Article Text

Original research
Sex differences in the association of pretransfusion haemoglobin and cognition in preterm infants
  1. Amanda M Benavides Mostek1,
  2. Edward F Bell2,
  3. Henry A Feldman3,
  4. Cassandra D Josephson4,5,6,
  5. Michael K Georgieff7,
  6. Peg Nopoulos1,2,8,
  7. Ravi Mangal Patel9,
  8. Sean R Stowell10,
  9. Martha Sola-Visner3,
  10. Amy L Conrad2
  1. 1Department of Psychiatry, University of Iowa Roy J and Lucille A Carver College of Medicine, Iowa City, Iowa, USA
  2. 2Stead Family Department of Pediatrics, University of Iowa Roy J and Lucille A Carver College of Medicine, Iowa City, Iowa, USA
  3. 3Department of Pediatrics, Division of Newborn Medicine, Boston Children's Hospital, Harvard Medical School, Cambridge, Massachusetts, USA
  4. 4Cancer and Blood Disorders Institute, Blood Bank and Transfusion Medicine, Johns Hopkins All Children's Hospital, St Petersburg, Florida, USA
  5. 5Department of Oncology, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
  6. 6Department of Pediatrics, Johns Hopkins University School of Medicine, Baltimore, Maryland, USA
  7. 7Department of Pediatrics, Division of Neonatology, University of Minnesota School of Medicine, Minneapolis, Minnesota, USA
  8. 8Department of Neurology, University of Iowa Roy J and Lucille A Carver College of Medicine, Iowa City, Iowa, USA
  9. 9Department of Pediatrics, Division of Neonatology, Children's Healthcare of Atlanta, Emory University School of Medicine, Atlanta, Georgia, USA
  10. 10Department of Pathology, Joint Program in Transfusion Medicine, Brigham and Women's Hospital, Harvard Medical School, Boston, Massachusetts, USA
  1. Correspondence to Dr Amanda M Benavides Mostek; amanda-benavides{at}


Objectives To assess sex-specific differences in the association between pre-transfusion haemoglobin values and early neurodevelopmental function.

Design Observational follow-up of infants with birth weights <1000 g and gestational ages 22–28 weeks who were enrolled in the NICHD Neonatal Research Network Transfusion of Prematures (TOP) Trial at 19 U.S. sites, 2012–2017.

Main outcome measures Pretransfusion haemoglobin values were obtained longitudinally through 36 weeks’ postmenstrual age. The infant’s mean pretransfusion haemoglobin was used as a marker of degree of anaemia (n=1655 measures). Measures of brain function were obtained at 22–26 months’ corrected age using the Bayley Scales of Infant & Toddler Development, third edition (BSID-III) (n=1290 BSID-III scores). Sex-specific estimates for the linear relation between pretransfusion haemoglobin and BSID-III scores were obtained from repeated-measures regression analysis, adjusted for gestational age, birth weight, study site, clinical characteristics, and demographic covariates.

Results The relation of pretransfusion haemoglobin with 24-month BSID-III scores showed significant, independent interactions with both (1) sex (p=0.046) and (2) retinopathy of prematurity (ROP; p=0.004). In 614 males, BSID-III scores were higher by 1.07 points per g/dL (95% CI 1.58 to 4.33; p=0.008), not differing significantly among the three subscales (cognitive, language and motor; p=0.94). In 247 infants with ROP, BSID-III scores were higher by 2.95 points per g/dL (95% CI 0.28 to 1.87; p<0.0001), uniformly across subscales (p=0.73). These associations were non-significant in 676 females (p=0.96) and 1043 infants without ROP (p=0.81).

Conclusions This study demonstrates sex-specific associations between mean pretransfusion haemoglobin (a marker of the severity of anaemia throughout the neonatal intensive care unit [NICU] hospitalisation) and early neurodevelopmental function at 22–26 months’ corrected age.

  • Neurology
  • Neonatology
  • Infant

Data availability statement

Data are available upon reasonable request. The data includes deidentified participant data from infants who were enrolled and participated in the TOP trial at multiple institutions. The data may be requested from Full inclusion and exclusion information, as well as study protocol, may be found under NICHD Neonatal Research Network Transfusion of Prematures Trial (NCT01702805).

