INTRODUCTION

In October 2012, a healthy young woman, 17 weeks pregnant, went to a hospital in Ireland with a complaint of backache. Clinical examination suggested impending pregnancy loss, which was further borne out when, within 12 hours, membranes ruptured. She requested that delivery be effected, but the request was not honored, because the fetus was still alive, albeit weeks too early for extrauterine viability.

Sixty hours later, she was in septic shock.

Six days after admission, she was dead.¹

Savita Halappanavar was not the only woman to die of obstetric sepsis in the year 2012, but she was one of the few whose name was widely known. The World Health Organization estimates that there are over 40,000 such deaths a year, worldwide,² more than 99% of them in developing regions of the world. Early recognition and appropriately swift care have the potential to drive these numbers lower.

BASIC SCIENCE

“There are known knowns: there are things we know we know. We also know there are known unknowns: that is to say we know there are some things we do not know. But there are also unknown unknowns – the ones we don’t know we don’t know … It is the latter category that tend to be the difficult ones.” (US Secretary of Defense Donald Rumsfeld, 2002)

The scientific understanding of sepsis is incomplete and continues to evolve. Interplay between pathogen and host is complex: the same microorganism that produces local infection in one individual can result in full-blown sepsis in another. This may be attributed to physiologic reserve, comorbidity, social behaviors, genetic determinants, host susceptibility, gene–environment interactions, pathogen–pathogen interactions, prevailing antimicrobial selection pressure, or other factors as yet unspecified.³

Sepsis is not synonymous with either bacteremia or infection, but should be understood as a matter of organ dysfunction which stems from a dysregulated host response to infection. Virtually any organ system can be affected: central nervous system (altered mental status); cardiovascular system (hypotension, myocardial dysfunction, or circulatory collapse); lungs (acute respiratory distress syndrome or ARDS; gastrointestinal system (paralytic ileus); liver (hepatic failure or abnormal transaminases); kidneys (oliguria or acute renal failure); hematologic system (thrombocytopenia or coagulopathy); and endocrine system (abnormal glucose handling, adrenal dysfunction.)

Our current understanding of sepsis pathophysiology includes roles for immune function (both innate and adaptive immunity), the microvascular circulation, the endothelium, and the gut, particularly its microbiome. Host and pathogen factors clearly affect both predisposition to, and recovery from, sepsis. In addition, host–pathogen communication modifies or mediates development of sepsis.

The innate immune system of humans, like that of other mammals, contains pattern-recognition receptors (PRRs), the best known of which are Toll-like receptors (TLRs.) PRRs are keyed to identify molecular motifs known as pathogen-associated molecular patterns (PAMPs), which are absent in higher organisms but expressed by pathogens; these include lipopolysaccharides, components of bacterial cell walls, and bacterial DNA.⁴ PRRs also recognize the host’s own danger-associated molecular patterns (DAMPs), also known as alarmins, such as heat shock proteins, fibrinogen, and high-mobility group box-1 protein (HMGB1), which are released by stressed or dead cells.^4,5,6,7 In response to PAMPs or alarmins, PRRs release a variety of inflammatory signals which are packaged into molecular complexes known as inflammasomes. Activated inflammasomes produce a series of proinflammatory responses, such as upregulating genes that code for cytokines, mediating the release of interleukins, and initiation of apoptosis. In concert with the proinflammatory response, the challenged host must also activate a systemic anti-inflammatory response, which includes epigenetic modification, immune cell regulation, interference with TLR signaling, and a vagally mediated cholinergic pathway.^4,5,7 A successful host response must balance pro-inflammatory and anti-inflammatory limbs in order to clear pathogens without destroying the host’s own tissues and organs.

Pathogen virulence varies in many ways, some of which are affected by host characteristics. For example, molecules produced by a stressed or compromised host are identified by a bacterial quorum sensing system, which then preferentially turns on replication and virulence genes.⁵ Genetic differences play a role in infection and sepsis beyond the obvious factors of age or comorbidities; specific polymorphisms in cytokine and TLR genes have been identified which increase susceptibility to various pathogens, and other polymorphisms in IL-8, protein C, and others increase the risk of organ failure and mortality when infection or sepsis is present.⁸

Disrupted endothelial integrity is a hallmark of sepsis, and barrier failure is mediated by HMBG1, among other DAMPs.⁷ Dysfunctional vascular endothelium predisposes to coagulopathy via effects on tissue factor, thrombin generation, and fibrinolysis.⁹ Mitochondrial dysfunction, failure of intercellular communication, and impaired cellular bioenergetics impede organ function.⁷

Prevailing concepts of pregnancy as an immunodeficient or immunoprivileged state are simplistic and may be incorrect.^10,11 While immune cells in the maternal decidua are altered so as to maintain tolerance of fetal and paternal antigens, this does not translate into a general immuno-incompetence. In fact, pregnancy does not affect B-cell numbers, T-cell numbers, subsets or function, complement, antibody-dependent cellular cytotoxicity, or levels of IgA, IgG or IgM.¹² However, trophoblast may regulate maternal immune cells in their development, differentiation, or function¹¹ and the placenta is known to harbor its own microbiome, which, oddly, resembles the oral microbiome.¹³ It is probably safe to say that as little as we understand about sepsis generally, we understand even less about sepsis in pregnancy.

