The Development of a Clinical Prediction
Tool to Support Clinicians in the
Assessment of the Risks of Fetal Asphyxia
and Failure to Progress in Term
Pregnancies
Uloma C Ogba
April 2015
The Development of a Clinical Prediction
Tool to Support Clinicians in the
Assessment of the Risks of Fetal Asphyxia
and Failure to Progress in Term
Pregnancies
Student Uloma C Ogba Collegekaart nummer: 10021671 E-‐mail: u.ogba@amc.uva.nl Mentor Sabine Ensing, MD AMC
Obstetrics and Gynecology/Department of Medical Informatics s.ensing@amc.uva.nl
Tutor
Anita Ravelli, PhD
Department of Medical Informatics, AMC-‐UvA a.c.ravelli@amc.uva.nl
Location of Scientific Research Project
Department of Medical Informatics/Department of Obstetrics and Gynecology AMC-‐UvA
Meibergdreef 15 1105 AZ Amsterdam
Practice teaching period October 2014-‐April 2015
Acknowledgments
I would like to express my profound gratitude to all those who made it possible for me to complete this thesis project successfully. My deepest and sincerest thanks to my mentor Sabine Ensing without whose supervision, guidance and input none of this would have been possible. From the inception till the completion of this research project Sabine helped me to clarify my research objectives, she was always available to provide help or feedback when I needed it and her warm and encouraging demeanor helped me maintain a positive attitude and work diligently throughout the entire process.
I would also like to thank Anita Ravelli and Ameen Abu-‐Hanna who were available to provide feedback and support when I needed it.
The development of the clinical prediction tool would not have been possible without the input of Ewoud Schuit, who conceived the idea for the project a while ago and also provided invaluable assistance while writing the code to develop the tool.
I am also grateful to the staff and support system of the Medical Informatics department at the AMC and to my fellow graduate students. All the tips, advice and suggestions that I have received during my time at the AMC helped me to navigate the Medical Informatics master’s program successfully.
Lastly, I could never thank my family enough for all their love and support, which has sustained me throughout my time in Amsterdam-‐ Leo, Ola, Okechukwu, Ndiya and Kachi, I am the luckiest girl in the world to be part of such an amazing and supportive family. And also to my best friends Cindy and Hodan, you mean the world to me, thank you for everything. To the rest of my friends and family in Amsterdam and abroad, thank you for always being there for me and for believing in me.
Contents
Summary 5
Samenvatting 6
1 Introduction 7
1.1 Objective and background information……….7
1.2 Outline of this thesis………...8
Bibliography………9
2 Prognosis and Prognostic Models in health care 10 2.1 Introduction………...10
2.2 Prognosis and Prognostic models………...10
2.3 Development of prognostic models………...…11
2.4 Performance of prognostic models: discrimination and calibration………12
2.5 Internal and external validation of prognostic models……….12
Bibliography………...………...13
3 Labor and its potential adverse outcomes; fetal asphyxia and failure to progress 14 3.1 Introduction………...14 3.2 Methods………...15 3.3 Results……….15 3.4 Discussion………...…………..20 3.5 Bibliography………..………….20 4 Evaluation of the existing prognostic models for fetal asphyxia and failure to progress 23 4.1 Introduction………..…….23 4.2 Methods……….…..24 4.3 Results……….24 4.4 Discussion……….27 Bibliography……….…………28 5. Development of a clinical prediction tool to support clinicians in the assessment of the risks of fetal asphyxia and failure to progress in term pregnancies 30 5.1 Introduction………..….30 5.2. Methods……….…….31 5.3 Results……….34 5.4 Discussion ………38 5.5 Conclusion………42 Bibliography……….……42
6 Discussion, conclusions and future recommendations 45
List of Abbreviations 43
Appendices
The development of a clinical prediction tool to support clinicians in the assessment of the risks of fetal asphyxia and failure to progress in term pregnancies.
Objective
During labor, problems can occur in the child (fetal asphyxia) or in the mother (failure to progress). Both problems can occur simultaneously. However, at present clinicians base their clinical decisions typically on one of these problems, rather than integrating both dimensions. The aim of this study was to develop a clinical prediction tool that simultaneously assesses the risk of failure to progress (FTP) and fetal asphyxia and can be applied in both primary and secondary obstetric care settings.