This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See:

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.


  • Large randomised controlled trials investigating comparative effectiveness between two different red blood cell (RBC) transfusion strategies in extremely low birth weight infants have shown no differences in primary gross neurodevelopmental outcomes or survival.


  • There are differences in early brain functioning at 24 months of age in response to haemoglobin level (ie, degree of anaemia) between sexes, as well as between infants that do or do not have retinopathy of prematurity (ROP).


  • Results from our observational study suggest that clinical providers may need to consider sex (and other neonatal factors) when making determinations for RBC transfusions as a treatment for neonatal anaemia in order to optimise neurodevelopment. Further investigations are necessary to validate these findings of sex-specific, distinct neurodevelopmental impairments that may result from different transfusion strategies, as well as elucidate the mechanisms of adverse brain outcomes, in order to provide sex-specific clinical transfusion practice guidelines.


As survival rates among infants born extremely prematurely increase, the identification of treatment modifications that may protect or improve neurodevelopmental outcomes is of clinical importance.1–6 A common therapy for preterm infants is red blood cell (RBC) transfusion to treat anaemia. The degree of anaemia varies under provider’s control, and early life anaemia in multiple populations has been associated with poorer neurodevelopment.7–9 Clinical trials have evaluated the comparative effectiveness of (1) minimising anaemia and maintaining a higher haemoglobin (Hb) level using RBC transfusion versus (2) tolerating more severe anaemia and minimising transfusions, with variable results.10–13 Recently, the Transfusion of Prematures (TOP)14 and Effects of Transfusion Thresholds on Neurocognitive Outcome (ETTNO)15 trials reported no differences in the primary outcome of neurodevelopmental impairment or death before 22–26 months corrected age in extremely low birth weight (ELBW) infants randomised to either a lower or higher Hb threshold for RBC transfusions. These findings suggest that Hb transfusion thresholds within the ranges used in clinical practice are not likely to result in adverse consequences on survival or gross neurodevelopment.16

However, the TOP and ETTNO trials focused on comparative effectiveness, and primary outcome measures were gross neurodevelopmental outcomes (eg, for TOP, defined as Bayley Scales of Infant & Toddler Development, third edition (BSID-III) cognitive score <85, cerebral palsy, vision impairment with acuity <20/200 or bilateral hearing deficits requiring amplification or cochlear implant). Previous studies with follow-up into adolescence have focused on more specific neurodevelopmental outcomes, such as memory, processing speed, language and affective regulation that correlates to brain regions at highest risk from early life anaemia.17 While the effects may be considered subtle, they still cause significant, lifelong impacts on functioning. One study of school-age children found that those who underwent more transfusions in infancy had lower associative verbal fluency, visual memory and reading scores.18 Further investigations found a sex-by-transfusion interaction, whereby more transfusions correlated with lower performance on cognitive measures and with white matter volume reductions in females only.19 20

Measures of cognitive functioning (ie, BSID-III) and brain structure (ie, MRI) were previously obtained on a subsample of the TOP Trial participants at 12 months of age. Using a continuous measure of pretransfusion Hb (mean ptHb), males with higher ptHb had higher white matter volumes, while females with higher ptHb had lower BSID-III scores.21 Additionally, higher number of RBC transfusions was associated with higher levels of monocyte chemoattractant protein 1 (MCP-1; a proinflammatory cytokine) in females only. These increases in MCP-1 were subsequently associated with lower BSID-III scores.22 Taken together, these findings suggest that higher transfusion thresholds in females lead to inflammatory responses affecting brain development, whereas in males, impaired brain outcomes are driven by anaemia.

The objective of this study was to use data from the entire TOP trial to assess whether neonatal ptHb predicted neurodevelopmental outcomes in a sex-specific manner, as suggested by Benavides, et al.21 We evaluated the association of ptHb level (as a continuous marker of transfusion status) with brain functional outcomes using the BSID-III at 22–26 months’ corrected age. All analyses evaluated sex-specific differences.



