Article Text

Download PDFPDF
A prospective national study of acute fatty liver of pregnancy in the UK
  1. M Knight1,
  2. C Nelson-Piercy2,
  3. J J Kurinczuk1,
  4. P Spark1,
  5. P Brocklehurst1
  1. 1
    National Perinatal Epidemiology Unit, University of Oxford, UK
  2. 2
    Guys and St Thomas’ Hospitals NHS Foundation Trust, London, UK
  1. Dr M Knight, National Perinatal Epidemiology Unit, University of Oxford, Old Road Campus, Oxford OX3 7LF, UK; marian.knight{at}


Objectives: To identify a national, population-based cohort of women with acute fatty liver of pregnancy (AFLP), to evaluate proposed diagnostic criteria and to document accurately the incidence, management and outcomes of the condition.

Subjects and methods: This was a population-based descriptive study using the UK Obstetric Surveillance System, carried out in all 229 hospitals with consultant-led maternity units in the UK. The participants comprised 57 women in the UK diagnosed with AFLP between February 2005 and August 2006 in an estimated cohort of 1 132 964 maternities (women delivering).

Results: The estimated incidence of AFLP was 5.0 cases per 100 000 maternities (95% CI 3.8 to 6.5 per 100 000). Fifty-five cases (90%) were confirmed by diagnostic criteria and clinical assessment, two (3%) by clinical assessment alone, representing 97% agreement (kappa statistic = 0.78). 18% of women had twin pregnancies and 20% were underweight (body mass index (BMI) <20). 60% of women were admitted to intensive care and 15% to a specialist liver unit. One woman received a liver transplant. One woman died (case fatality rate 1.8%, 95% CI 0% to 9.4%). There were seven deaths among 67 infants (perinatal mortality rate 104 per 1000 births, 95% CI 43 to 203).

Conclusions: The largest population-based cohort of women with AFLP to date has been identified. Diagnostic criteria previously proposed agree substantially with clinical diagnosis. The incidence estimate from this study is lower than documented by earlier hospital-based studies, but maternal and neonatal outcomes are better than previously reported, possibly related to improved ascertainment. Women with twin pregnancies appear to be at higher risk, but further studies are needed to investigate the risk associated with low BMI.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Acute fatty liver of pregnancy (AFLP) is a rare but potentially lethal condition of late pregnancy which may be part of a spectrum of disorders related to pre-eclampsia. It remains a cause of maternal mortality in the UK1 and elsewhere.2 Because the condition is rare, it is difficult to study, and the existing literature consists predominantly of small hospital-based case series3 4 or historical cohorts identified retrospectively over a number of years.5 6 Both these study types have limitations. Incidence estimates from these studies vary widely between 1 in 900 and 1 in 16 000 deliveries7 8; maternal case fatality estimates range between 12% and 18%, and neonatal mortality estimates range from 7% to 58%.9 There has been no comprehensive national study of the epidemiology of this condition. Because the evidence base is poor, information available to counsel women with the condition is limited, and guidelines for management are lacking.

The condition is closely related to the syndrome of haemolysis, elevated liver enzymes and low platelets (HELLP). However, widely accepted diagnostic criteria do not exist for AFLP, and the cases included in different studies are not necessarily diagnostically uniform. A set of standard diagnostic criteria has been proposed in a recent UK study,3 but these criteria have not been evaluated in a large series of cases.

The aims of this study were to identify a national, population-based cohort of women with AFLP using the UK Obstetric Surveillance System (UKOSS), to evaluate the diagnostic criteria proposed by Ch’ng et al3 and to document accurately the incidence, management and outcomes of this condition.


We identified cases through the monthly mailing of UKOSS10 between February 2005 and September 2006. Clinicians were asked to report any woman diagnosed with AFLP with symptoms and signs consistent with AFLP, or any woman in whom AFLP was confirmed by biopsy or postmortem.

