November 30, 2018

expert reaction to screening study on uk prevalence of fetal alcohol spectrum disorders (FASD)

Research published Preventative Medicine suggests that fetal alcohol spectrum disorders (FASD) could be a significant public health concern in the UK.

A Before the Headlines accompanied this Round Up.

Dr Virginia Beckett, Consultant Obstetrician and spokesperson for the Royal College of Obstetricians and Gynaecologists (RCOG), said:

“The results of this large study are a cause for concern, but it is important to note that this is an estimate of prevalence, rather than a formal diagnosis of foetal alcohol spectrum disorder. More studies are required to conclusively determine how many children might be affected and that alcohol exposure was the cause of the impairments.

“It is well known that alcohol consumption during pregnancy can lead to long-term harm to the baby – the higher the amount of alcohol consumed, the greater the risk. Foetal alcohol spectrum disorder is directly linked to how much and how often a woman drinks during pregnancy.

“As there is no proven safe amount of alcohol women can consume during pregnancy, avoiding alcohol is recommended, particularly for women trying to conceive or those in the first three months of pregnancy.

“Women should be reassured that the risk of harm to their baby is likely to be low if they have consumed small amounts of alcohol before knowing she is pregnant. It will be unlikely in most cases that her baby has been affected.

“If women are pregnant and struggling with an alcohol problem, they are advised to speak to their midwife, doctor or pharmacist as specialist help and support is available. It’s never too late to stop drinking, and stopping at any point during pregnancy can help reduce the risk of problems in the baby.”

Dr James Doidge, Senior Research Associate, University College London (UCL), said:

“Screening tests are designed ‘screen out’ people who are unlikely to have a disease, not to identify people who do have it. By their nature, screening tests identify large numbers of ‘false positives’, sometimes many times more false positives than true positives. A positive screening test cannot be equated with a positive diagnosis and the prevalence of a positive screening test can bear little relation to the prevalence of the diagnosis. A big problem in the FASD setting is that FASD are not clearly defined and there is no gold-standard diagnostic test. Without clear definitions and diagnostic tests, there is little point talking about prevalence. On the other hand, alcohol consumption during pregnancy is certainly a risk factor for many poor outcomes, whether those outcomes are classified as FASD or not. This study is more about understanding the extent of exposure to alcohol during pregnancy and the possible extent of the problems that this causes.

“Regarding the missing data, it is very common for small numbers of participants in longitudinal surveys such as ALSPAC to have data that are ‘complete’. These subsets of participants with complete data are usually not at all representative of the wider cohort (they may be healthier, for example) and should not be the focus of any analysis. Researchers need to consider all of the data that were collected, whether complete or incomplete, and make plausible assumptions about the data that are missing.”

Prof Kevin McConway, Emeritus Professor of Applied Statistics, The Open University, said:

“This research provides some potentially very useful information relating to FASD, but I really don’t think it tells us anything very helpful about how common these conditions are in the UK. What’s certainly useful is the information on the numbers of children who were exposed to alcohol before birth, and on how many of them had evidence of impairments of learning and behaviour that can arise in FASD, but as the researchers rightly point out, it isn’t possible to prove (or disprove) that the alcohol exposure was the cause of the impairments.

“To summarise my views: yes, this study does provide some useful information, but I don’t think we can rely on it to estimate how common FASD is in the UK. There are good reasons why the researchers’ estimates of up to 17% of children having FASD symptoms is very likely to be an overestimate of the percentage of children who actually have FASD in the UK, and quite possibly it overestimates a lot. Does that mean that FASD is not really a problem in the UK? That’s absolutely not the case – the relative high rates of drinking alcohol during pregnancy mean that FASD could well be more of a problem here than elsewhere. I agree with the researchers that research using more reliable methods of finding out the extent of FASD should be done. Is the Chief Medical Officer’s advice to avoid alcohol in pregnancy changed by any of this? Absolutely not.

“Why do I think the study is less useful in providing information on how common FASD is in British children? The researchers used a screening tool that they had developed, which classified between 6% and 17% of the 13,500 children in their study as having symptoms of FASD, depending on which exact method they used. They do correctly point out that the screening tool “is not the same as a formal diagnosis”, but I think it’s worth looking further into the implications of that. In almost all screening situations, where a relatively simple method is used to look for people who might possibly have a disease or condition, the percentage of people with a positive screening test is higher than the percentage of people who eventually turn out to have the disease in question, and it’s often a lot higher.”

Dr Christopher Steer of the Royal College of Paediatrics and Child Health (RCPCH) said:

“I would consider this research to be of high quality from a recognised authoritative source.

“The screening methodology is well designed, ingenious, and conservative in its approach, and the large numbers of subjects in this birth cohort add considerable strength, relevance, and significance to the findings.