SEPSIS: DEFINITIONS AND DIAGNOSIS

Terminology related to sepsis has changed in recent years. When reading medical literature, keep in mind that the authors’ definitions may not be those used in current practice. Infection is not synonymous with sepsis. Neither bacteremia nor the outdated term septicemia should be used interchangeably with the term sepsis, though the medical literature on obstetric sepsis often looks at bacteremia instead (see Box 1 for definitions). A significant problem underlying the definition of sepsis is the absence of a clear diagnostic marker: sepsis is currently understood as a clinical syndrome of “physiologic, pathologic, and biochemical abnormalities induced by infection.”¹⁴

An early attempt to codify definitions of sepsis came in 1991, when a consensus conference created four categories of severity, starting with systemic inflammatory response syndrome (SIRS) and proceeding to septic shock.¹⁵ A redefinition following a 2001 conference only served to complicate matters further by adding multiple possibilities for variables which could be used to identify sepsis¹⁶ (see Box 2). Many of these variables were predicated on sophisticated monitoring or laboratory assessment, which limited applicability. In the years following the second consensus conference, a great deal of variability in criteria complicated epidemiology and research, and an increased understanding of molecular mechanisms underpinning sepsis, coupled with a need to distinguish sepsis per se from simple infection, prompted further reconsideration by an expert panel in 2014 and 2015. The current SEPSIS-3 definitions are radically simpler (see Box 1), and categories have been collapsed from four into two: sepsis and septic shock only.

Sepsis has recently been redefined as “life-threatening organ dysfunction caused by a dysregulated host response to infection”.¹⁴ The focus has shifted from signs of infection per se to signs of organ dysfunction or failure. Following release of the SEPSIS-3 guidelines, an expert panel at the World Health Organization arrived at a consensus definition of maternal sepsis in line with current understanding: “Maternal sepsis is a life-threatening condition defined as organ dysfunction resulting from infection during pregnancy, childbirth, post-abortion, or post-partum period”.¹⁷

A full assessment for organ dysfunction is based on the SOFA (sequential organ failure assessment) score: an increase of ≥2 points above baseline identifies organ dysfunction. In a patient not already known to have organ dysfunction, the baseline SOFA score is assumed to be zero (Table 1). However, clinicians outside the ICU may not be familiar with the SOFA score, and since four of the variables require laboratory testing, scoring may not be quick enough for early identification. In response, the Sepsis-3 group developed a quick SOFA (qSOFA) score which can be applied at the bedside.¹⁴ This model assigns 1 point for any of the following:

• Systolic blood pressure ≤100 mmHg;
• Respiratory rate ≥22/min;
• Altered mental status.

A qSOFA score of 2 or 3 does a good job of identifying patients with suspected infection who are at risk of poor outcome, inasmuch as some organ compromise is already present. The predictive value is less than that of the full SOFA score, but the qSOFA score is immediately available at bedside, requires no laboratory assessment, and is easy to repeat. The panel recommended that an elevated qSOFA score should prompt clinicians to further investigate for organ dysfunction, to begin or ramp up therapy, to increase the frequency of monitoring (or consider critical care referral), and in patients not already believed to be infected, to consider the possibility of infection/sepsis as a cause. It is important to note that fever is not necessary or sufficient to the diagnosis of sepsis.

Conventional critical care scoring systems have not had a good track record when applied to obstetric populations. For example, APACHE-II overestimates mortality in these patients by a factor of 4,¹⁸ which reflects the absence of pregnant women from the sample set originally used to develop the prognostic model; the normal physiology of pregnancy is not accounted for. Before uncritically adopting the SOFA and qSOFA scores as screens for sepsis in pregnancy, it would be ideal to validate them in obstetric patients (a task obviously made more difficult by the absence of a diagnostic test for sepsis). It probably makes sense to adjust both the SOFA and qSOFA scores for the physiologic alterations of normal pregnancy. Though mental status and respiratory rate are not affected by pregnancy, other parameters often are. A working group from the Society of Obstetric Medicine of Australia and New Zealand (SOMANZ) has proposed obstetrically modified scores which they call the omSOFA and omqSOFA.¹⁹ The obstetrically modified qSOFA score (omqSOFA) also evaluates three parameters at the bedside and considers a score of 2 or 3 to be abnormal:

• Systolic BP <90 mmHg;
• Respiratory rate ≥25/min (this cutoff was based on regional bedside maternity observation charts);
• Not alert.