Infants were enrolled in the NICHD Neonatal Research Network TOP Trial, a 19-site, two-armed randomised clinical trial comparing liberal and restrictive transfusion guidelines based on Hb thresholds in ELBW infants.23 Infants fulfilled the following inclusion criteria: (a) birth weight ≤1000 g and gestational age ≥22 but <29 completed weeks, and (b) <48 hours of age. Exclusion criteria included: (a) considered non-viable by attending neonatologist, (b) cyanotic congenital heart disease, (c) parental opposition to transfusion of blood, (d) parents with hemoglobinopathy or congenital anaemia, (e) fetal transfusion, (f) twin-to-twin transfusion syndrome, (g) isoimmune hemolytic disease, (h) severe acute haemorrhage, acute shock, sepsis with coagulopathy or need for perioperative transfusion, (i) prior blood transfusion on clinical grounds beyond the first 6 hours of life.

Demographic and clinical characteristics recorded by TOP included birth weight, gestational age, neonatal diagnoses (eg, intraventricular haemorrhage (IVH), sepsis, retinopathy of prematurity (ROP)) and severity of illness (Score for Neonatal Acute Physiology with Perinatal extension (SNAPPE score)), as well as race, insurance status and mother’s education. Missing data on maternal education were estimated by multiple imputation for 284 subjects (16%), using a single-chain Markov Chain Monte Carlo procedure based on the following covariates: private insurance, singleton birth, outborn status and black and white race. The mean of 25 imputations was used for further analysis.

Transfusion, blood banking and sampling

A standard transfusion volume of 15 mL/kg was given according to criteria determined by the TOP Trial protocol.24 All RBCs transfused were tested and screened according to the policies of the hospital’s blood bank. Hb values recorded prior to each transfusion (ptHb) were used as a marker of the degree of anaemia prior to transfusion in each infant. Infants were randomly assigned to higher or lower Hb thresholds for transfusion according to the TOP study protocol. TOP Trial infants in the higher transfusion threshold group had mean Hb levels approximately 2 g/dL higher than those in the lower threshold group.

Measures of development

The BSID-III were used to evaluate early cognitive function in these infants. Infants returned for a follow-up visit at approximately 24 months’ corrected age to undergo BSID-III assessment, including the subscale assessments of cognition, fine motor, gross motor, receptive language and expressive language function. Examiners were trained on standardised administration, certified by a gold standard examiner before testing, and recertified after 2.5 years.

Statistical analysis

We employed a single comprehensive repeated-measures model for analysis, incorporating all hypothesised influences. Each infant’s BSID-III scale scores were assembled into a three-entry vector for use as the dependent variable, allowing estimation and accounting for correlation among the scales. The primary independent variable was a single mean ptHb concentration, obtained by averaging all of the levels that triggered the infant’s TOP transfusions. The regression model was sex-stratified and site-stratified and adjusted for covariate effects of gestational age, birth weight, maternal education, severity of illness (SNAPPE score), ROP, sepsis, necrotising enterocolitis (NEC) and IVH. By addition of interaction terms, the model allowed the effects of all covariates to vary among scales, and the effect of ptHb to be modified by all covariates. Non-significant interactions were discarded from the final fitted model. Estimates of the linear trend versus ptHb were calculated from the fitted model, stratified by levels of the significant effect modifiers.

To assess the possibility of bias from missing data, the analysis was repeated with inverse probability weighting.25 In the full sample of 1824 infants, a logistic regression model was constructed using the demographic and clinical covariates to predict an infant’s probability (P) of providing both ptHb and BSID-III. The primary analysis was repeated with data weights 1/P, giving greater influence to those infants most closely resembling those with missing data and thus compensating for the latter’s absence.

Robust regression with bisquare weighting and 95% efficiency were used to identify outliers and reduce the influence of extreme values. P<0.05 was taken as the threshold for statistical significance. SAS software (V.9.4, Cary, North Carolina) was used for all statistical computations.