Methods
To develop the prediction models for fetal asphyxia as and failure to progress, data on term singleton pregnancies from the Perinatal Registry of the Netherlands (PRN) between 2000 to 2010 were used. Bootstrapping techniques were used for internal validation. Discrimination (AUC) and calibration (graph, c-‐statistics) were used to assess the predictive performance of both models.
Results
Two prediction models: one for fetal asphyxia and one for obstetric intervention due to failure to progress were developed. In a summary graph, the predicted probabilities of fetal asphyxia versus the predicted probabilities of an intervention due to FTP and the number of women within each 10th percentile combination of both outcomes were shown. The probability of fetal asphyxia varied between 0.1% and 13%, whereas the probability of an obstetric intervention due to failure to progress varied between 0.3% and 100%. Overall, the chance of an intervention due to FTP increased with an increased risk of fetal asphyxia. However, in some cases the risk of FTP is high while the risk of fetal asphyxia is low and vice versa. To aid clinicians in the use of the tool we suggested a list of eight interventions (in increasing order of the number of women within each 10th percentile combination) for various combinations of a range of predicted probabilities for both outcomes.
Conclusion
In women with a singleton term pregnancy in cephalic presentation a graph combining the risks of fetal asphyxia and an intervention due to FTP could aid clinicians in the choice of interventions during labor and delivery in both primary and secondary obstetric care settings.
Key words: fetal asphyxia; failure to progress; prediction models Samenvatting
De ontwikkeling van een klinisch predictie instrument om clinici te ondersteunen bij het bepalen van de risico’s op feutale asphyxie en stagnatie van de bevalling.
Doel
Tijdens de bevalling kunnen verschillende problemen optreden zoals featale asphyxie of stagnatie van de bevalling. Beide problemen kunnen tegelijkertijd voorkomen. Echter clinici baseren hun klinische beslissing vaak op een van deze twee problemen in plaats van beide dimensies te integreren. Het doel van deze studie is het ontwikkelen van een klinisch predictie instrument dat zowel het risico op feutale asphyxie als stagnatie van de bevalling voorspelt.
Methode
Met gebruik van data uit de Nederlandse prenatale registers (PRN) hebben we twee predictie modellen ontwikkelt om feutale asphyxie en stagnatie van de bevalling te voorspellen. Op basis van deze twee predictie modellen hebben we een tweedimensionaal predictie instrument ontwikkelt waar voor iedere bevalling zowel de kans op feutale asphyxie als stagnatie van de bevalling weergegeven wordt op een tweedimensionale schaal.
Resultaten
We lieten de predictieve kans op feutale asphyxie versus de predictieve kans van een interventie door stagnatie van de bevalling en het aantal vrouwen binnen de combinatie van beide uitkomsten van elk 10de percentiel, zien. De kans op feutale asphyxie varieerde tussen de 0.1% en 13%, waar de kans op een obstetrische interventie door stagnatie van de bevalling varieerde tussen de 0.3% en 100%. De kans op een interventie door stagnatie van de bevalling nam toe bij een toenemend risico van feutale asphyxie. Echter, in sommige gevallen is het risico van stagnatie tijdens de bevalling hoog terwijl het risico op feutale asphyxie laag is en vice versa. Om de clinici te ondersteunen in het gebruik van het instrument hebben we een lijst van acht interventies opgesteld (in toenemende volgorde van het aantal vrouwen binnen de combinatie van elk 10de percentiel) voor verschillende combinaties van een reeks predictieve kansen voor beide uitkomsten.
Conclusie
Bij vrouwen met eenling zwangerschap die zich presenteren in een eerste of tweede lijn obstetrische (verloskundige instelling, obstetrie) zorginstelling, kan een grafische weergave van de risico’s op feutale asphyxie en het inzetten van een interventie door het stagneren van de bevalling als bruikbaar instrument worden ingezet door de clinici om deze te begeleiden in de keuze voor het inzetten van een interventie tijdens de bevalling.
