean delivery. 32,42,43 The incidence of retained placenta, which is the leading cause to manual removal, is 0.1% to 3%. It is less common in developing countries, however with a high fatality rate of up to 10%. 44,45 There are no randomized controlled trials comparing antibiotic prophylaxis and non-antibiotic use or placebo to prevent PPE after manual removal of placenta at vaginal birth 46 but a systematic review of non-randomized studies showed no significant reduction in incidence. 47 Importance of this study PPE contributes considerable to maternal morbidity and mortality and there might be an opportunity for prevention. Antibiotics are suggested prophylaxis but the evidence of profit of routinely administration of antibiotics to all women after uncomplicated vaginal birth is low. However, antibiotics are used as prophylaxis to prevent endometritis in some events such as caesarean section and intrapartum fever. It might also be beneficial in other specific situations that evolve during labour, also in women with vaginal delivery, such as manual removal of placenta and delayed long labours. To choose antibiotic treatment or prophylaxis, knowledge of microorganisms causing infection, is crucial. Objectives In a Swedish setting, we aimed to identify clinical variables associated to postpartum endometritis in women with vaginal onset of labour and, to describe the microbiology of the condition. 10

Material and methods Study design and settings This is a case-control study from the Department of Obstetrics and Gynaecology at Uppsala University Hospital in Sweden, which is a referral centre with approximately 4000 deliveries yearly. The period studied was January 1 st to December 31 st 2017. The study also includes a descriptive analysis of microorganisms present in cervical cultures of the cases. Study population Cases were women who had a vaginal onset of labour who within 14 days postpartum were diagnosed with postpartum endometritis according to the medical records (ICD-10 code O85.9). Since there are no universally accepted criteria for the condition, the outcome was based on the clinical diagnosis by the physician. Exclusion criteria included elective caesarean sections, preterm gestational age (less than 34 weeks), multiple gestation and, fever (temperature of 38.0 C or higher) at arrival to the delivery ward. The first two women registered at the delivery unit after each study case were chosen as controls and matched for parity. The same exclusion criteria as for the cases applied for the controls. Data collection Data were collected from medical records during January to March 2018. A database containing data of diagnoses entered from the healthcare information system Cambio COSMIC was searched for women diagnosed with postpartum endometritis during the set period of time. Controls were found in a register of all women in labour organized by order of arrival to the ward, maintained by the delivery unit at Uppsala University Hospital. Medical records, including partograms, were reviewed and data of chosen variables were retrieved. To isolate potentially pathogenic microorganisms, cervical cultures were taken from most of the cases at the time of return to the hospital. Urine cultures were also taken from most of the patients, but those results were of no importance in this study. Endometrial cultures were not taken from any of the women why results of the cervical cultures were the chosen method to identify microorganisms. The results from the cultures were reviewed to find present microorganisms. Ethical approval No application to the Research Ethics Committee was needed since this study is a quality control of the medical care conducted at Uppsala University Hospital and contributes to 11

development of the care. However, permission to review medical records of patients was given by Masoumeh Isfahani Rezapour, Operational manager of the Department of Women s health and Sune Larsson, Director of research and education at Uppsala University Hospital. There was no interaction with any of the subjects. Variables Information on parity, age, Body Mass Index (BMI), infertility treatment, tobacco use, intercurrent diseases, pregnancy complications, gestational age, duration of rupture of membrane, induction of labour, colour of amniotic fluid, duration of different stages of labour, oxytocin treatment, epidural anaesthesia, number of vaginal examinations, attendance of student, mode of delivery, foetal presentation, foetal birth weight, maternal blood loss, maternal blood transfusion, manual removal of placenta and uterine atony. A list of variables with definitions and methods of collection and calculation can be found in Appendix I. Statistical analysis IBM SPSS Statistics Subscription 2017 for Windows was used to perform statistical analysis. Chi Square test or Fisher s exact test were applied to compare distribution of categorical variables between groups. Fisher s exact test was chosen when the number of values in any group were smaller than four. For continuous variables, Mann-Whitney U test was applied when data were non-parametric and T-test when data were parametric. P-value <0.05 was considered statistically significant. Logistic regression was used to calculate odds ratio (OR) with 95% confidence interval (CI). Multivariate logistic regression was used to identify independent predictors since some of the variables are interrelated. Variables entered in the multivariate analysis were post-term pregnancy (42 weeks or later), induction of labour, second stage of 3 hours or longer, caesarean section and manual removal of placenta. Second stage of 3 hours or longer were used instead of the continuous variable of duration of second stage since a cut-off value is of more clinically use. 3 hours were the cut-off value with the lowest p-value (p 0.06) and therefore used in the multiple regression. Women who did not reach second stage because of caesarean section were not considered in the variable second stage of 3 hours or longer. The same applied to women delivering by caesarean section in the variable manual removal of placenta. The variable oxytocin treatment was excluded from the multivariate analysis since 12