With a screening omqSOFA score of 2 or 3, the group recommends that sepsis be considered and further testing for infection and organ dysfunction be undertaken. They also recommend that an omSOFA score ≥2 be understood as organ dysfunction and sepsis should therefore be considered. At the time of this writing (summer 2018) no research has yet been published using either the omqSOFA or omSOFA score, though they seem a promising start.

Septic shock, the second category in the Sepsis-3 guidelines, is defined as “a subset of sepsis in which underlying circulatory and cellular metabolism abnormalities are profound enough to substantially increase mortality.”^14,20 Operationally, this means hypotension (MAP ≤65 mmHg) requiring vasopressors, plus a serum lactate level above 2 mmol/L, despite adequate volume resuscitation. In patients meeting all of these criteria, Sepsis-3 found mortality rates ranging from 35% to 54%.^14,20 Mortality rose in proportion to serum lactate, though higher lactate levels do not necessarily equate to worse tissue perfusion: lactate is a marker of cellular and metabolic stress. Serum lactate levels do not change in pregnancy, and hyperlactatemia does predict higher morbidity among pregnant women presumed to have sepsis.²¹ Identifying the threshold for pathologic hypotension in a pregnant woman is, however, more difficult. In a longitudinal study of blood pressure in 57 normal women during pregnancy, Grindheim et al. determined that the MAP averaged 76–81 mmHg, depending on gestational age, and systolic BP averaged 104–109 mmHg.²² Extending the normal range to 2 standard deviations below the mean, 65 mmHg would thus be a reasonable lower boundary for MAP in normal pregnancy, and 88–92 mmHg for systolic BP. At this time, no one has validated a blood pressure cutoff for septic shock in pregnancy, but the SOMANZ group¹⁹ recommends no change to the MAP cutoff of 65 mmHg that applies outside of pregnancy. Blood pressure in pregnancy is, of course, subject to variation based on maternal position, so it is important to confirm that hypotension is not due purely to vena cava compression when a woman is lying supine. At later gestational ages there may also be a role for the electronic fetal heart rate tracing in the diagnosis of hypoperfusion: this may help the clinician decide whether a given maternal BP is adequate or too low.

EPIDEMIOLOGY OF SEPSIS

Is pregnancy protective or predispositive to sepsis? This is a difficult question to answer, since both the incidence and the mortality of sepsis depend on age, making it difficult to find an appropriate comparison group of women of reproductive age. Analysis of the US National Inpatient Sample for “severe sepsis” hospitalizations in 2007 calculated the incidence as 35/100,000 among adults 18–24, 44 per 100,000 for age 25–29, 66 per 100,000 for age 30–34, 78/100,000 for age 35–39, 118/100,000 for age 40–44, and >800/100,000 for those aged >65.^23, In each age group, the incidence at least doubled between 2000 and 2007, though mortality declined: the overall case–fatality rate of 39.6% in 2000 dropped to 27.3% in 2007. The authors were unable to extract the specific rate for reproductive-age women from this paper. A study of 101,064 patients with “severe sepsis” between 2000 and 2012, drawn from a database of adult ICU patients in Australia and New Zealand, calculated mortality in the subgroup of adults aged ≤44 years (n = 15,471) as 22.1% in 2000, declining to 7.3% in 2012, an impressive 67% drop.²⁴ Survival does not, however, equal intact survival. In the US study, only about 28% of survivors of sepsis were discharged home, while twice as many were discharged to a skilled nursing facility;²³ most of the patients in this study were much older than the reproductive-age group to which this chapter relates. In the Australia/NZ study, 12% of the young adult survivors (≤44) were not discharged home; they were transferred to another hospital or to a rehabilitation facility.²⁴

Epidemiology of sepsis in pregnancy and puerperium

Early work on sepsis in pregnancy relied on case series. Population-based studies are now, fortunately, available. As in sepsis epidemiology generally, there are still questions of case ascertainment and sepsis definition, but uniquely in maternal sepsis, one must make a decision about what to use for the denominator: researchers have used live births, delivery hospitalizations, and estimated total pregnancies. Numbers from large case–control or cohort studies do not line up entirely with coding-based administrative datasets.

In The Netherlands, a national cohort study found the incidence of sepsis in pregnancy and the puerperium to be 2.1 per 10,000 deliveries, between 2004 and 2006,²⁵ with a case–fatality rate of 7.7%. The United Kingdom Obstetric Surveillance System (UKOSS), in a large case–control study, estimated 4.7 per 10,000 maternities were complicated by severe sepsis, including septic shock, between 2011 and 2012.²⁶ Less than one-third of these, however, were admitted to intensive care (ICU), which may reflect either a low degree of severity or a high threshold for ICU admission. Data for the US are derived from large administrative databases, either at the state or national level.^27,28,29 Figures for maternal sepsis drawn from the National Inpatient Sample (NIS) range from 2 to 4 per 10,000 delivery hospitalizations in the early 2000s.^28,29