The UKOSS methodology has been described in detail elsewhere.10 In brief, every month UKOSS case notification cards were sent to nominated reporting clinicians in each hospital in the UK with a consultant-led maternity unit, with a simple tick box to indicate whether they had seen a woman with AFLP. They were also asked to return cards indicating a “nil report“, in order that we could monitor card return rates and confirm the denominator to calculate the incidence rate. When a clinician returned a card indicating a case, they were then sent a data collection form asking for details of disease presentation, management and outcomes. They were also asked to report exclusion of other causes of liver dysfunction such as acute viral hepatitis and paracetamol overdose. All data collected were anonymous.

Data were double entered into a customised database. We assessed cases against the diagnostic criteria for AFLP proposed by Ch’ng et al3 (“Swansea criteria”, see box) to confirm the diagnosis objectively. As the Swansea criteria have not been systematically assessed for their performance in a population case series, two of the authors (CNP, MK) reviewed all the cases to determine independently whether AFLP was the most likely diagnosis.

Additional case ascertainment

To ensure all cases were identified, we independently contacted all radiology departments and liver units, who were asked to report any cases of AFLP, reporting only their year of birth and date of diagnosis. Where a case was identified which had apparently not been reported through UKOSS, the relevant UKOSS reporting clinician was contacted and asked to complete a data collection form. In addition, the Confidential Enquiry into Maternal and Child Health (CEMACH) was contacted at the end of the study and asked to identify any maternal deaths from AFLP occurring during the study period, again providing only their year of birth and date of diagnosis. These were compared with maternal deaths reported through UKOSS. No additional new cases were identified through these three other sources.

Criteria for diagnosis of AFLP3*

Six or more of the following features in the absence of another explanation:

  • Vomiting

  • Abdominal pain

  • Polydipsia/polyuria

  • Encephalopathy

  • Elevated bilirubin (>14 μmol/l)

  • Hypoglycaemia (<4 mmol/l)

  • Elevated urate (>340 μmol/l)

  • Leucocytosis (>11×109/l)

  • Ascites or bright liver on ultrasound scan

  • Elevated transaminases (aspartate aminotransferase or alanine aminotransferase >42 IU/l)

  • Elevated ammonia (>47 μmol/l)

  • Renal impairment (creatinine >150 μmol/l)

  • Coagulopathy (prothrombin time >14 s or activated partial thromboplastin time >34 s)

  • Microvesicular steatosis on liver biopsy

*Figures in parentheses indicate ranges used in the current study

Statistical analyses

Statistical analysis was carried out according to a prespecified study protocol. All analyses were carried out using STATA v9 software.

Rates with 95% CI were calculated using the most recently available birth data (2005) as a proxy for 2005 and 2006,11 with the total number of maternities (women delivering) during the study period estimated from this source as 1 132 964. A kappa statistic (which takes into account chance agreement) was calculated for the comparison of the diagnostic criteria and the clinical assessment of the presence of AFLP.

Ethics committee approval

The UKOSS general methodology (04/MRE02/45) and this study (04/MRE02/71) were approved by the London Multi-centre Research Ethics Committee.


All 229 UK hospitals with consultant-led maternity units contributed data to UKOSS during the study period—that is, 100% participation. All women diagnosed with AFLP deliver in consultant-led maternity units; this study therefore effectively covered the entire cohort of UK maternities during the study period. Ninety-two percent of monthly report cards were returned, and data collection was complete for 97% of cases (fig 1). Additional reports were received from 31% of intensive care units and 26% of liver units, although no new cases were identified through these sources.

Figure 1 Case reporting and completeness of data collection. AFLP, acute fatty liver of pregnancy.

Data collection forms were received for 61 women. Of these, 55 cases (90%) were confirmed by both the Swansea criteria and clinical assessment, and a further two (3%) by clinical assessment alone (table 1). These two cases each met five of the Swansea criteria; the cases confirmed by both assessments met between six and 11 criteria, with a median of eight criteria. Thus a total of 57 cases were defined as confirmed cases. A further four cases were not confirmed by either the Swansea criteria or clinical assessment. The clinical diagnoses in these women were: HELLP syndrome (five of the Swansea criteria met), obstetric cholestasis (five of the Swansea criteria met), pre-eclamptic liver dysfunction (four of the Swansea criteria met) and in one woman the diagnosis was unclear (three of the Swansea criteria met). There was thus 97% agreement between the clinical assessment and the use of the Swansea criteria, with a kappa statistic of 0.78, indicating substantial agreement.

