Another Day, Another Lie

And yes, another abortion post, but as the anti-abortion lobby keeps on lying, I’m just going to have keep right on pointing out the truth.

So what have we got for today? Well it’s this article in the Telegraph which claims that:

Hundreds of children are surviving after being born within the legal limit for abortion, official figures have disclosed.

All of which is complete and utter nonsense – it’s actually 52 children (in 2005) and not the 250 that the article claims:

Department of Health data show that 909 children were born between 22 and 24 weeks of pregnancy during 2005. Of those, 250 survived for at least a year.

The figures will be used by pro-life campaigners as further proof of the need to cut the abortion time limit when MPs vote on the issue in the coming weeks.

All it proves is that the churnalist responsible for writing this article need to go on a basic course in how to read statistics.

These are the official figure, as obtained by David Amess MP via a written question in the House of Commons:

<

Live births and infant deaths by gestational age at birth: babies born in England and Wales, 2005
Births (number) Deaths (number) Rates per 1,000 live births (percentage)
Gestational age (weeks) Live births Under 7 days Between 7 and 28 days 28 days and over but under 1 year Deaths under 1 year Under 7 days Between 7 and 28 days 28 days and over but under one year Deaths under 1 year
All 645,887 1,695 544 961 3,200 2.6 0.8 1.5 5.0
22 152 140 3 1 144 921.1 19.7 6.6 947.4
23 283 198 34 7 239 699.6 120.1 24.7 844.5
24 474 168 65 43 276 354.4 137.1 90.7 582.3
25 499 95 43 38 176 190.4 86.2 76.2 352.7
Stillbirths by gestational age at birth: England and Wales, 2005
Gestational age (weeks) Number
24 291
25 230
Notes:
1. The tables show figures for gestational ages 22 to 25 weeks for live births, deaths under seven days of age), deaths between seven and 28 days old, deaths at 28 days and over but under one year and deaths under one year old for babies born in 2005 in England and Wales (the latest year for which figures are available). Data are not available by gestational age at birth on survival beyond one year.
2. The tables includes figures for stillbirths at gestational ages 24 and 25 weeks as a stillbirth is defined as a baby born dead after 24 completed weeks gestation.
Source:
Office for National Statistics

All of which is written up as follows:

The data, for births in England and Wales, showed that eight of the 152 children born after 22 weeks’ gestation lived for a year or more.

At 23 weeks, 44 of 283 children survived. At 24 weeks, almost half – 198 of 474 – of babies survived. Of the 201,173 abortions in England and Wales in 2006, 1,262 were at 22 weeks or more.

And the problem with this write-up is a very simple one – its makes the old ‘week zero’ error. The legal limit for abortion in the UK, other than in cases of serious disability, is 24 weeks and 0 days, which means that a neonate show in the statistics as having been born at 24 weeks gestation was born AFTER the upper limit for abortion not before.

So, to correct the error, the actual number of live births taking place inside the legal limit for abortion was actually only 435 and the number that survived for more than a year a mere 52. The additional 474 live births and 198 neonates that survived for a year or more which get included in the false claims made by the article should not have been included and, of course, the article has nothing to say on the prevalence of disabilities and other developmental problems in the survivors either.

All of which means that the actual survival rate for neonates born prior to the legal limit is a matter of 5.2% at 22 weeks (8 out of 152) and at 23 weeks its 15.5% (44 out of 283), giving an average of 11.95%.

(As you might expect, I gave the Telegraph a go by posting a polite comment pointing out this basic error – a comment that has, as yet, failed to appear)

Nadine Dorries, of course, gets into the act by spouting her usual, utterly ill-informed nonsense:

Nadine Dorries, a Tory backbencher who is campaigning for lower abortion time limits, said the national figures masked a “postcode lottery” where properly equipped neonatal units could deliver much higher survival rates for premature babies.

She added: “Even more viable babies are being aborted than these figures would suggest. If you go into labour at 22 weeks, there is something wrong with you or your baby. Most babies aborted at 22 weeks are healthy, so their survival chances would be even higher.”

None of which is true as but for the abortion the foetuses which Dorries contends would have had an even better chance of survival would be highly unlikely – there were 645,887 live births recorded in the UK in 2005 of which 435 occurred prior to the 24 week legal limit for abortion, while there were a total of 885 abortions at the same gestational age giving a 0.6% of any one of those aborted foetuses being born, naturally, before the 24 week limit and a 0.07% chance of a being a lucky survivor.

The Telegraph also reintroduces another old ‘friend’ into the debate, Professor John Wyatt of University College London Hospital and, not uncoincidentally, also a member of the Christian Medical Fellowship:

John Wyatt, a University College London professor and a member of the anti-abortion Christian Medical Fellowship, has claimed that survival rates are as high as 42 per cent at 23 weeks and 72 per cent at 24 weeks’ gestation.