By linking California vital statistics to hospital discharge data for all in-hospital live births between 2005 and 2007, Acosta et al. calculated a rate of “uncomplicated sepsis”, including “septicemia”, of 5.0 per 10,000, plus another 4.5 per 10,000 who had “severe sepsis”, including septic shock, for a total of nearly 10 cases of maternal sepsis per 10,000 live births.³⁰ There seems to be a real trend toward increasing sepsis rates in the obstetric population. The NIS data showed a doubling in rates between 2003 and 2008.²⁸ A detailed analysis of pregnancy-associated severe sepsis at delivery hospitalization in Texas also demonstrated a doubling in pregnancy-associated severe sepsis, from 6 per 10,000 in 2001 to 12 per 10,000 in 2010, using total estimated pregnancies as the denominator.³¹

In the Texas study, delivery hospitalizations accounted for only about one-third of pregnancy-associated severe sepsis hospitalizations: when abortions and fetal losses were included, the incidence of pregnancy-associated severe sepsis rose from 11 per 10,000 estimated pregnancies in 2001 to 26 per 10,000 in 2010. The authors pointed out that studying severe sepsis only at the time of delivery hospitalization severely underestimates the burden of disease in an obstetric population.³¹ Mortality in this Texas cohort was 9.7% in 2001 and 12% in 2010; of survivors, only 72–78% were simply discharged to home. The remainder required transfer to a long-term care facility, a short-term care facility, or additional care at home. How this plays out for the new mother, the new baby, or the family, is not detailed.

Case series and cohort studies have described patterns among antepartum and postpartum sepsis. Over a 30-year period, 66 women with maternal sepsis were admitted to ICU in a single hospital in a suburb of Paris, two-thirds of whom were antepartum, the majority from a “nonpelvic” source, mostly lung or urinary tract.³⁰ More recently, in two Dublin hospitals between 2005 and 2012, 276 cases of bacteremia were identified among pregnant and postpartum women: 17% were antepartum, 47% were postpartum, and 36% were described as “intrapartum”.³² Among the antepartum cases, 41% were attributed to genital and 46% to urinary sources. Ninety percent of intrapartum cases arose from a genital source; among the postpartum cases, 54% were genital and 25% urinary. It must be pointed out, however, that bacteremia does not equal sepsis, as a significant proportion of clinical sepsis has no positive blood culture, and bacteremia alone is not commonly associated with organ failure; in the Dublin study, only 2.6% of women actually required ICU admission. At the other end of the spectrum of severity, in a series of maternal deaths due to sepsis, 39% of cases were antepartum in onset, though these included first-trimester losses and abortion.³³

Unfortunately, population-based studies do not always separate antepartum from postpartum cases. The few we do have come from several high-income countries. In The Netherlands, 43% of severe maternal sepsis was antepartum, with a case–fatality rate of 11%, compared to the case–fatality rate of 3% for postpartum sepsis.²⁵ These researchers classified causes as obstetric (56%) and nonobstetric (44%): of the nonobstetric causes, urinary tract infection accounted for about 32% and pneumonia for nearly 18%; almost 15% to appendicitis. In the UK, 37% of severe sepsis occurred antepartum, and the risk of death was 1.5%, no different from the postpartum risk of 1.3%.²⁶ Antepartum sepsis was attributed to the urinary tract in 34%, to the genital tract in 20%, and to pneumonia in 9%. In 30%, no source was identified. Postpartum, 37% were from the genital tract, 12% from the urinary tract, 14% from wound infection, about 4% were respiratory in origin, and in 24%, no source was identified.

An audit of 646 pregnant and “recently pregnant” women admitted to intensive care units (ICU) in England, Wales, and Northern Ireland with a diagnosis of severe sepsis or septic shock (using the 2001 criteria) identified respiratory infection as the most common cause overall (approximately 40%); these women also had a longer length of stay in ICU and a higher absolute risk of death, compared with other causes of sepsis.³⁴ Among pregnant women, who accounted for approximately 34% of the sample, leading causes of sepsis resulting in ICU admission were pneumonia, pyelonephritis, and appendicitis, while among the “recently pregnant” (within 6 weeks), the predominant causes were genital tract infection, wound infection, or “septicemia”. Both incidence and mortality of sepsis rose linearly with social deprivation. Overall, the maternal mortality rate was 4.6% and the stillbirth rate 7.2%.³⁴

In the near future, we expect to have a great deal more information about the epidemiology of maternal sepsis. The World Health Organization launched a huge cross-sectional study of maternal (and newborn) sepsis across Latin America, Africa, Europe, Oceania, and Asia, during a single week in November 2017.³⁵ During that week, facilities in participating geographical areas collected information on all women admitted to hospitals and healthcare facilities with suspected or confirmed infection at any stage of pregnancy – antepartum, in all trimesters; intrapartum; and postpartum and postabortal (including miscarriage) up to 42 days after pregnancy ended. Nearly 3000 women are likely to have been included in this observational study, the results of which are expected to be published in 2019. The study investigators hope to be able to validate identification criteria for maternal sepsis as well as address quality improvement strategies for management.