Kernwoorden: feutale asphyxie; stagnatie van de bevalling; predictie modellen
Introduction
The Scientific Research Project (SRP) is a mandatory part of the Master program of Medical Informatics, and results in a Master thesis. Its goal is to develop the student’s scientific problem-‐focused approach and to improve their ability to pursue lifelong learning. Students will train in undertaking critical assessment of scientific biomedical and biomedical informatics literature, formulating clear research questions, and independently resolving information problems in biomedicine and reporting on it. This thesis describes the SRP called ‘The development of a clinical prediction tool to support clinicians in the assessment of fetal asphyxia and failure to progress in term women”, performed at the Department of Medical Informatics in conjunction with the Department of Obstetrics and Gynecology at the Amsterdam Medical Center (AMC). In this section, first we will describe the goal of this SRP and provide some background information, next we will state the research questions to be answered by this SRP and finally we will provide the outline for this thesis.
1.1 Objective and background information
Prognosis is defined as the prediction of the future course or outcome of disease processes [1,2]. For example it might be important to raise the question: what are the risks of maternal morbidity and fetal mortality during labor? Because of various sources of uncertainty it is usually more useful to estimate the probability of an event, such as mortality, than to predict the event itself. Prognosis is required, among other options, for making informed treatment decisions, and clinicians are constantly prognosticating events although they might not always be conscious of this fact. While some clinicians rely on intuition to arrive at a prognosis, others rely on some formal or semi-‐formal instrument such as a risk-‐calculator. Some of the approaches for outcome prediction are subjective, relying on the clinicians’ assessments, while others rely on models based on collected data.
In obstetric care, the caregiver constantly needs to weight the risk of an adverse pregnancy outcome i.e. neonatal and maternal morbidity or even mortality, when deciding between continuing labor versus obstetric interventions. Two main adverse outcomes that can occur are failure of labor to progress (in the woman) and fetal asphyxia (in the child). Both outcomes can occur simultaneously. In both cases, an intervention like a cesarean section or an assisted vaginal delivery might be necessary. Recently, two prognostic models were developed to predict the risk of fetal asphyxia [3] as well as non-‐progressive labor [4,5]. These prediction models were developed on Dutch cohorts of high-‐risk pregnancies and the information gleaned from their analyses will serve as the basis for this research project.
During this traineeship we will develop two new models to predict the risk of fetal asphyxia as well as non-‐progressive labor, based on a thorough understanding of the prediction models that have already been developed and also by including low-‐risk populations and incorporating new registered variables we think may also be candidate predictive factors for both outcomes. The aim of this traineeship will be to develop a two-‐dimensional prediction tool, based on the new prediction models, in which for each laboring woman the two predictions will be expressed as coordinates
on a two-‐dimensional plane e.g. 3% risk of fetal asphyxia and 20% risk of non-‐ progressive labor. As a complement to aid clinicians in the use of the tool, we will also suggest a list of possible interventions that may be employed, depending on the combination of predictions derived for each individual woman to whom the tool is applied during labor.
1.2 Outline of this thesis: description of the studies and research questions
The first part of the study presented in chapter two of this thesis provides some background information of prognosis and prognostic models in health care. The research questions posed in this section of the study are:
• How are prognostic models developed and evaluated?
The second part of the study presented in chapter three focuses on the adverse outcomes of labor that are of particular interest in this study: fetal asphyxia and failure to progress. The research questions posed in this section of the study are:
• What is fetal asphyxia? How is it defined, diagnosed and what are the risk factors associated with fetal asphyxia?
• What is failure to progress? How is it defined, diagnosed and what are the risk factors associated with non-‐progressive labor?
The third part of the study presented in chapter four of this thesis is a summary of the existing prognostic models for fetal asphyxia and non-‐progressive labor. Knowledge of the development of these prognostic models will guide us in the development of new prediction models for fetal asphyxia and failure to progress and subsequently aid in the development of the clinical prediction tool, which will be discussed in the fourth part of this thesis. The research questions posed in this section of the study are:
• What are the existing prognostic models for fetal asphyxia and non-‐progressive labor?