Demographics and clinical characteristics

Infants in this study were enrolled as part of the multicentre NICHD Neonatal Research Network TOP Trial (NCT01702805).23 Enrolment dates were between 31 December 2012 and 12 April 2017, with follow-up through 3 February 2020. Of 1824 eligible infants, 1655 had at least one TOP-triggered transfusion with a recorded prior Hb measurement. The total number of transfusions was 8133, ranging from 1 to 25 per infant, median 4, mean 5. Of those infants, 1290 provided at least one BSID-III subtest score at the 24-month assessment (online supplemental figure 1). Online supplemental table 1 shows the characteristics of TOP infants not included in this analysis because of missing ptHb or BSID-III assessment. Infants who lacked ptHb—many because they received no transfusion—were generally healthier than those in the analysis sample, as indicated by higher birth weight, lower SNAPPE scores, and less frequent IVH, NEC and sepsis. Those who provided ptHb but no 24-month neurodevelopmental assessment were generally sicker than those in the analysis sample, as indicated by higher SNAPPE scores and more frequent IVH, NEC and sepsis.

Supplemental material

Supplemental material

Table 1 shows demographic and clinical characteristics of the 1290 infants with Hb measurements triggering transfusion under the TOP protocol who also provided BSID-III assessments at 24 months, separated based on sex. There was no significant difference between males and females in gestational age at birth, chronological or corrected age at 24-month assessment, race, maternal education or insurance status. Severity of illness as measured by SNAPPE score did not differ between sexes, nor did incidence of IVH, NEC, adverse head ultrasound findings at 36 weeks or periventricular leukomalacia. The females showed lower mean birth weight and lower incidence of sepsis and ROP.

Table 1

Demographic and clinical characteristics of TOP subsample used in analysis of 2-year neurodevelopmental outcome versus mean pre-transfusion haemoglobin, by sex

Pretransfusion Hb and 2-year neurodevelopment

Mean ptHb levels did not differ between the sexes. Twenty-four-month BSID-III scores were consistently higher, on average, in females (table 2). The three scales were strongly correlated, with pairwise correlations between 0.65 and 0.71 (online supplemental table 2).

Supplemental material

Table 2

Levels of pretransfusion haemoglobin and 2-year neurodevelopmental outcome in TOP subsample, by sex (unadjusted)

Results of regression analysis are shown in two parts: those relating the 24-month neurodevelopmental measures to the infant’s mean ptHb (table 3) and those involving covariates (online supplemental table 3). For each BSID-III subscale, the estimated trends in response to increasing ptHb went in opposite directions for males and females (figure 1). In males, higher ptHb was associated with higher BSID-III scores, by approximately one scale point per g/dL (p=0.008). The effect did not differ across scales (p=0.94). By contrast, ptHb was not significantly associated with BSID-III scores in females. Virtually identical results were obtained in an alternative analysis weighted to compensate for missing data and in an analysis employing only the first transfusion-triggering ptHb for each infant.

Supplemental material

Figure 1

Relation of 24-month neurodevelopmental measures to mean pretransfusion haemoglobin level in 1290 infants enrolled in the Transfusion of Prematures study. Interactions with sex (above) and ROP (below) were mutually adjusted and independently significant. From repeated-measures regression analysis of the 3-component BSID-III, adjusted for multiple clinical and demographic covariates. Trends did not differ significantly among cognitive, language and motor scales (p>0.70); mean of the three scales is shown. Band indicates 95% CI. BSID-III, Bayley Scales of Infant & Toddler Development third edition; ROP, retinopathy of prematurity.

Table 3

Association of 2-year neurodevelopmental outcome with pretransfusion haemoglobin in TOP subsample, by sex and by ROP status

ROP likewise acted as an effect modifier on the relation between ptHb and BSID-III (table 3, figure 1). In infants with ROP, greater ptHb was associated with higher BSID-III scores, by nearly three scale points per g/dL (p<0.0001), an effect not differing significantly across scales (p=0.73), whereas BSID-III scores were not significantly related to ptHb in infants without ROP.

Covariate influences are detailed in online supplemental table 3. Gestational age and IVH affected the subscales differentially, as did maternal education level (all interaction p<0.05). Birth weight, sepsis, NEC and SNAPPE score were significantly associated with BSID-III scores, but their influence did not differ significantly among subscales. These and the abovementioned findings for ptHb, sex and ROP were mutually adjusted and independent.