most of the women who were induced got oxytocin treatment, why both variables were not included. OR with 95% CI was calculated. 13

Results During the study period there were 4,169 births (3,477 vaginal births (83.4%), 374 emergency caesareans (9.0%) and 318 elective caesareans (7.6%)) at Uppsala University Hospital. Out of these, 65 women (1.6%) were diagnosed with postpartum endometritis in total, of which 56 women delivered vaginally, 8 women by emergency caesarean and 1 woman by elective caesarean. The incidences of PPE were 1.6% after vaginal delivery, 2.1% after emergency caesarean and 0.3% after elective caesarean (1.3% after all caesarean deliveries in total). 57 women met our inclusion criteria and formed the case group (Figure 1). Total number of deliveries at Uppsala University hospital 2017 N = 4169 Women not diagnosed with PPE n =4104 PPE diagnosed >14 days postpartum n = 7 Elective caesarean section n = 1 Study population n = 57 Figure 1. Flowchart of cases. Table 1 describes maternal characteristics. There were no significant differences between the groups in terms of age, BMI and tobacco use. The total number of women with intercurrent diseases did not differ. However, psychiatric diagnoses were more common among cases, 22.8% vs 10.5% (p 0.03). Pregnancy complications did not differ between groups in total but fewer cases suffered from premature contractions, 15.8% vs. 30.1% (p 0.04). 14

Table 1. Maternal characteristics in the case and control groups. Characteristics Cases n = 57 Controls n = 114 p-value First vaginal delivery previous caesarean delivery 35 (61.4) 5 (8.8) 66 (57.9) 5 (4.4) 0.66 0.25 Age (years) 30.3 ± 6.1 30.0 ± 4.9 0.75 BMI (kg/m 2 ) 24.2 (21.9-27.7) 23.8 (21.5-26.6) 0.74 Infertility treatment 7 (12.3) 6 (5.3) 0.11 Tobacco use 11 (19.3) 15 (13.8) 0.35 missing Intercurrent diseases Diabetes mellitus Hypertensive disorders 1 17 (29.8) 0 (0.0) 7 (12.3) 5 31 (27.2) 5 (4.4) 9 (8.0) Pregnancy complications Colonization of Group B Streptococcus 48 (84.2) 1 (1.8) 102 (90.3) 9 (8.0) Values are given a n (%), means ± standard deviation (SD) or medians (quartiles). Bolded = significance. * Fischer s exact test. BMI = Body Mass Index IVF = In vitro fertilisation 0.72 0.17* 0.36 0.25 0.17* In Table 2 variables related to labour are displayed. Using univariate analyses, six clinical variables related to labour were found associated with PPE. Post-term delivery (42 weeks or later) was significantly more common among cases (15.8%) than controls (5.3%) but the mean gestational age did not differ between groups. There was a higher rate of induction of labour among cases (36.8% vs. 21.9%) and prostaglandins were used more often. We found no difference in duration from start of induction to delivery. Duration of second stage was significantly longer (p 0.03) in the case group. The median duration of second stage was 97 minutes (interquartile range (IQR) 37-184 minutes) among cases and 48 minutes (IQR 16-153 minutes) among controls. There were no differences regarding total duration of labour, duration of first stage or duration of bearing down effort. The use of oxytocin infusion was more frequent among cases (71.9% vs. 53.5%) but we found no difference regarding diagnosis of uterine inertia. There were no differences in rates of amniotomy, epidural use, prolonged rupture of membranes or number of vaginal examinations performed, even after correction for examinations performed by students. Cases more often delivered by caesarean section (14.0% vs. 5.3%). The median amount of blood loss was 400 ml (300-800 ml) among cases and 300 ml (250-456 ml) among controls (p 0.06). There was no significant difference between blood loss of 1000 ml or above, which is the limit used for postpartum haemorrhage. Manual removal of placenta was needed among 5 cases (8.8%) compared to 1 control (0.9%) (p 0.02). Controls had longer hospital stay, 2 days (IQR 1-3 days) vs. 1 day (IQR 1-2 days), after delivery (p 0.02). 15