Table 1 Comparison of clinical assessment of acute fatty liver of pregnancy (AFLP) with use of the Swansea criteria in cases reported to the UK Obstetric Surveillance System (n, %)

There were 57 confirmed cases in an estimated 1 132 964 maternities,11 representing an estimated incidence of 5.0 cases per 100 000 maternities (95% CI 3.8 to 6.5).

Diagnosis and management

Characteristics of women with AFLP are shown in table 2. Of particular note, 18% of the women were pregnant with twins.

Table 2 Characteristics of cases

Forty-two women (74%) were diagnosed antenatally, the remainder postnatally (see fig 2). The diagnosis was made at a median gestation of 36 weeks (range 22–40) in women diagnosed antenatally. The majority of women diagnosed antenatally (41/42, 98%) were delivered within 4 days of diagnosis, with most (25/42, 60%) delivered within 24 h of diagnosis. One woman, who had AFLP in a previous pregnancy, was diagnosed 76 days before delivery, although she did not meet the Swansea criteria until the week of delivery.

Figure 2 Gestation at diagnosis (antenatal cases) or delivery (postnatal cases).

The women diagnosed postnatally delivered at a median gestation of 36 weeks (range 28–39), and all were diagnosed within 4 days of delivery. The majority (12/15, 80%) were diagnosed within 48 h of delivery.

Forty-eight women (84%) experienced prodromal symptoms; 34 (60%) had vomiting, 32 (56%) had abdominal pain, 7 (12%) had polydipsia and 5 (9%) had encephalopathy.

The diagnostic features of the women are shown in table 3. Twenty-four (52%) of the 46 women tested had coagulopathy in the absence of thrombocytopenia (platelet count <100×1012). Forty-five women had an abdominal ultrasound scan performed. Ascites or bright liver was seen in 12 women (27%). No woman had a liver biopsy performed. Eleven women had labour induced (19%); 10 of the women diagnosed antenatally and one of the women diagnosed postnatally. Four of the women induced antenatally went on to have caesarean deliveries, two because of fetal distress and two for failed induction. Forty-two women in total were delivered by caesarean section (74%) (table 4), with just over half (23) receiving a general anaesthetic (GA) for the procedure. All three caesarean sections in labour in women diagnosed antenatally were carried out under GA for fetal distress. Eighteen of the prelabour caesarean sections in women diagnosed antenatally were carried out under GA; the indications for GA were fetal distress (1), fetal distress in the presence of maternal coagulopathy (3), maternal coagulopathy alone (12) and maternal AFLP but with no documented coagulopathy (2). Two caesarean sections were carried out under GA in women diagnosed with AFLP postnatally; these were both for fetal distress.

Table 3 Diagnostic features of women with acute fatty liver of pregnancy
Table 4 Mode of delivery and stated indication for caesarean section in women with acute fatty liver of pregnancy (AFLP)


One woman died (case fatality 1.8%, 95% CI 0% to 9.4%). The same woman received a liver transplant; no other transplants were undertaken. Sixteen women were reported to have other severe morbidities, including eight with renal failure (two requiring dialysis) and four who required ventilation. Other co-morbidities reported in individual women were postpartum haemorrhage, septicaemia, a cerebrovascular accident, convulsions and multiorgan failure. Sixty percent of women (34) were admitted to intensive care for a median of 3 days (range 1–8); seven women were subsequently transferred to a specialist liver unit. In total, 10 women (18%) were admitted to a specialist liver unit, where they were treated for a median of 9.5 days (range 4–44).

There were six stillbirths and one neonatal death amongst the 67 infants, giving a perinatal mortality rate of 104 per 1000 total births (95% CI 43 to 203).