Risk factors for sepsis and for sepsis-related mortality

In the general adult population (not specifically in pregnancy), major risk factors associated with sepsis are age and medical comorbidity.^23,36 Since pregnant and postpartum women are young, studies of younger adults are preferable to studies of the general adult population as a comparison; in a large observational study of sepsis and septic shock from Australia and New Zealand, which included 15,471 adults under age 44, young adults without comorbidities (as defined by the APACHE-II or APACHE-III chronic health classification system) had an 8% risk of in-hospital mortality, compared to a 24% risk for those with comorbidity.²⁴

Most epidemiologic data on sepsis, including mortality, comes from high-income countries, though these make up only about 13% of the world population.³⁷ At the national and cross-national level, there does appear to be both a higher incidence of sepsis and a higher case–fatality rate in low- and middle-income countries (LMIC), compared with high-income countries.^38,39 Cross-national comparisons focused specifically on maternal sepsis are even harder to come by, though the GLOSS study³⁵ is likely to remedy that. In one survey focused on maternal deaths across 40 LMIC, using a national data collection tool called emergency obstetric and newborn care (EmONC) assessments, case–fatality rates for postpartum sepsis were reported to range from a low of ≤1% (Afghanistan, Bangladesh, Cambodia, Burundi, Djibouti, Gambia, Liberia, Namibia, Niger, Ecuador, Guyana, Haiti, Nicaragua, and Panama), to ~5–10% (Angola, Chad, Benin, Eritrea, Ethiopia, Ghana, Guinea, Malawi, Mauritania, Mozambique, Niger, Senegal, Togo), to 15% or higher in Mongolia, Zambia, Congo, and Lesotho.⁴⁰ Data collection is, of course, more limited in those LMIC where the rate of healthcare facility-based deliveries is low.

In high-income countries with near-universal hospital-based deliveries and robust statistical data collection, demographic and medical factors have been studied as contributors to the risk of sepsis in obstetrics. Though many risk factors have been put forward, the literature is not consistent. Drawing from large population-based studies, there seems to be a modest increase in risk (adjusted odds ratio less than 2.0) for nulliparas, African ancestry, age ≥35, and lower socioeconomic status, as represented by low levels of education, residence in a high-poverty neighborhood, and public or no insurance.^{25,26,27,30,31} Obesity was a slight risk factor in one study,³¹ with an odds ratio of 1.4, but not in others.^25,26,27

Specific obstetric factors associated with higher risks for sepsis include:

• Assisted reproductive technology (odds ratio [OR] 5.8);²⁵
• Multiple pregnancy (OR 1.8–6.5) in some studies,^25,27 though not in others;³⁰
• Cervical cerclage (OR 3.4–9.8);²⁷
• Preterm premature rupture of membranes (PPROM) and preterm delivery (OR 2.1–3.0) in some studies,^27,28 but not others;^25,26
• Cesarean delivery, both prelabor and in-labor (OR 1.8–8.1);^25,26,28,29
• Retained placenta or products of conception (OR 4.5) in one²⁷ study, but not in another;²⁶
• Peripartum hysterectomy (OR 56) and blood transfusion (OR 10.9) were associated with significantly increased risk for sepsis;²⁸
• Gestational diabetes was protective where separately analysed (OR 0.54);³¹
• Pregestational diabetes was associated with a range of risks, from no increased risk²⁷ to moderate risk (OR 1.8–4).^28,31

Medical comorbidity played a significant role in sepsis risk, in studies where it was explicitly assessed. Chronic renal disease (adjusted odds ratio [aOR] 33), chronic liver disease (aOR 55), and especially congestive heart failure (aOR 135)²⁷ outweighed all other risk factors, including HIV infection (aOR 3.2–4.2).^27,31

Microbiologic assessment

Information is limited on the pathogens responsible for maternal sepsis, since in many reports, microbiology is not specifically addressed. In the United Kingdom Obstetric Surveillance System,²⁶ which is among the largest and most meticulous reports, the clinical laboratory was only able to identify the microorganism in 64% of maternal sepsis cases, and the clinician could identify the source in only 74%. Remarkably, in 16%, neither the inciting organism nor the source of sepsis was known. These figures are consistent with the overall experience of sepsis in a general adult population, in which blood cultures are negative in two-thirds of patients with sepsis, and cultures from all sites are negative in one-third.⁴¹ The most frequently isolated organisms in maternal sepsis are Escherichia coli, group A and group B streptococcus,^25,26 though staphylococci, Gram-negatives, anaerobes, and many other organisms have been reported.^32,42 Mixed infections are also possible; in 15% of maternal sepsis deaths in which organisms could be identified, infection was polymicrobial.³³