Table 2. Clinical variables related to labour in the case and control groups. Characteristics Cases n = 57 Controls n = 114 p-value Gestational age (weeks) Post-term delivery ( 42 weeks) 39.6 ± 1.7 9 (15.8) 39.7 ± 1.3 6 (5.3) 0.69 0.02 Prolonged rupture of membrane ( 24 h) 11 (19.3) 13 (11.4) 0.16 Induction of labour 21 (36.8) 25 (21.9) 0.04 Time of induction of labour (min) 937 (511-1520) 708 (300-1624) 1.00 Cervical ripeness at induction Favourable Unfavourable Method of induction of labour Amniotomy Oxytocin Foley catheter Prostaglandin Double-balloon catheter Meconium-stained amniotic fluid missing Total duration of labour (min) First stage missing Second stage Second stage 3 hours Bearing down effort missing 5 (23.8) 16 (76.2) 4 (7.0) 1 (1.8) 5 (8.8) 11 (19.3) 0 (0.0) 11 (21.2) 5 360 (169-610) 263 (137-438) 4 97 (37-184) 15 (28.3) 23 (11-50) 8 10 (40.0) 15 (60.0) 9 (7.9) 1 (0.9) 3 (2.6) 10 (8.8) 2 (1.8) 13 (11.5) 1 271 (113-559) 217 (110-432) 12 48 (16-153) 17 (15.6) 20 (12-34) 9 0.24 1.00 1.00* 0.12* 0.05 0.55* 0.10 0.30 0.46 0.03 0.06 0.17 Oxytocin treatment 41 (71.9) 61 (53.5) 0.02 Uterine inertia Primary Secondary missing 24 (42.1) 9 (15.8) 15 (26.3) 0 39 (34.5) 15 (13.3) 24 (21.2) 1 0.33 0.66 0.46 Epidural anaesthesia 35 (61.4) 53 (46.5) 0.07 Number of vaginal examinations 6 (4-10) 5 (3-9) 0.27 Operative delivery Vacuum extraction Caesarean section (vaginal onset) Foetal presentation Vertex, occiput anterior Vertex, occiput posterior Breech 6 (10.5) 8 (14.0) 51 (89.5) 5 (8.8) 1 (1.8) 3600 (3342-4003) 9 (7.9) 6 (5.3) 105 (92.1) 7 (6.1) 2 (1.8) 3533 (3214-3836) 0.57 0.05 Birth weight (g) 0.78 missing 5 6 Blood loss (ml) 400 (300-800) 300 (250-456) 0.06 Blood loss 1000 ml 8 (14.0) 8 (7.0) 0.14 Blood transfusion 3 (5.3) 2 (1.8) 0.36* Manual removal of placenta 5 (8.8) 1 (0.9) 0.02* Uterine atony 8 (14.0) 8 (7.0) 0.14 Values are given as n (%), means ± standard deviation (SD) or medians (quartiles). Bolded = significance. * Fischer s exact test 16