• How were these models developed and evaluated? How well do these models perform and what are their limitations?
The fourth part of the study presented in chapter five of this thesis is the development of a new prognostic models to predict the risk of fetal asphyxia and failure to progress using a cohort of high and low risk singleton, term pregnancies. By combining these models we developed a two-‐dimensional prediction tool, in which for each laboring woman the two predictions were expressed as coordinates on a two-‐ dimensional plane. The research questions posed in this study are:
• Which combination of predictive factors in the prognostic model for fetal asphyxia performs best?
• Which combination of predictive factors in the prognostic model for failure to progress performs best?
• How can the two-‐dimensional clinical decision tool be used to guide clinicians in the choice of interventions during labor?
The final part of the study, which will be presented in chapter six, provides the discussion and conclusions we arrived at during the study as well as recommendations to develop the clinical tool further.
1. A Abu-‐Hanna and PJF Lucas. Prognostic models in medicine: AI and statistical approaches. Method Inform Med, 40:1_5, 2001.
2. PJF Lucas and A Abu-‐Hanna. Prognostic methods in medicine. Artif Intell Med, 15:105_119, 1999.
3. Michelle E.M.H. Westerhuis et al. Prediction of Neonatal Metabolic Acidosis in Women with a Singleton Term Pregnancy in Cephalic Presentation. American Journal of Perinatology 2012; 29:167-‐174.
4. E Schuit et al. A clinical prediction model to assess the risk of operative delivery. BJOG 2012; 119:915-‐923
5. S. Katherine Laughton et al. Using a Simplified Bishop Score to Predict Vaginal Delivery. Obstet Gynecol 2011; 117:805-‐11
Chapter 2
Prognosis and Prognostic models in health care
2.1 Introduction
In this chapter background information on the domain will be given, beginning with an explanation of the term 'prognosis' and the meaning of the term in healthcare [1,2]. This will be followed by an explanation of the development and the evaluation and validation of prognostic models.
2.2 What is prognosis?
Prognosis simply means foreseeing, predicting or estimating the probability or risk of future conditions. In healthcare, prognosis commonly relates to the probability or risk of an individual developing a particular state of health (or an outcome) over a specific period of time, based on his or her clinical and non-‐clinical profile. Outcomes may be specific events, such as death or complications, or quantities, such as disease progression, changes in pain or quality of life.
In practice, clinicians do not predict the course of an illness but the course of an illness in a particular individual. Prognosis may be shaped by a patient’s age, sex, history, symptoms, signs and other test results. Besides, prognostication in healthcare is not limited to those who are ill. Healthcare professionals regularly predict the future of healthy individuals e.g. using the Apgar score to determine the prognosis of newborns and prenatal testing to determine the risk that a pregnant woman will deliver a baby with Down’s syndrome.
Given the variability among patients and in the etiology, presentation, and treatment of diseases and other health states, a single predictor or variable rarely gives an adequate estimate of prognosis. Clinicians-‐implicitly and explicitly-‐ use multiple predictors to estimate a patient’s prognosis. Prognostic studies therefore need to use a multivariable approach in design and analysis to determine the important predictors of the studied outcome probabilities for different combinations of predictors, or to provide tools to estimate such probabilities. These tools are commonly called prognostic models, prediction models, prediction rules, or risk scores. They enable care providers to use combinations of predictor values to estimate an absolute risk or probability that an outcome will occur in an individual. A multivariable approach also enables researchers to investigate whether specific prognostic factors or markers that are, say, more invasive or costly to measure, have worthwhile added predictive value beyond cheap or simply obtained predictors.
Prognostic models are used in various settings and for various reasons. The main reasons are to inform individuals about the future course of their illness (or their risk of developing illness) and to guide doctors and patients in joint decisions on further treatment. An example from obstetric care would include the use of a simplified Bishop score to predict the chances of vaginal delivery [1,2].
2.3 Development of prognostic models
In developing a prognostic model there are certain non-‐statistical and statistical aspects that should be taken into account [3]. The non-‐statistical characteristics of the multivariable study aimed at developing a prognostic model include:
Objective
The main objective of a prognostic study is to determine the probability of the specified outcome with different combinations of predictors in a well-‐defined population.