The current study demonstrates that there are sex-specific differences in brain function at 24 months of age in response to Hb level in a large sample of preterm infants. All BSID-III scores were directly related to Hb level in males only, in whom higher ptHb values were associated with higher BSID-III, while in females, there was no such association. In previous work, a similar pattern of sex differences emerged; females in the higher Hb threshold group and males in the lower Hb threshold group had the worst outcomes in regard to brain structure (volume of white matter) and brain function (cognitive skills).19 20 26 The effect sizes detailed in table 3 are consistent with larger, positive effects on all three components of BSID-III in males, as compared with smaller, negative or null effects in females.

Additionally, the presence or absence of ROP modified the relationship between ptHb and BSID-III, as infants with ROP showed that higher BSID-III scores were associated with higher ptHb, a pattern not seen in infants without ROP. While rates of ROP were higher in the males, these infants also had larger, positive effects of ptHb on their cognitive outcomes. ROP is a developmental disorder of vascular proliferation that occurs in the retina of preterm infants, with systemic inflammatory responses (ie, several proinflammatory cytokines) thought to be an important factor in its pathogenesis.27 28 Therefore, it is possible that inflammatory responses to RBC transfusion may also be playing a role in infants with ROP.

As speculated in a subsample that was assessed at 12 months old,21 a potential mechanism for impaired brain development in males is anaemia, resulting in hypoxia and iron deficiency.29–32 Mouse models of phlebotomy-induced anaemia demonstrate that anaemic males have preferential reduction of synaptic plasticity and cell growth genes in a dose-dependent manner.33 This effect is accompanied by more predominant behavioural abnormalities in males than females.34 Additionally, a randomised trial of placental transfusion strategies reported better fine motor and social scores for males, but not females, undergoing delayed cord clamping (which increases Hb levels and improves iron stores in the first several months of life).35

A limitation of studies investigating neurodevelopmental outcomes after transfusion in human subjects is the inability to determine whether anaemia itself is the cause of the observed deficits in brain function, since most ELBW infants will receive at least one RBC transfusion during their birth hospital stay. However, future work relating biomarkers of brain iron deficiency and oxygenation to outcomes may provide further insight. A recent subanalysis of the Preterm Erythropoietin Neuroprotection Trial (PENUT) showed that vigorous iron supplementation of anaemic ELBW preterm infants who did not receive recombinant human erythropoietin improved neurodevelopmental outcomes, suggesting a role for iron deficiency anaemia in poorer outcomes.36 Based on the data from the current study, the PENUT trial substudy, and the mouse models of phlebotomy-induced anaemia, it is hypothesised that the risk to the brain from anaemia outweighs the potential risk of RBC transfusion in males.33 34 36

The degree of anaemia and neurodevelopmental functioning was not found in females, which is consistent with mouse models where female mice did not show changes to synaptic plasticity genes until reaching a much lower level of Hb concentration than used in the TOP trial.33 Previous work using higher Hb thresholds found that the more transfusions a female receives, the lower her cognitive functioning.19 20 It is hypothesised that higher RBC transfusion thresholds may lead to inflammation in females, thus affecting brain development. Pro-inflammatory cytokine production and endothelial activation after RBC transfusion may underlie several neonatal morbidities.37 38 Two potential factors could be that: (1) white matter in females is more vulnerable to inflammation and/or (2) inflammatory processes are more robust in females. The greater activation of proinflammatory gene pathways in the brains of anaemic female mice lends support to the second factor.33 Neonatal inflammatory responses to RBC transfusion may be different between sexes, with a rise in several cytokines limited to females only,22 and elevations of MCP-1 associated with lower BSID-III cognitive, language and motor scores at 12 months corrected age (a cytokine known to be associated with brain injury in premature infants).39 The extent to which brain inflammation is a direct consequence of anaemia, as opposed to being influenced by anaemia’s impacts on other tissues (eg, the intestine), remains to be defined.40

This study was limited by possible selection bias for participants who were lost to follow-up, unable to provide valid BSID-III scores, or did not undergo RBC transfusion. Participants from the original sample without BSID-III scores were generally sicker than infants included in the study, while those that did not receive transfusion were generally healthier (online supplemental table 1). However, an alternative analysis suggested that the missing participants did not introduce substantial bias.