“The authors clearly state that the screening “diagnosis” of FASD has been reached “on a balance of probability”, pointing to the requirement for active FASD surveillance studies, so far not achievable in the UK.

“The UK has one of the highest pregnancy drinking rates in the world, at just over 40%, only exceeded by Ireland, Belarus and Denmark. Not surprising therefore, to identify such a large number of children affected by alcohol harm in utero. The figure of 17% is certainly strikingly high, well above previous indirect indications pointing to levels of between 2 and 5%.

“Under current circumstances many cases of FASD remain undiagnosed. Lack of early detection and necessary support increases the risk of long-term disability and secondary difficulties affecting learning, behaviour, health, and indeed in the more severe examples of FASD, life expectancy. FASD remains the most common, and potentially preventable cause of learning and behavioural difficulties in the world.

“The UK Chief Medical Officers’ advice and message “No Alcohol, No Risk” is therefore supported by this research, which also underlines the pivotal importance of awareness raising and preventative strategies to reduce the prevalence of FASD in the UK.

“This study also in my opinion underlines an urgent need for UK based active FASD surveillance studies, screening children in mid childhood at which point reliable diagnosis utilising evidence based approach is entirely feasible.”

Prof Jean Golding, Emeritus Professor of Paediatric and Perinatal Epidemiology, University of Bristol, said:

“The authors of this study have created a screening tool with the aim of identifying children with the fetal alcohol spectrum disorder. They have used very detailed and complex data collected from their pregnancies until the children were 15 years old and developed an algorithm to identify affected children.

“However, there is no hard evidence given that this algorithm works. The women had to have drunk some alcohol in pregnancy for their children to be considered; the algorithm then took account of the presence of reduced growth, behaviour, mental and/or motor development to come to a conclusion that the child was adversely affected by the mother having drunk alcohol in pregnancy. This might be appropriate if the authors had shown that using this algorithm among children of women who had not drunk alcohol in pregnancy revealed a very much smaller proportion of affected children. As it stands, this study cannot be considered to provide an appropriate indicator of the incidence of the fetal alcohol effects.”

Additional Info from Prof Kevin McConway:

“A more familiar example is breast screening, where women in a certain age range are called in to have a mammogram (a kind of breast X-ray) to screen them for breast cancer. According to the leaflet that the NHS gives to women who are called for breast screening, about 4 in every 100 women who have breast screening will be called back for more tests – in the jargon  they “screen positive”, but only about 1 of these 4 women will eventually turn out to have breast cancer.  That is, the rate of screening positive is about 4 times as big as the rate of actually having breast cancer. That’s despite the fact that the screening mammography is actually pretty accurate.

“How can that be? Well, there are two ways that a woman might screen positive. Either she really has breast cancer (so she’s a “true positive”), or she really doesn’t have a breast cancer but the screen has indicated that she has (so that, in the jargon, she’s a “false positive”). The chance of bring a false positive, if the women really doesn’t have cancer, is actually pretty low, but the great majority of women who are screened do not have a cancer, and any of them could turn out to be a false positive. So overall, the number of positives is quite high, and there are about three times as many false positives as true positives. (There’s another way the screening result can be wrong – it could be a “false negative” where the woman does not screen positive but turns out to have a cancer after all. However, just for comparing the numbers who screen positive with the real number of cancers, false negatives make very little difference because they can arise only in women who really do have cancer, and the great majority of women being screened do not have a cancer so that the number of false negatives is very small.)

“Breast screening is considerably more accurate than the new screening tool for FASD developed in this research, in that the chance of a positive screen is high when a woman really does have a cancer, and is low when she really doesn’t have cancer. So does the number of children screened positive for FASD exceed the true number who really have it to a similar extent as with breast screening? I can’t say, because that depends on information that the researchers didn’t have (and nor do I). It depends on the chances of the two possible types of screening error (false positives and false negatives), and though the researchers report those from the data they used to develop the screening tool, those results apply only to the small number (31) of patient profiles used for the development, and they are incompatible with the results that were reported from the 13,500 children in the main study. (Also, though much less important in my view, expert diagnoses used in developing the tool were not what the researchers would describe as a “formal FASD diagnosis”, since they were based only on certain details about the 31 patients and not on a “specialised in-person evaluation”.) But most crucially, the extent to what the screening results might overestimate the real prevalence of FASD depends on the real prevalence of FASD, and the whole point is that we don’t know that real prevalence. However, in pretty well all screening contexts, the proportion of positive screens is higher than the real prevalence of the disease, possibly to a considerable extent, and I think that’s likely to have happened here too.”

‘Screening prevalence of fetal alcohol spectrum disorders in a region of the United Kingdom: a population-based birth-cohort study.’ by Cheryl McQuire et al. was published in Preventative Medicine at 00:01 UK time on Friday 30th November.

Declared interests

Prof Jean Golding:  ‘I am closely involved with ALSPAC but have not been involved in this work.’