Study sample
The study sample includes people at risk of developing the outcome of interest, defined by the presence of a particular condition.
Study design
The best design to answer prognostic questions is a cohort study. A prospective study is preferable as it enables optimal measurement of predictors and outcome.
Predictors
Candidate predictors can be obtained from patient demographics, clinical history, physical examination, disease characteristics, test results and previous treatments. Studied predictors should be clearly defined, standardized and reproducible to enhance generalizability and application of the study results to practice. Predictors should be measured using methods applicable to daily practice. Finally, prognostic studies should only include predictors that will be available at the time when the model is intended to be used i.e. if the aim is to predict a patient’s prognosis at the time of diagnosis, predictors that will not be known until actual treatment has started are of little value.
Outcome
Prognostic studies should focus on outcomes that are relevant to patients such as death, complications, treatment response or quality of life. The period over which the outcome is studied and the methods or measurement should be clearly defined. Finally, outcomes should be measured without knowledge of the predictors under study to prevent bias [2].
Some of the important statistical aspects to consider when developing a prognostic model include:
Selecting candidate predictors
Studies often measure more predictors than can sensibly be used in a model and selection is required. Predictors already reported as prognostic would normally be included. Predictors that are highly correlated with others contribute little independent information and may be excluded beforehand. However, predictors that are not significant in a univariable analysis should not be excluded as candidates.
Evaluating data quality
In principle, data used for developing a prognostic model should be fit for the purpose. Measurements of candidate predictors and outcomes should be comparable across clinicians or study centers. Modern statistical techniques such as multiple imputations can handle data sets with missing values.
Selecting variables
No agreement exists on the best method for selecting variables. In the full model approach all the candidate variables are included in the model. This model is claimed to avoid over fitting and selection bias and provide correct standard errors and p values. The backward elimination approach starts with all the candidate variables. A nominal significance level, often 5% is chosen in advance. A sequence of hypothesis tests is applied to determine whether a given variable should be removed from the model. Backward elimination is preferable to forward selection whereby the model is built up from the best candidate predictors [3].
2.4 Performance of prognostic models: calibration and discrimination
Prognostic models yield scores to enable the prediction of the risk of future events in individual patients or groups and the stratification of patients by these risks. A good model may allow the reasonably reliable classification of patients into risk groups with different prognoses. However, to show that a prognostic model is valuable, it is not sufficient to show that it successfully predicts outcome in the initial development data. We need evidence that the model performs well for other groups of patients.
The performance of a logistic regression model may be assessed in terms of calibration and discrimination.
Calibration can be investigated by plotting the observed proportions of events against the predicted risks for groups defined by ranges of individual predicted risks. Ideally, if the observed proportion of events and predicted probabilities agree over the whole range of probabilities, the plot shows a 45° line. This plot can be accompanied by the Hosmer-‐Lemeshow test. The overall observed and predicted event probabilities are by definition equal for the sample used to develop the model.
Various statistics can summarize discrimination between individuals with and without the outcome event. The area under the receiver operating curve or the equivalent c (concordance) index is the chance that given two patients, one who will develop an event and the other will not, the model will assign a higher probability of an event to the former.
2.5 Internal and external validation of prognostic models
A model’s validity can be assessed using new recent data from the same source as the derivation sample (“apparent validation”), but a true external validation of the prediction model’s generalizability requires evaluation on data from elsewhere [4].
Internal validation
One validation strategy is internal validation. One approach to internal validation is to randomly split the dataset into two parts and the model is developed using the first portion (often called the training set) and its predictive accuracy is assessed on the second portion.
Another validation approach is non-‐random splitting which may be preferable as it reduces the similarity of the two sets of patients. If the available data are limited, the model can be developed on the whole dataset and techniques of data re-‐use, such as cross-‐validation and bootstrapping can be used to assess performance. Bootstrapping is a method of estimating properties of an estimator, such as its variance, by measuring those properties when sampling from an approximate distribution e.g. a resample with replacement, of the observed dataset [4].