New techniques including molecular techniques are emerging as a supplement to culture-based identification for bloodstream infections. Peptide nucleic acid fluorescent in situ hybridization (PNA-FISH) stains, matrix-assisted laser desorption-ionization/time-of-flight (MALDI-TOF) mass spectroscopy, and polymerase chain reaction (PCR) based systems are commercially available and can provide pathogen identification from blood cultures in under 3 hours from the time cultures are positive.⁴³ Additionally, it is possible to apply molecular techniques directly to whole blood samples, using PCR for the most common Gram-positive and Gram-negative bacteria plus several pathogenic fungi, which could shorten turnaround time still further.⁴³ Concordance between whole-blood PCR and blood culture based techniques is nowhere near 100% at this time, however; detection failures by PCR probably relate to the choice of organisms included in the panel. Interestingly, about 11% of patients with a clinical suspicion of bacteremia but negative blood cultures are reported to have a positive PCR.⁴⁴ Adding molecular and other new techniques to culture-based testing would likely reduce the number of cases in which no organism is identified in suspected sepsis.

Altered pregnancy physiology and the recognition of sepsis

Normal human pregnancy is a state of expanded plasma volume, increased cardiac output, and peripheral vasodilation; these are also features of sepsis. In an earlier era, when the Swan-Ganz catheter was routinely employed in severe sepsis, the hemodynamic picture of sepsis was so similar to that of normal pregnancy that obstetricians had to caution intensivists about jumping to conclusions.

None of the extant definitions of sepsis were modeled on pregnant patients, and they do not account for the alterations in normal physiology which are characteristic of pregnancy. When nonpregnant norms are used, either overdiagnosis or underdiagnosis of sepsis may occur. An analysis of normal maternal physiologic parameters⁴⁵ compared to the 1992 sepsis criteria¹⁵ showed that SIRS and sepsis cutoffs for respiratory rate, heart rate, PaCO₂ and white blood cell count overlapped with the normal range for pregnancy, labor, and/or the early puerperium, which vitiates their value in diagnosing sepsis in obstetrics. The high false-positive rate is one problem, but additionally, the obstetrician may under-react to sepsis signals, being accustomed to a degree of tachycardia or leukocytosis in normal pregnancy.

Recognition of sepsis

A high index of suspicion is needed to recognize sepsis in pregnancy and the puerperium. Because there is no diagnostic test, expedited clinical recognition is key. Infection may be recognized and sepsis nevertheless missed; or the patient may be identified as having organ dysfunction but an underlying infection may go undiagnosed. To reiterate the WHO definition: “Maternal sepsis is a life-threatening condition defined as organ dysfunction resulting from infection during pregnancy, childbirth, post-abortion, or post-partum period”.¹⁷ The full SOFA score is shown in Table 1, the qSOFA score in Table 2, the obstetrically modified SOFA score in Table 3 and the obstetrically modified qSOFA score in Table 4. Using either the qSOFA score from the Sepsis-3 investigators,¹⁴ or the omqSOFA score proposed by the SOMANZ group¹⁹ (Tables 2 and 4), pregnant and postpartum patients should be screened at the bedside for sepsis. Though fever is often related to infection, it is neither sensitive nor specific in sepsis. Early warning systems have been adapted for use in maternal care and can provide a formal review of maternal status, repeated at regular intervals.^46,47,48

If a patient screens positive for sepsis with a rapid bedside assessment, further evaluation is indicated, specifically, a careful search for infection and organ dysfunction. History, physical examination, and laboratory testing should be deployed in a focused way. Cultures should be obtained from blood and from any suspected source, but the process should not be allowed to delay initiation of antimicrobials, as delay in therapy increases mortality.

TREATMENT

Antibiotics should be initiated as soon as possible, ideally within the first hour after sepsis is suspected or diagnosed. The Surviving Sepsis Campaign (SSC) (http://survivingsepsis.org), a project of several international organizations, has released several iterations of guidelines since 2004. Earlier guidelines had both a 3-hour and a 6-hour bundle of practices to implement, but the most recent set, released in 2018, is explicit in recommending a 1-hour bundle.⁴⁹ Recognition, resuscitation, and treatment should be initiated immediately. The components of this bundle are shown in Box 3.

Delay in starting antimicrobials increases mortality from sepsis.⁵⁰ Although the relationship between delay and death has not been tested specifically in an obstetric population, there is no reason to suspect it would be different. In a review of 50 pregnant or postpartum women admitted to ICUs in New Zealand with a diagnosis of “severe sepsis”, 50% of cases were judged to have been preventable, most of them owing to delay in treatment; delay in recognition or diagnosis also was a factor.⁵¹ Mortality from sepsis rises linearly with each hour of delay. Once cultures have been obtained and antibiotics started, the clinician should begin a search for a locus of infection which may be amenable to source control: for example, an infected wound should be opened, retained products of conception should be evacuated, etc.