Crude and adjusted ORs (95% CIs) from the multivariable logistic regression are shown in Table 3. Statistical significance persisted for manual removal of placenta (adjusted OR 11.3 95% CI 1.2-102.4). Table 3. Crude and adjusted odds ratios for intrapartum variables. Characteristics Crude OR (95% CI) Adjusted OR (95% CI) Post-term delivery 42 weeks 3.4 (1.1-10.0) 3.2 (0.9-11.4) Induction of labour 2.1 (1.0-4.2) 1.3 (0.6-3.1) Second stage 3h 2.1 (1.0-4.7) 2.0 (0.9-4.8) Caesarean section 2.9 (1.0-8.9) 7.0 (0.7-68.5) Manual removal of placenta 13.3 (1.6-113.3) 11.3 (1.2-102.4) OR = odds ratio. CI = confidence interval. Cervical cultures were obtained from 51 cases (89.5%). In 24 of the cultures (47.1%) one or more microorganisms were found. Two of the cultures contained two microorganisms each, one with G vaginalis and S aureus and the other one with group G streptococcus (GGS) and S aureus. The other 22 cultures containing microorganisms obtained only one microorganism each. Table 4 shows the panorama of microorganisms. The most frequent microorganism was S aureus, found in 5 cultures (9.8%). The three most frequent microorganisms after S aureus were GBS, GGS and G vaginalis, found in 4 cultures each. Two of the cultures were not tested for G vaginalis. Table 4. Microorganisms of the cervical cultures. Type of microbe n (% of cultures taken) Staphylococcus aureus 5 (9.8) Group B Streptococcus (Streptococcus agalactiae) 4 (7.8) Group G Streptococcus 4 (7.8) Gardnerella vaginalis 1 4 (7.8) Escherichia coli 3 (5.9) Bacteroides fragilis 3 (5.9) Group A Streptococcus (Streptococcus pyogenes) 1 (2.0) Group C Streptococcus 1 (2.0) Clostridium perfringens 1 (2.0) Cultures with any microorganism 24 (47.1) Cultures without microorganisms 27 (52.9) No cervical culture taken 6 1 two isolates were not tested for Gardnerella vaginalis. 17

Discussion Potential predictive factors PPE is the most common obstetric infection 48, contributing to maternal morbidity and mortality. The aim of our study was to evaluate potential predictive factors for PPE in women with vaginal onset of labour. The total rate of PPE was 1.6% in Uppsala in 2017. The incidences were 1.6% after vaginal deliveries, 2.1% after emergency caesareans and 0.3% after elective caesareans. The incidences correlate with rates found in the literature but the difference in rate between vaginal and caesarean delivery is smaller than in other studies. 7,11 Whether the small difference between incidences in our study is due to a high incidence of PPE after vaginal deliveries or if the incidence of post-caesarean endometritis is low in our study population cannot be known for sure. Compared to the incidences in Uppsala in 2015, both incidences are lower 9. The same goes for a cross-sectional study of 1871 women in Denmark from 2007 to 2008, with a comparable antibiotic use during labour and delivery as in our study. In their study, the incidences in total, after vaginal delivery and after elective and emergency caesarean delivery were higher. The correlation between vaginal and caesarean delivery was, as in other studies, larger. 10 However, in comparison to the incidences in a study 7 from Israel in 2000, the incidences in our study are higher. However, in the study from Israel, the use of prophylactic antibiotics was broader. At Uppsala University Hospital, routine antibiotic prophylaxis is used in the event of prolonged rupture of membrane (18 hours or longer) 49, preterm labour (less than 37 weeks) 24, elective and emergency caesarean deliveries 50, intrapartum fever (temperature of 38.0 C or higher) 23 and known GBS colonization of the vagina 24. Our study demonstrates a strong association between manual removal of placenta after vaginal birth and development of PPE. The association persisted in a multivariate analysis with other potential confounding variables (adjusted OR 11.3). The result is in agreement with a previous study of manual placental removal after vaginal birth showing an increased rate of PPE after the procedure (adjusted OR 2.9). 32 Another randomized controlled study of 62 caesarean births presented the same association (OR 8.8, p 0.05). 43 In contrast to our result, a study from 1963 presented no increased risk of PPE after manual removal of placenta. 51 However, in that study, about 30% of the patients with manual removal were administered antibiotics for unstated reasons. During manual placental removal, potential pathogenic bacteria contaminates the uterine cavity which could be one explanation to why it is associated with PPE. However, that would not explain the association with manual removal of 18