Quality assessment
There are a few frameworks that have been developed for the purpose of assessing the quality of studies describing the development or validation of prediction models [5,6]. One such framework is the Transparent Reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD) statement [7].
Bibliography
1. A Abu-‐Hanna and PJF Lucas. Prognostic models in medicine: AI and statistical approaches. Method Inform Med, 40:1_5, 2001
2. Karel GM Moons et al. Prognosis and prognostic research: what, why and how? BMJ 2009;338:b375
3. Karel GM Moons et al. Prognosis and prognostic research: Developing a prognostic model. BMJ 2009;338:b604
4. Douglas G Altman et al. Prognosis and prognostic research: validating a prognostic model. BMJ 2009; 338:b605
5. Medlock S et al. Prediction of mortality in very premature infants: A systematic review of prediction models. PLoS ONE 2011; 6(9) e: 23441
6. Leushuis E et al. Prediction models in reproductive medicine: a critical appraisal. Hum. Reprod. Update (2009) 15 (5): 537-‐552.
7. Collins SG at el. Transparent reporting of a multivariable prediction model for individual prognosis or diagnosis (TRIPOD): the TRIPOD statement. BMJ 2015; 350:g7594 Chapter 3
Labor and its potential adverse outcomes; fetal asphyxia and failure to progress
3.1 Introduction
Labor is the process by which the fetus and the placenta leave the uterus. Delivery can occur in two ways: vaginally or by a cesarean section. According to the American Pregnancy Association, labor usually occurs in three stages [1]. The first stage is defined as the time from the onset of labor until the cervix is completely dilated to 10cm. The second stage of labor is defined as the period after the cervix is dilated to 10cm until the baby is delivered. The third stage of labor involves the delivery of the placenta. The first stage of labor is the longest and involves three phases: early, active and transition labor phase. The early labor phase is the time from the onset of labor until the cervix is dilated to about 3cm. The active labor phase continues from 3cm until the cervix is dilated to 7cm. The transition phase continues from 7cm until the cervix is fully dilated to 10cm.
For laboring women the preferred pathway and outcome would be a spontaneous labor resulting in a vaginal delivery. However, in some cases labor can take different pathways and involve different interventions, some of which may result in adverse outcomes for the mother and child. For instance, induction of labor is a common and essential element of obstetric practice with an incidence of approximately 20% of pregnancies [2,3]. If induction is successful the laboring woman will achieve the active phase of labor. Some women may also experience a failure of labor to progress; in first stage of labor during the active phase this implies no cervical dilation for at least 2 hours and in the second stage of labor this implies no descent of the fetus’s head for at least one hour despite adequate uterine contractions. If fetal distress is suspected or failure to progress is observed during labor, the clinician might choose to intervene through an assisted vaginal delivery or by performing a cesarean section.
Each of these interventions: induction of labor, instrumental delivery and cesarean section should be used with caution as they may increase the risk of adverse outcome for either the mother or the baby. For example, induction of labor has been associated with increased rates of epidural anesthesia, emergency cesarean delivery and adverse neonatal events such as requirement for resuscitation [4,5]. Similarly, instrumental delivery is a strong risk factor third-‐ and fourth-‐degree perineal injuries [6]. Cesarean delivery itself is associated with an increased risk of respiratory morbidity in babies, even after 37 weeks’ gestation [7].
For the purpose of this SRP, the outcomes of interest are fetal asphyxia and failure of labor to progress [8]. The research questions posed in this chapter are:
• What is fetal asphyxia? How is it defined, diagnosed and what are the predictive risk factors for fetal asphyxia?
• What is failure to progress? How is it defined, diagnosed and what are the interventions and predictive risk factors for non-‐progressive labor?
We searched PubMed from January 1, 1990 until October 1, 2014 for publications studying fetal asphyxia (also referred to in some instances as intrapartum asphyxia, birth asphyxia or metabolic acidosis) and failure to progress (also referred to in some instances as non-‐progressive labor or dystocia) and offering a definition, means of diagnosis and classification. We also searched for publications studying the risk factors (maternal and obstetric risk factors) that are associated with fetal asphyxia and failure to progress.