The choice of antibiotic depends on presumed source, likely microorganism, local patterns of antibiotic resistance, and the hospital’s stock of agents, but should always be broad spectrum to start. The Infectious Disease Society of America convened a working group in 2016 to make recommendations (not specific to pregnancy) as to appropriate initial empirical treatment of bacterial sepsis;⁵² the group offered options for both monotherapy and combination therapy (see Boxes 4 and 5). Suggestions for monotherapy included third- to fifth-generation cephalosporins, piperacillin-tazobactam, or carbapenems. The working group recommended, as combination therapy, that either an aminoglycoside or aztreonam be given for Gram-negative coverage, and an additional agent (e.g., a penicillin, glycopeptide, linezolid, daptomycin) to cover Gram-positive and anaerobic bacteria (see Box 5).

Other specialty organizations have made antibiotic recommendations specific to antepartum or postpartum sepsis, or at least to specific types of maternal infection. The Royal College of Obstetricians and Gynaecologists (RCOG) recommends piperacillin/tazobactam, or a carbapenem plus clindamycin, as broad coverage in postpartum sepsis⁵³ (Box 6). The Society of Obstetric Medicine Australia and New Zealand (SOMANZ) offers recommendations for treatment of maternal sepsis with unknown source, divided into those cases that are community-acquired and those that are hospital-acquired;¹⁹ the recommendations in New Zealand are the same either way, while in Australia they are different (Box 7).

The World Health Organization (WHO) rather tepidly suggests ampicillin plus once-daily gentamicin for suspected chorioamnionitis, and clindamycin plus gentamicin for postpartum endometritis, though they do not offer recommendations for maternal sepsis per se (Box 8).⁵⁴ Nor has the American College of Obstetricians and Gynecologists published specific recommendations for antibiotic treatment of maternal sepsis, though they have weighed in on intra-amniotic infection generally⁵⁵ (Box 9).

Antibiotic coverage should be narrowed and focused once culture results are available, which in most cases will take several days. The duration of treatment will depend on patient response and on the particular source or pathogen involved.

Source control is an important tool for addressing infection.⁵⁶ If the host inflammatory response is successful in isolating a bolus of microorganisms, along with debris from tissue fluid, coagulation elements, and host inflammatory cells, within a fibrin capsule, the abscess thus formed locks pathogens away from systemic dissemination, but also limits the ability of host defenses to attack those pathogens. Local tissue may become necrotic. Uncontrolled abscess rupture carries a great risk of morbidity.

A diligent search for a focus of infection amenable to control should be carried out when sepsis is suspected. Physical examination should be supplemented, where needed, with appropriate diagnostic imaging. Source control is generally predicated on abscess drainage or debridement of necrotic, infected tissue, but in cases of obstetric sepsis with a uterine source, it will specifically be necessary to empty the uterus: delivery of the fetus and placenta is indicated. Postpartum (and post-abortal), it is important to ensure that no placental tissue or infected clot has been retained. With maternal sepsis from a nonuterine source, it is generally preferable to undertake source control via the intervention with the least likelihood of physiologic derangement; thus percutaneous abscess drainage, for example, is preferable to laparotomy. In under-resourced settings, general anesthesia and an open surgical procedure may not even be feasible in a timely fashion.

Fluid management in sepsis

Fluid resuscitation is the first intervention if hypotension or hypoperfusion is present. Fever, venodilation, and capillary leakage all tend to leave the septic patient inadequately preloaded. The Surviving Sepsis Campaign (SSC) recommends an initial bolus of 30 ml/kg of a balanced isotonic salt solution,⁴⁹ but this recommendation may be overly aggressive in pregnancy, where colloid oncotic pressure is lower and the risk of pulmonary edema is higher. Only about 50% of hypotensive septic patients are fluid responders in any case, and in those who are not, aggressive fluid loading may produce diastolic dysfunction, pulmonary edema, and higher mortality.⁵⁷ Point-of-care ultrasound has been used to identify fluid responsiveness either by measuring the diameter of the inferior vena cava (IVC diameter less than 1.5 cm predicts fluid responsiveness, while IVC diameter >2.5 cm suggests the patient is already fully fluid-loaded) or by measuring extravascular lung water.⁵⁸ These tools have not been extensively evaluated in pregnancy, though obstetricians are well aware of the potential for IVC compression by a gravid uterus when a pregnant patient is supine, and reflexively will effect left uterine displacement when she is hypotensive in this position.