placenta in caesarean deliveries reported by McCurdy et al. 43 The potential benefits of antibiotic prophylaxis to prevent PPE after manual placental removal at vaginal birth have not been investigated in any randomized controlled studies, comparing placebo or non-antibiotic use with antibiotic prophylaxis. 46 In the univariate analysis, delivery at 42 weeks or later was associated with PPE and a tendency of an association was seen in the multivariate logistic regression analysis. There were only 15 women with post-term deliveries in our study. The small number of women in this group could be part of the reason to the fact that significance was not reached in the multivariate analysis. There are few studies of the relation between post-term deliveries and PPE but at least one has shown a significant correlation. 40 Caesarean delivery is a well-studied risk factor for developing PPE and antibiotic prophylaxis is given to women with either elective or emergency caesareans. 20 Regardless of the use of prophylactic antibiotics routinely given intrapartum to women with caesarean deliveries in our study population there is still a significant association between caesarean deliveries with a vaginal onset of labour and PPE in the univariate analysis. Significance did not persist in the multivariate analysis. The reason for a potential remaining association despite antibiotic prophylaxis could be that the choice of antibiotic or route of administration are not the most adequate. There are studies comparing different routes of prophylactic antibiotics but most are of low evidence, primarily due to limitations in study design or number of participants. 52 Therefore, there might be an opportunity for future research to optimize the prophylactic treatment. The association between prolonged second stage of labour and risk of PPE found in the univariate analysis, with a median twice as long in the case group, did not persist in the multivariate analysis. Neither did any cut-off value (more than 2, 3 or 4 hours). The association found in the univariate analysis is consistent with a study of perinatal and maternal outcomes associated with duration of second stage in multiparous women, which demonstrated a greater incidence of PPE after a second stage of 2 hours or longer. However, the adjusted OR from the multivariate logistic regression in that study was not significant. 53 Other studies have presented rates of maternal infection or puerperal febrile morbidity in total, without making distinction between different infections. In these studies, significant 19

association was shown also after control for confounding variables in multivariate analysis. 54 56 Both induction of labour and oxytocin infusion were associated to PPE in our study. Since these groups contain almost the same patients, the association to an increased risk for PPE can be due to one or the other even a third interrelated factor. There are no studies reporting an association between induction of labour or oxytocin use and PPE but the studies investigating the possible correlations are few. In a study from the United States from 2013 there are no significant association. 48 Postpartum anaemia are known as a risk factor for PPE 42. Level of haemoglobin was not measured in our study, however intrapartum and postpartum blood loss were. Worldwide, many cut-off values are used to define pathological obstetric bleeding. World Health Organization (WHO) uses the definition of any blood loss in excess of 500 ml the first 24 hours postpartum whereas American College of Obstetricians and Gynecologists (ACOG) defines pathological obstetric bleeding as cumulative blood of 1000 ml or more or blood loss accompanied by signs or symptoms of hypovolemia within 24 hours. 57,58 In our study we used the cut-off value of 1000 ml since the guidelines of treatment of obstetric bleeding postpartum uses ACOGs definition of pathological bleeding and would be the only one of clinical importance. Bleeding of 1000 ml or more were not significantly associated with PPE. However, that might be explained by the small number of women with a bleeding of that volume. Remaining clinical variables investigated in our study were not significantly associated to PPE in the univariate analysis. For some variables the differences in incidences between groups were clinically important although not significant in this material. Other studies have shown significant differences for some of these variables, such as PROM, MSAF and vaginal operative deliveries. 13,37,41 The reason to the disagreements in results could be the small number of participants in our study leading to retaining of a false null hypothesis (type-ii error). Microorganisms in cervical cultures There is quite a variety of microorganisms present in the cervical cultures in our study. The relatively high prevalence of S aureus could be caused by contamination since S aureus is part 20