The following keywords were used in the database search: fetal asphyxia, failure to progress, risk factors, cesarean section, instrumental vaginal delivery and labor/delivery. The search was narrowed down to articles written in English.
All the titles and abstracts of the articles were reviewed and articles were included if they described a descriptive study of fetal asphyxia or failure to progress and/or their associated risk factors. The full text articles of these studies were retrieved and all of them were reviewed.
First we looked at the definition of asphyxia in general and fetal asphyxia in particular and also the definition of failure to progress. Next we looked at the methods used to diagnose fetal asphyxia and failure to progress. Finally we looked at the maternal and obstetric risk factors (both antepartum and intrapartum characteristics) that are associated with (and possibly predictive of) fetal asphyxia and failure to progress.
3.3 Results
After full text reviews, 19 articles were found eligible for inclusion in the study. Of these, 8 articles focused primarily on the definition and diagnosis of fetal asphyxia and 4 articles focused primarily on the risk factors associated with fetal asphyxia. Three articles focused primarily on the definition and diagnosis of failure to progress and 4 articles focused primarily on the risk factors and obstetric interventions associated with failure to progress
3.3.1 Fetal asphyxia
Definition of fetal asphyxia
Specific evidence of asphyxia can be provided by means of a blood gas and acid-‐base assessment. When this has been done, the timing of the asphyxia event can be determined more accurately. For this reason, the use of the term perinatal or birth asphyxia is discouraged. The effects of fetal asphyxia resulting from compromised maternal-‐fetal blood gas exchange before delivery should be differentiated from newborn asphyxia resulting from cardiorespiratory complications after delivery. Accurate diagnosis and precise timing are important if strategies to prevent or modify the effects of asphyxia are to be developed.
A task force set up by the World Federation of Neurology Group for the prevention of cerebral palsy and related neurologic disorders defined asphyxia as a condition of impaired gas exchange leading, if it persists, to progressive hypoxemia, hypercapnia and metabolic acidosis[9]. By this definition asphyxia may affect the fetus and
newborn. In some cases, the asphyxia is short-‐lived with no apparent pathological effects. Significant asphyxia leading to tissue oxygen debt with accumulation of fixed acids results in metabolic acidosis. Thus for clinical purposes the task force proposed the following definition for fetal asphyxia: a condition of impaired blood gas exchange leading to progressive hypoxemia and hypercapnia with a significant metabolic acidosis.
A diagnosis of an asphyxiating event should not be made unless there is some evidence of an interruption of oxygen supply or blood flow to the fetus. These events can be secondary to problems from the mother (e.g. hypotension, toxemia, uterine rupture), the placenta or umbilical cord (e.g. abruption, infection or inflammation, or umbilical cord compression), or the fetus or infant (e.g. central nervous system depression, anomalies, infection) [10]. The term “asphyxia” should not be used unless the neonate meets all of the following conditions: umbilical cord arterial pH less than 7, Apgar score of 0 to 3 for longer than 5 minutes, neurological manifestations (e.g. seizures, coma or hypotonia), and multisystemic organ dysfunction [11]
The Apgar score was developed by Dr. Virginia Apgar in 1952 as an objective tool that measures five signs of physiologic adaptation: heart rate, respiratory effort, muscle tone, reflex irritability and color. A score is a sum of the values assigned to the infant at 1 and 5 minutes of life, with a score of 7 or more indicating that the baby is in good to excellent condition [12]. A retrospective analysis of 151,891 neonates born over a 10-‐year period revealed a mortality rate of 24.4% for infants with five-‐minute Apgar scores of 0 to 3 versus a mortality rate of 0.02% for infants with five-‐minute Apgar scores of 7 to 10 suggesting that an Apgar score of 3 or less at 5 minutes of life does predict a higher rate of mortality [13].
Diagnosis of fetal asphyxia
Laboratory and clinical studies suggest that the threshold for a significant metabolic acidosis is a base deficit between 12 and 16 mmol/L. Asphyxia with significant metabolic acidosis is associated with seizures in the fetal lamb. Clinical studies have also showed an association between severe academia and multiorgan complications in the newborn.