A quick and reversible test of fluid responsiveness can be performed with passive leg raising to 45 degrees: within a few minutes after passive leg raising, fluid responders can be identified by hemodynamic improvement, while those who do not improve are probably better treated with vasopressors.⁵⁹ Limited data are available on this maneuver during pregnancy, however, and it may be technically difficult to execute or may not result in an augmentation of cardiac output in late pregnancy.⁶⁰

Early goal-directed therapy (EGDT) in sepsis generated some enthusiasm in the early years of the 21st century. This required a central venous catheter be inserted and aggressive fluid resuscitation to a target central venous pressure 8–12 mmHg, accompanied by aggressive transfusion to hematocrit of ≥30%, and dobutamine as needed to reach a goal measure of mixed venous oxygen saturation.⁶¹ It was never tested in pregnant patients, and has since been demonstrated to worsen outcomes in adult sepsis generally.^64,62,63 EGDT has no place in contemporary management of sepsis in obstetric patients.

The purpose of vasopressors is to constrict the pathologically dilated systemic circulation commonly seem in sepsis, and thereby maintain adequate perfusion. SSC recommends norepinephrine as the first-line agent and does not specify a hard target for mean arterial pressure (MAP), instead suggesting it be individualized.⁴⁹ Intensivists often aim for a MAP of ≥65 mmHg, but this may be too high for a previously healthy young patient, and is probably too high in pregnancy. Values obtained from Grindheim²² in healthy pregnant subjects show 2 standard deviations below the mean to be about 65 mmHg, and 3 SD to be about 60 mmHg. As it happens, many pregnant patients have another easily accessible measure of perfusion available, as the fetal heart rate tracing is exquisitely sensitive to uteroplacental perfusion. Determining the target MAP in a septic pregnant patient must be individualized, with reference to overall organ perfusion, including (though not limited to) the fetal heart rate tracing.

Norepinephrine has been studied in human pregnancy and is used to maintain BP under regional anesthesia for cesarean.⁶⁵ Acknowledging that data are limited in the setting of pregnancy-associated septic shock, norepinephrine nevertheless seems to be safe for fetuses, at least in low doses, in contrast to ongoing maternal shock, which is demonstrably unsafe. Additional reassurance as to its safety profile comes from a dual-perfused single-cotyledon model of human placenta, which showed that norepinephrine did not interfere with perfusion on the fetal side.⁶⁶ The clinician should not hesitate to administer norepinephrine to a septic pregnant patient when it is otherwise indicated.

OBSTETRIC AND PERINATAL/NEONATAL OUTCOMES AFTER MATERNAL SEPSIS

Preterm delivery is common after critical maternal illness, including sepsis, even when the source is not uterine. This is consistent with the pathophysiology of sepsis, in which inflammatory mediators are released systemically,^4,6,7 and one might speculate that these mediators contribute to triggering labor. In an Irish series of pregnant and postpartum women with bacteremia – described as sepsis, but given low rates of organ failure and ICU admission, these women were less sick than we would currently understand as sepsis – the rate of preterm birth was 16.8%, nearly 3 times the rate in the control group at the institutions.³² This rate included women diagnosed with bacteremia antepartum, intrapartum, and postpartum. Examining only those women with antepartum bacteremia, 69% either miscarried or delivered preterm. The outcome was, as expected, even worse for women with antepartum bacteremia of uterine origin: all delivered within 24 hours of onset. Among women with a nongenital source of bacteremia in the antepartum period, it appears that 12% miscarried, 33% delivered soon after onset, and the remainder delivered between 1 week and 7 months after onset, but the exact details cannot be obtained from the paper. Bacteremia during pregnancy was associated with a 29% risk of preterm delivery in a French study, with an overall 10% rate of fetal mortality; when maternal bacteremia occurred during the second trimester, the fetal death rate was 40%.⁶⁷ In a small but careful study focused specifically on E. coli bacteremia during pregnancy, the same researchers determined the rate of fetal death was 27% overall, despite adequate antibiotic therapy, and even when the microbiologic virulence score was low.⁶⁸ In both these papers, maternal condition seems overall to have been good, with low rates of ICU admission and organ failure, and no deaths. One would anticipate that in women who are overtly septic, by current definitions involving organ failure, maternal and perinatal outcomes would be considerably worse.

Long-term outcomes

In the longer term, adults who have survived sepsis continue to have higher mortality rates than those who did not have sepsis.⁶⁹ There are no data available on long-term maternal outcome after critical illness, including sepsis. Quality of life is impaired among sepsis survivors.⁶⁹ Ongoing weakness, myopathy, and neuropathy are common among ICU survivors, as is psychological distress, e.g., depression and post-traumatic stress disorder (PTSD). More than 50% of sepsis survivors have a degree of long-term cognitive impairment,⁷⁰ which should be borne in mind when planning hospital discharge. Almost nothing is known about recovery from critical illness during pregnancy, though a recent qualitative study points up weakness, frustration, isolation, anxiety, panic attacks, depression, and PTSD after the critical-illness hospitalization.⁷¹ Critical maternal illness also affects families and caregivers in ways which are only beginning to be evaluated. There is a great need for more research on these outcomes.

PRACTICE RECOMMENDATIONS