of the normal flora in most healthy individuals, frequently found in the skin and mucous membranes. 59 The fact that two out of five cultures containing S aureus contained another potentially more virulent bacteria, makes S aureus an even more questionable pathogen even if it is a major bacterial pathogen in humans at other infections. Variations in methods to identify microorganisms could be one explanation to the different results. In this study, we used cervical cultures to isolate potential microorganisms causing infection since endometrial cultures were not obtained from any of the cases. Studies have shown that microorganisms can be isolated from the cervix of 70% to 90% of afebrile women without clinical endometritis. 60,61 Therefore it is uncertain if the bacteria present in cervical cultures are pathogenic or not, especially in cases with low-virulence microorganisms. Eschenbach et al. 60 attempted to investigate whether cervical cultures could predict the presence of microorganisms in endometrial cultures in febrile women. Their study showed that isolation of many facultative bacteria did not predict the presence of the same bacteria in endometrial cultures, including the usually considered potentially virulent E coli. However, isolation of some common virulent facultative microorganisms, including GBS and G vaginalis, and anaerobic bacteria were associated to a relatively high predictive value. This was a small study and therefore, the result should be viewed with caution. The conclusion of the study was, that to isolate microorganisms from the endometrium, endometrial cultures are the most useful culture in women with PPE and that the value of cervical cultures in these cases is limited. The most practical alternative clinically, would be transcervical endometrial cultures. Since cervical cultures were used in our study the results must be considered with caution. However, the panorama of microorganisms found in our study is consistent with other studies of microorganisms present during PPE. Strengths and limitations A strength of this study was that medical charts were scrutinized which enabled for control of a large number of confounding factors and the possibility to investigate multiple exposures. However, despite the usage of multivariate logistic regression to control for potential confounders there might have been remaining confounders not identified in our study and therefore not controlled for. Our study is not without limitations. First, the cases in our study were only women with a diagnosis of postpartum endometritis registered in the Cambio COSMIC healthcare 21

information system. Women seeking medical treatment in clinics without Cambio COSMIC are missing. Fortunately, almost all clinics in Uppsala uses Cambio COSMIC and most women are recommended to seek medical care at Uppsala University Hospital if complications occur in the puerperium. Secondly, the population size is rather small which might make the study underpowered to detect differences between groups, especially for those variables with a small effect size or low prevalence. In a larger study, more variables would probably show significant differences between groups. Thirdly, all data were obtained retrospectively from medical records only. The method used in clinical work is not always the most reliable for measuring the variables of interest. An example is the calculation of duration of labour which is based on results of vaginal examinations. Clinically, vaginal examinations are not performed if it is not necessary. Therefore, the exact moment of start and stop of different stages cannot be certain. The number of vaginal examinations differs and the time interval between is not standardized, depending on many reasons, why the inaccuracy of duration can vary between groups. The event of missing values can also be a problem when data is collected retrospectively, since regaining these values subsequently is not possible. Fourthly, the study was not blinded when data were collected, why bias cannot be excluded. However, the data of exposures were prospectively registered in the medical records, which reduces the possible risk of bias. Most variables have little room for interpretation during collection. There are also some limitations regarding the observations of microorganisms potentially causing PPE. First, because of the use of cervical cultures instead of endometrial cultures, the predictive value of microorganisms present in the endometrium is relatively low for some organisms. The validity of this method might therefore be low. A strength though, is that the predictive value of other organisms associated with PPE are relatively high. Secondly, the part of the study investigating potential microorganisms causing PPE did not have any controls since cervical cultures were not obtained from women in the control group. This is an inevitable error in studies of which data are collected retrospectively since there is no routine of obtaining cervical cultures from women postpartum without signs of infections. The consequence is that you cannot know for sure if the bacteria present in the cervical culture is pathogenic or harmless colonisation of the vagina. 22

Conclusion In this case-control study evaluating associations between clinical variables during pregnancy and delivery and PPE, manual removal of placenta was found independently associated with PPE. Several other clinical variables such as post-term delivery (in gestational week 42 or later), induction of labour, oxytocin treatment, prolonged duration of second stage and caesarean delivery were identified to be associated in the univariate analysis. To enable prevention, knowledge of risk factors of PPE is important and might be helpful in decisions of interventions during labour as well as monitoring postpartum. Despite the importance of PPE and the large number of studies, the microbiology of the infection is still not clearly defined, mostly due to suboptimal methods to isolate bacteria in the endometrium and contamination of cervical, vaginal and skin bacteria. This study is small, and further research is necessary to interpret the results in practice. Larger studies are needed to identify risk factors for PPE and well-designed randomized controlled trials with high power are needed to evaluate the value of antibiotic prophylaxis in special events during vaginal delivery, for example in the event of manual placental removal. Acknowledgement I would like to express my sincere appreciation to my supervisor Maria Jonsson for her support, patience and professional guidance throughout this project. Thank you for sharing your knowledge and experience in science and for your commitment and dedication to this study and obstetrics. 23