This study investigated the association of average ptHb, a longitudinal measurement obtained throughout an infant’s hospitalisation, with cognitive functioning at 24 months. Our results showed sex-specific differences in outcomes, with males receiving transfusion at lower Hb thresholds showing lower cognitive outcome scores. Possible mechanisms for these findings include multifactorial effects of anaemia on the male brain as well as possible inflammatory responses in females. Future investigations will be necessary to validate these sex-specific outcome differences by elucidating the mechanisms of adverse brain outcomes in the neonatal period and later in development.

Data availability statement

Data are available upon reasonable request. The data includes deidentified participant data from infants who were enrolled and participated in the TOP trial at multiple institutions. The data may be requested from Full inclusion and exclusion information, as well as study protocol, may be found under NICHD Neonatal Research Network Transfusion of Prematures Trial (NCT01702805).

Ethics statements

Patient consent for publication

Ethics approval

All NICHD Neonatal Research Network (NRN) clinical centres and Research Triangle Institute (RTI) have Institutional Review Board (IRB) committees that convene on a fixed schedule every month to review protocols and associated informed consent forms and data collection procedures for all research to ensure that they are in compliance with all applicable human subject regulations. IRB approval was granted prior to beginning of this study, and study progress and procedures were reviewed by the IRB at least annually. The legal guardians of all subjects were fully informed about the details of this research study. Specific consent forms were developed for each protocol and reviewed and approved by the IRBs at RTI and at each clinical centre. RTI reviewed all forms used at the clinical research sites to ensure that essential elements of consent were presented and complied with federal law. All NRN clinical centres and the Data Coordinating Center (DCC) at RTI have Federal wide Assurance (FWA) by the Office for Human Research Protections (OHRP) of the Department of Health and Human Services (HHS). This FWA is an agreement between each centre, RTI and the U.S. government that all research with human subjects will be conducted according to appropriate federal regulations and allows us to undertake its own Institutional Review Board (IRB) review and monitoring of research with human subjects. Each NRN institution participating in the TOP trial held an FWA, which ensured that the institution's human research activities, overseen by their regulatory authorities, complied with the requirements set forth in 45 CFR 46, as well as the terms of Assurance. The DCC at RTI was responsible for obtaining appropriate clearances at our institution with respect to HIPAA and Human Subjects Research regulations and verifying similar clearances at the clinical centres. Additionally, the Institutional Review Board for the local institution of primary author includes IRB # 201211720—'Transfusion of Prematures (TOP) Trial'—Edward F. Bell, MD (Principal Investigator), and IRB #201211734—'Preterm Transfusions: Brain Structure and Function Outcomes'—Peggy Nopoulos (Principal Investigator). As mentioned above, informed consent was obtained from parents and/or legal guardians for all participants in this study.


We would like to thank the NICHD Neonatal Research Network (NRN) and RTI International for making these data available for analysis.


Supplementary materials


  • Contributors All authors (AMBM, EB, HAF, MG, CDJ, SRS, MS-V, RMP, PN and ALC) have provided substantial contributions to the conceptualisation/design, interpretation of the data, reviewing, editing and approval of the final manuscript and are responsible for all aspects of the reported research. EB, MKG, CDJ, SRS, MS-V and PN substantially contributed to funding acquisition for this study. AMBM drafted the initial manuscript. HAF provided substantial contribution to the methodology and formal analysis of the data. PN and ALC provided direct supervision, oversight and resources for this study and act as guarantors of this study.

  • Funding All phases of this study were supported by grants (U01 HL112776, U01 HL112748 and 5P01HL046925-22, Sub-Project ID 5779 to Drs. Kirpalani, Bell, Das and Sola-Visner) from the National Heart, Lung, and Blood Institute (NHLBI), grants (UG1 HD068244, UG1 HD053109, and U24 HD095254, to the Neonatal Research Network [NRN] site investigators) from the National Institute of Child Health and Human Development (NICHD), grant (5T32MH019113-26) from the National Institute for Mental Health (NIMH) and grant support through cooperative agreements from the NIH, the NICHD, the NHLBI, the National Center for Research Resources and the National Center for Advancing Translational Sciences.

  • Competing interests None declared.

  • Patient and public involvement Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research.

  • Provenance and peer review Not commissioned; externally peer-reviewed.

  • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.