This is the largest prospective, population-based study to estimate the incidence of AFLP, from an estimated cohort of >1.1 million maternities. The UK incidence of AFLP as estimated from this study is 5.0 cases per 100 000 maternities, or approximately 1 case per 20 000 births. One other regional population-based study in the UK,13 which included only three cases of AFLP, reported an incidence of 6.1 cases per 100 000 maternities (95% CI 1.3 to 17.9 per 100 000), with which our estimate is compatible. A second, hospital-based, study7 reported an incidence >20 times greater, at 111 cases per 100 000 births (95% CI 51 to 201 per 100 000), based on nine cases of AFLP. The first study was in the context of a project to report a variety of severe maternal morbidities in the South East Thames region, identified through multiple sources including maternity computer databases and labour ward records, as well as staff (principally obstetric and midwifery) reporting. These methods are similar to those we employed. The second study was a prospective study of all pregnant women referred for biochemical testing of liver function in one unit in south west Wales. Each woman or her case records were reviewed by one of two doctors and a diagnosis made using the same criteria as we used. The consequent estimate of disease incidence is much higher than ours and than other estimates. This difference may be explained by the fact that this study was hospital-based; affected women are likely to be referred to this tertiary centre from surrounding district hospitals. Hence the denominator number used to estimate the incidence almost certainly underestimated the true denominator of births from a number of referral hospitals from which the cases were referred and reflects the difficulties of estimating disease incidence from non-population-based sources. It is also possible that more mildly affected women were diagnosed with AFLP in the Welsh study. This is supported by the observed laboratory values; the women in our study had higher levels of transaminases and bilirubin. It is possible that some cases in the UK were not notified to our study; however, we were not able to identify any additional cases through the three alternative reporting sources we used, and we feel it is unlikely that there was significant under-reporting.

Incidence estimates of AFLP from non-UK studies are also based on hospital case series.46 8 14These estimates vary from 1 in 4000 deliveries to 1 in 16 000 births. All are compatible with our estimate due to their small case numbers and hence wide confidence intervals, with the exception of the largest case series,5 a retrospective review of 28 cases from Los Angeles County Hospital, USA between 1982 and 1997. These authors estimated that AFLP occurred in 15 of every 10 000 deliveries in their population (95% CI 10.0 to 21.7). This difference may again be explained by referrals of women from outside the area to this tertiary hospital; questions have also been raised by other authors15 as to whether all the cases included were definitely AFLP.

AFLP was first reported as a unique pregnancy syndrome more than 60 years ago.16 However, there is no clearly established clinical definition to allow consistent comparison across studies and which allows the syndrome to be clearly distinguished from related disorders such as HELLP syndrome. We used the diagnostic criteria proposed by Ch’ng et al3 in Swansea. In our large, national cohort, there was substantial agreement between the clinical diagnosis of cases and the Swansea criteria; we therefore suggest that these diagnostic criteria are adopted as an objective measure for future studies, enabling more consistent interstudy comparisons.

Nearly 20% of the women in this cohort had twin pregnancies. In comparison, only 1% of maternities in England and Wales in 2005 were twin pregnancies;12 we can therefore estimate that women with a twin pregnancy have a >14-fold increase in the risk of AFLP (estimated odds ratio (OR) 14.3, 95% CI 6.4 to 28.6). A similar proportion of women with twin pregnancies has been observed in other studies,2 and an increased risk has also been observed in triplet pregnancies.17 These observations are consistent with the quantified increase in risk of related conditions such as pre-eclampsia observed in multiple pregnancies.18 Although this substantial increase in risk is associated with a rare condition, the observation is of particular importance in the light of increasing multiple pregnancy rates following assisted reproductive techniques in the UK and elsewhere.12

Women with AFLP were also more likely to be primiparous and older than the general population of women giving birth in England and Wales. There was also a suggestion of an inverse relationship between body mass index (BMI) and AFLP. In our cohort, 20% of women were classified as underweight with a BMI <20, and 7% were obese with a BMI of ⩾30. There are no national comparative data on BMI in pregnant women; however, a database analysis of 325 000 women with singleton pregnancies in the North West Thames region of London between 1989 and 199719 reported that 12% of women were underweight and 10% obese. A comparison between these data and ours would give an estimated OR of 1.4 for AFLP in underweight women (95% CI 0.6 to 2.9) and 0.5 in obese women (95% CI 0.1 to 1.6). This relationship has not been observed before, and is the opposite of the relationship observed between BMI and pre-eclampsia.19 20 Because of the rarity of AFLP, our national study covering >1.1 million maternities identified only 57 cases and does not have the power to determine whether this is a statistically significant relationship or just a chance finding. A multinational study would be required to investigate this further.