References 1. Adair, F. L. The American Committee on Maternal Welfare: Meeting held at Atlantic City, June 12, 1935 Chairman s address. American Journal of Obstetrics and Gynecology 30, 868 871 (1935). 2. Filker, R. & Monif, G. R. The significance of temperature during the first 24 hours postpartum. Obstet Gynecol 53, 358 361 (1979). 3. Mackeen, A. D., Packard, R. E., Ota, E. & Speer, L. Antibiotic regimens for postpartum endometritis. Cochrane Database Syst Rev CD001067 (2015). doi:10.1002/14651858.cd001067.pub3 4. Rimawi, B. H., Soper, D. E. & Eschenbach, D. A. Group A streptococcal infections in obstetrics and gynecology. Clin Obstet Gynecol 55, 864 874 (2012). 5. Hussein, J. & Walker, L. Puerperial sepsis in low-and middle- income settings: past, present and future. Maternal and Infant Deaths. Chasing Millennium Development Goals 4, 131 147 (2010). 6. Acosta, C. D. et al. Severe maternal sepsis in the UK, 2011-2012: a national case-control study. PLoS Medicine 11, e1001672 (2014). 7. Chaim, W., Bashiri, A., Bar David, J., Shoham Vardi, I. & Mazor, M. Prevalence and clinical significance of postpartum endometritis and wound infection. Infectious Diseases in Obstetrics and Gynecology 8, 77 82 (2000). 8. Calhoun, B. C. & Brost, B. Emergency management of sudden puerperal fever. Obstet. Gynecol. Clin. North Am. 22, 357 367 (1995). 9. Sveus. Värdebaserad uppföljning av förlossningsvård: analys från framtagande av nya uppföljningssystem : presentation av historiska resultat för att tydliggöra möjligheter och utmaningar i framtida uppföljning. (Sveus, 2015). 24

10. Ahnfeldt Mollerup, P., Petersen, L. K., Kragstrup, J., Christensen, R. D. & Sørensen, B. Postpartum infections: occurrence, healthcare contacts and association with breastfeeding. Acta Obstetricia et Gynecologica Scandinavica 91, 1440 1444 11. Burrows, L. J., Meyn, L. A. & Weber, A. M. Maternal Morbidity Associated With Vaginal Versus Cesarean Delivery: Obstetrics & Gynecology 103, 907 912 (2004). 12. Bebell, L. M. et al. Antimicrobial-resistant infections among postpartum women at a Ugandan referral hospital. PLoS One 12, (2017). 13. Casey, B. M. & Cox, S. M. Chorioamnionitis and endometritis. Infect. Dis. Clin. North Am. 11, 203 222 (1997). 14. Watts, D. H., Eschenbach, D. A. & Kenny, G. E. Early Postpartum Endometritis: The Role of Bacteria, Genital Mycoplasmas, and Chlamydia trachomatis. Obstetrics & Gynecology 73, 52 60 (1989). 15. Spiegel, C. A., Amsel, R., Eschenbach, D., Schoenknecht, F. & Holmes, K. K. Anaerobic bacteria in nonspecific vaginitis. N. Engl. J. Med. 303, 601 607 (1980). 16. Lamy, C. et al. [Management of post-partum infections]. J Gynecol Obstet Biol Reprod (Paris) 41, 886 903 (2012). 17. Gilstrap, L. C. & Cunningham, F. G. The bacterial pathogenesis of infection following cesarean section. Obstet Gynecol 53, 545 549 (1979). 18. Say, L. et al. Global causes of maternal death: a WHO systematic analysis. The Lancet Global Health 2, e323 e333 (2014). 19. El-Mahally, A. A. et al. Risk factors of puerperal sepsis in Alexandria. J Egypt Public Health Assoc 79, 311 331 (2004). 20. Smaill, F. M. & Grivell, R. M. Antibiotic prophylaxis versus no prophylaxis for preventing infection after cesarean section. Cochrane Database Syst Rev CD007482 (2014). doi:10.1002/14651858.cd007482.pub3 25