Routine blood gas and acid-‐base assessment of umbilical artery blood at delivery has demonstrated an umbilical artery base deficit > 12mmol/L in 2% and > 16 mmol/L in 0.5% of the total population. This assessment means that 98% of newborns do not have a significant asphyxial episode during labor and delivery. However, for the 2% of newborns that have been exposed to asphyxia this may affect their outcome.
Evidence of a significant metabolic acidosis can establish that exposure to asphyxia has occurred. This will also suggest the degree of metabolic acidosis at the time of sampling. However, it does not necessarily indicate the severity of the asphyxial exposure to the fetus. The duration of the asphyxia is generally not known. Also the nature of the exposure (i.e. continuous or intermittent) or whether the asphyxia during labor and delivery is the last in a series of insults is not known.
The importance of an asphyxial exposure is influenced by the fetal response. The response to asphyxia is a concentration of the fetal circulation with increased blood flow to the brain, heart and adrenals. If the hypoxia is sustained fetal cardiovascular decompensation will occur. Laboratory studies in fetal lamb have shown that fetal
cardiovascular decompensation results in decreased cerebral blood flow and cerebral oxygen metabolism and brain damage. These studies also suggest that the fetal response is not necessarily proportional to the exposure [14].
For the 2% of newborns who have been exposed to an asphyxial event, a classification of the severity of the exposure is required to predict the long-‐term outcome in the child. The severity of intrapartum fetal asphyxia can be classified by determining the short-‐term outcome as expressed by newborn encephalopathy and other newborn organs system complications. This alternative is acceptable because the duration of the asphyxia itself cannot be determined and the clinical measures of fetal cardiovascular compensation and decompensation are not available. Also, the susceptibility to the exposure may depend on the different characteristics of the fetus i.e. gestational age, small for gestational age vs. normal for gestational age.
The clinical signs of newborn encephalopathy associated with intrapartum fetal asphyxia occur more often on the first day after delivery, with decreasing frequency on the second and third day after delivery. Newborn encephalopathy was classified as minor if it consisted of jitteriness and irritability; moderate if it consisted of lethargy or abnormal tone and severe if it consisted of coma or abnormal tone and multiple seizures. Cardiovascular complications were classified as minor it there was bradycardia or tachycardia (defined by the 95% confidence limits for heart rate for term and preterm newborns), moderate if there was hypertension or hypotension (defined by the 95% confidence limits for blood pressure for term and preterm newborns), and severe if there was abnormal electrocardiographic or echocardiographic findings. Respiratory complications were classified as minor if needing supplementary oxygen, moderate if requiring continuous positive airway pressure or transient ventilation (<24 hours), or severe if requiring mechanical ventilation for >24 hours. Renal complications were classified as minor if hematuria was observed, moderate if there was an elevation of serum creatinine (>100 µmol/L) and severe with clinical evidence of oliguria (<1 ml/kg/hr) or anuria. The classification of mild, moderate and severe intrapartum fetal asphyxia is based on the evidence that at this time early-‐onset newborn encephalopathy is the best single predictor of long-‐ term outcome. The long-‐term outcome examined in most studies of fetal asphyxia has been severe handicap. Luckily, most survivors of intrapartum fetal asphyxia do not have major sequelae. However, it is unclear whether there is a range of casualty after fetal asphyxia [15].
A study by Ruth and Raivio compared Apgar scores, cord blood pH and cord lactate levels in more than 900 infants and looked at outcome. They found that 11% of the infants with cord blood acidosis had an Apgar score below 7, whereas 41% of infants with an Apgar of less than 7 had an acidosis. In the end, the sensitivity and positive predictive value of a low pH for adverse outcomes were 21% and 8% respectively. Cord blood lactate levels were at 12% and 5% respectively. The sensitivity and positive predictive value of the Apgar score values were 12% and 19% respectively. Thus, in situations where fetal blood sample values are not readily available Apgar scores may be used to obtain a comparable diagnosis of incidences of fetal asphyxia i.e. an asphyxial event is assumed to have occurred if the Apgar score is less than 7 at 5 minutes of life [16].