As would be anticipated, all the women in this series had raised transaminases and bilirubin levels, and 98% also had a raised white cell count. More than 80% had either raised urate or coagulopathy, or both. More than half of the women had abnormal levels of each of blood glucose, platelets or creatinine, or reported prodromal vomiting. Vigil-De Gracia, in a comparison of 75 women with HELLP syndrome and 10 women with AFLP,21 also found low glucose levels, raised bilirubin levels, a raised white blood cell count and prodromal vomiting to be useful distinguishing features of AFLP in comparison with HELLP. Transaminase levels were not consistently significantly different between AFLP and HELLP patients in their study, although transaminitis in AFLP is generally reported to be higher than in HELLP or pre-eclampsia.9 Other investigators have also found coagulopathy to be a useful distinguishing feature of AFLP.22 23 In our cohort, >50% of women tested had coagulopathy without thrombocytopenia, also suggesting that these features may be used to discriminate between the two conditions.

Nearly 80% of the women with AFLP had an abdominal ultrasound examination. Classical features of ascites or bright liver were only seen in a quarter of these. This observation is reflected in other studies which report that hepatic ultrasound is not sufficiently sensitive or specific to make a definite diagnosis.23 Diagnosis by CT scanning was not reported in our study, although this has been recommended where the diagnosis of AFLP remains unclear after ultrasound examination.9 We did not identify any women who underwent MRI scanning; other series have suggested that the role of MRI still has to be defined.24 Similarly, none of the women in our study underwent a liver biopsy. This is unsurprising given the high proportion of women with coagulopathy and current recommendations that biopsy is not undertaken routinely.9

The current recommended management of AFLP is supportive,9 with expedited delivery in women diagnosed antenatally. All the women in our study were diagnosed within 4 days of delivery, with the exception of one woman who was diagnosed very early in pregnancy with recurrent AFLP. This woman was the only woman reported to have recurrent AFLP and, although we cannot calculate a reliable recurrence rate from these data, this suggests the rate is relatively low. There are no current recommendations concerning the mode of delivery in women with AFLP. Our data would suggest that current practice in the UK favours caesarean section delivery. Three-quarters of the women in our study were delivered by caesarean section; >80% of caesarean deliveries were carried out electively before labour onset. This high rate is perhaps surprising given the high rate of reported coagulopathy, but may be a reflection of the perceived severity of the condition. The choice of method of anaesthesia for caesarean delivery is also difficult, with the options of GA which has a potentially negative effect on hepatic encephalopathy versus regional methods with increased risks of haematoma formation in the presence of coagulopathy.25 In this study, half of the women had GA and half had regional anaesthesia for caesarean section. No women who had regional anaesthesia were reported to have had complications of the procedure, including five women who received regional anaesthesia in the presence of documented coagulopathy. Worsening encephalopathy was not reported as a problem in any of the women who had a GA.

The severe morbidity associated with AFLP is reflected in the high proportion of women (65%) who were admitted to an intensive care or specialist liver unit. One in six women was admitted to a regional specialist liver unit. We are unable to estimate what additional proportion was cared for within obstetric high dependency units. However, only one woman died, representing a case fatality rate of only 1.8%, although with relatively small numbers involved the imprecision of this estimate is reflected in the wide confidence intervals ranging from 0% to 9%. There were no other deaths identified through UKOSS or through the UK CEMACH. This case fatality rate is lower than widely quoted rates of 12–18%,9 including a series from a UK specialist liver unit with a case fatality of 13% (95% CI 4% to 29%).26 There are several reasons possible for the better maternal outcome we observed. All of the case series have relatively small case numbers and therefore the confidence intervals for their estimates of case fatality overlap with those estimated from our national cohort study. In addition, as we have previously noted, the majority of studies are hospital-based series from specialist or tertiary referral units; the spectrum of cases seen in these hospitals is likely to be more severe than that of the population as a whole. It is also possible that care for women with AFLP has improved over time, leading to a lower case fatality in our study in comparison with older studies.