21. Fishman, S. G. & Gelber, S. E. Evidence for the clinical management of chorioamnionitis. Seminars in Fetal and Neonatal Medicine 17, 46 50 (2012). 22. Chapman, E., Reveiz, L., Illanes, E. & Bonfill Cosp, X. Antibiotic regimens for management of intra-amniotic infection. in Cochrane Database of Systematic Reviews (John Wiley & Sons, Ltd, 2014). doi:10.1002/14651858.cd010976.pub2 23. Rezapour Isfahani, M. Feber under förlossning (Swedish). Region Uppsala (2017). 24. Rezapour Isfahani, M. Grupp B Streptococcer GBS under graviditet och förlossning. Profylax (Swedish). Landstinget i Uppsala län (2016). 25. Wojcieszek, A. M., Stock, O. M. & Flenady, V. Antibiotics for prelabour rupture of membranes at or near term. Cochrane Database Syst Rev CD001807 (2014). doi:10.1002/14651858.cd001807.pub2 26. Prevention av tidiga infektioner med grupp B-streptokocker (GBS) hos nyfödda (Swedish). Socialstyrelsen (National Board of Health and Welfare). (2008) 27. Verani, J. R., McGee, L., Schrag, S. J. & Division of Bacterial Diseases, National Center for Immunization and Respiratory Diseases, Centers for Disease Control and Prevention (CDC). Prevention of perinatal group B streptococcal disease--revised guidelines from CDC, 2010. MMWR Recomm Rep 59, 1 36 (2010). 28. Bonet, M., Ota, E., Chibueze, C. E. & Oladapo, O. T. Routine antibiotic prophylaxis after normal vaginal birth for reducing maternal infectious morbidity. Cochrane Database Syst Rev 11, CD012137 (2017). 29. Ventola, C. L. The Antibiotic Resistance Crisis. P T 40, 277 283 (2015). 30. Karsnitz, D. B. Puerperal infections of the genital tract: a clinical review. J Midwifery Womens Health 58, 632 642 (2013). 26

31. Meaney-Delman, D., Bartlett, L. A., Gravett, M. G. & Jamieson, D. J. Oral and Intramuscular Treatment Options for Early Postpartum Endometritis in Low-Resource Settings: A Systematic Review. Obstetrics & Gynecology 125, 789 800 (2015). 32. Ely, J. W. M., Rijhsinghani, A., Bowdler, N. C. & Dawson, J. D. S. The Association Between Manual Removal of the Placenta and Postpartum Endometritis Following Vaginal Delivery. Obstetrics & Gynecology 86, 1002 1006 (1995). 33. Newton, E. R., Prihoda, T. J. & Gibbs, R. S. A Clinical and Microbiologic Analysis of Risk Factors for Puerperal Endometritis. Obstetrics & Gynecology 75, 402 406 (1990). 34. Jacobsson, B., Pernevi, P., Chidekel, L. & Jörgen Platz-Christensen, J. Bacterial vaginosis in early pregnancy may predispose for preterm birth and postpartum endometritis. Acta Obstet Gynecol Scand 81, 1006 1010 (2002). 35. Watts, D. H., Krohn, M. A., Hillier, S. L. & Eschenbach, D. A. Bacterial vaginosis as a risk factor for post-cesarean endometritis. Obstet Gynecol 75, 52 58 (1990). 36. D angelo, L. J. & Sokol, R. J. Time-Related Peripartum Determinants of Postpartum Morbidity. Obstetrics & Gynecology 55, 319 323 (1980). 37. Tran, S. H., Cheng, Y. W., Kaimal, A. J. & Caughey, A. B. Length of rupture of membranes in the setting of premature rupture of membranes at term and infectious maternal morbidity. American Journal of Obstetrics and Gynecology 198, 700.e1-700.e5 (2008). 38. Berman, S. M. et al. Low birth weight, prematurity, and postpartum endometritis. Association with prenatal cervical Mycoplasma hominis and Chlamydia trachomatis infections. JAMA 257, 1189 1194 (1987). 39. Ott, W. J. Primary cesarean section: factors related to postpartum infection. Obstet Gynecol 57, 171 176 (1981). 27