The outcome we observed for the fetus was poorer than the maternal outcome, with a perinatal mortality rate of 104 per 1000 births, >10 times the overall national rate. Only one neonatal death occurred, giving a neonatal case fatality rate of 2%, which also compares favourably with quoted rates of 7–58%.9 The majority of the baby deaths in this study occurred antepartum; six infants were stillborn (9%), which is similar to or lower than rates quoted in other studies (7–66%).5 6 Further investigation into earlier diagnosis and delivery may help to determine whether any of these stillbirths are potentially preventable.


We have identified the largest population-based cohort of women with AFLP published to date, in a population of >1.1 million births. The diagnostic criteria proposed by Ch’ng et al3 agree substantially with clinical diagnosis in this series, and we suggest that they are adopted for future studies to ensure diagnostic consistency and thus comparability. The incidence estimate from this study is lower than documented by earlier hospital-based studies, but maternal and neonatal outcomes are better than previously reported. The improvement in outcomes may be due to improved care over time or due to improved case ascertainment and the wider generalisability of the results of this national population-based survey. Women with twin pregnancies appear to be at higher risk of AFLP, but further studies are needed to investigate the risk associated with low BMI.


This study would not have been possible without the contribution and enthusiasm of the UKOSS reporting clinicians who notified cases and completed the data collection forms. We would particularly like to thank Carole Harris who administered the data collection. We would also like to acknowledge the members of the UKOSS Steering Committee who provided advice throughout the study and supplied useful comments on earlier drafts of the paper. The support of the Royal College of Obstetricians and Gynaecologists, Royal College of Midwives, Obstetric Anaesthetists Association, Faculty of Public Health, National Childbirth Trust and the Confidential Enquiry into Maternal and Child Health has greatly contributed to the success of UKOSS.

Appendix 1: UKOSS steering committee

Catherine Nelson-Piercy (Chair), Guys and St Thomas’ Hospital; Jenny Furniss (Vice-chair), Lay Member; Sabaratnam Arulkumaran, Royal College of Obstetricians and Gynaecologists; Jean Chapple, Faculty of Public Health; Cynthia Clarkson, National Childbirth Trust; Andrew Dawson, Nevill Hall Hospital; James Dornan, Royal College of Obstetricians and Gynaecologists; Kate Fleming, Confidential Enquiry into Maternal and Child Health; Shona Golightly, Confidential Enquiry into Maternal and Child Health; Ian Greer, Hull York Medical School, York; Mervi Jokinen, Royal College of Midwives; Gwyneth Lewis, Department of Health; Richard Lilford, Department of Public Health and Epidemiology, University of Birmingham; Mary Mackintosh, Confidential Enquiry into Maternal and Child Health; Margaret McGuire, Scottish Executive Health Department; Richard Pebody, Health Protection Agency; Kate Taylor-Weetman, National Childbirth Trust; Derek Tuffnell, Bradford Hospitals NHS Trust; James Walker, National Patient Safety Agency; Steve Yentis, Chelsea and Westminster Hospital; Members from the National Perinatal Epidemiology Unit: Carole Harris, Marian Knight, Jennifer Kurinczuk, Peter Brocklehurst.



  • MK designed the study, coordinated data collection, coded the data, carried out the analysis and wrote the first draft of the paper. CNP provided clinical input to the study and contributed to the analysis and writing of the paper. JJK assisted with the design of the study, supervised the data collection and analysis, and contributed to writing the paper. PS assisted with data coding, and conducted validation of the data and some analysis. PB had the original idea for the surveillance system, provided advice at every stage of the study and contributed to the writing and editing of the paper. PB will act as study guarantor.

  • Funding: MK is funded by the National Coordinating Centre for Research Capacity Development of the Department of Health. JJK was partially funded by a National Public Health Career Scientist Award from the Department of Health and NHS R&D (PHCS022). The National Perinatal Epidemiology Unit is funded by the Department of Health in England. The views expressed in this paper are those of the authors and do not necessarily reflect the views of the Department of Health

  • Competing interests: None.

  • The UKOSS general methodology (04/MRE02/45) and this study (04/MRE02/71) were approved by the London Multicentre Research Ethics Committee.

Linked Articles

  • Digest
    Robin Spiller Magnus Simren