Anti-abortion campaigners point out that survival rates can be even higher in specialist hospital units.

Last month, a separate study, carried out at University College London Hospital, found that one third of babies born between 22 and 25 weeks’ gestation survived in the early 1980s but this had risen to 71 per cent by the late 1990s.

The flaws in Wyatt’s arguments were neatly dissected here, by Ben Goldacre, and a quick review of the UCLH website reveals that its ‘separate’ study isn’t quite as separate as the Telegraph suggests:

The study was conducted by Professor John Wyatt, a UCLH consultant and professor at the UCL EGA Institute for Women’s Health, in conjunction with the hospital’s neonatal team, and is part of a larger, ongoing follow-up programme for all extremely premature babies cared for at UCLH.

And although the study got an even bigger write-up in the Telegraph a few weeks ago:

A study at one of Britain’s top neonatal units found that one third of babies born between 22 and 25 weeks’ gestation survived in the early 1980s but this had risen to 71 per cent by the late 1990s.

On neither occasion did the Telegraph see to mention Wyatt’s involvement in the study or that the study uses Wyatt’s data, even though its limitations are acknowledged by Wyatt, himself, in UCLHs press release:

“We fully acknowledge that single centre studies such as this one have statistical limitations because of the relatively small numbers involved and they can also suffer from the problems of possible selection bias. However, they are hugely important because they provide information on the survival rates that can be achieved with consistent levels of staffing and resources, and with consistent policies. They also allow counselling of women who deliver at the hospital to be based upon data that are relevant to their individual situation.”

To recap a little on Ben’s remarks about Wyatt’s data, the UCLH press release includes the following:

During the study, of the 357 infants admitted to the neonatal unit at 22-25 weeks gestation, 163 (45 per cent) survived to discharge from the unit. 161 were alive for medical assessment at one year of age, which included a neurological examination, a developmental assessment and tests of visual and auditory function. Approximately one quarter of the infants showed signs of impairment leading to disability at one year of age – a figure which stayed fairly constant over the assessment period.

The data covers a period of 20 years from 1981 – 2000, so we’re only dealing with data from around 17-18 live births a year and only then from those neonates who survive for long enough to be admitted to the hospital’s neonatal ICU, which excludes those neonates who die in the delivery room before they can be admitted to the ICU – and the exclusion will, of course, act to inflate the survival rates quoted in the study.

Reading back over the section of the Science and Technology Committee Report that deals with Wyatt’s evidence, one finds some rather worrying and not just a little perplexing anomalies in his evidence. The S&TC notes that:

39. In his written evidence Professor Wyatt stated that “Data from a prospectively-defined long-term follow-up study at the Neonatal Intensive Care Unit at University College London Hospital has shown survival rates in the period 1996 to 2000 of 42% at 23 weeks and 72% at 24 weeks.” The reference given is to a 2004 abstract (Riley et al, 2004) which does not contain the data mentioned. Professor Wyatt also told us in oral evidence that the denominator in the Riley study was all live births, but the denominator of that study was in fact admissions to NICU which will include transfers and exclude deaths in the delivery room (see below). In a further two memoranda to the Committee, Professor Wyatt clarified that the 42% survival figure for 23 weeks had not been published in a peer-reviewed journal and confirmed that it was not even in the abstract given as a reference. He further explained that he had “for the benefit of the Committee” therefore gone back to reanalyse the data prospectively collected in 1996-2000 and excluded transfers and added back in deaths in the delivery room.

Hardly the most impressive start, but when pressed by the Committee, Wyatt did manage to go back over his evidence and produce some fresh numbers:

40. These data therefore represent, according to Professor Wyatt, all babies that were born at UCLH, including those that showed signs of life but died before admission to the NICU, between 1996 and 2000:

Table 4: Survival rates at UCLH by gestation week

Gestational age (completed wks) Total born alive at UCLH Number survived to 1 year of age Percentage Survivors
22 8 4 50%
23 13 6 46%
24 22 18 82%
25 26 20 77%

Apart from the obvious question of how the survival rate is higher at 22 weeks than it is at 23 week, which the committee commented on:

41. These impressive survival figures are higher than the national average but they illustrate a difficulty with data on extremely premature neonates at individual hospitals which is that there are very few births at these gestations and consequently the confidence we can place on the percentages is quite low. This is demonstrated very clearly in these data, where the chance of survival appears higher at 22 weeks than 23 weeks and at 24 weeks than 25 weeks, which is obviously not the case. As Professor Wyatt put it: “If you have a very small number, you have a large statistical error”.

There is also the question, which the committee declined to comment on explicitly, of just exactly it is that these figures are actually higher than the numbers he originally gave given these exclude transfers from other hospitals, who would be only those neonates with the best chance of survival as they, of course, have to survive the transfer before being admitted to UCLH, while adding back in deaths in the delivery room, the inclusion of which should have lowered the survival rates cited in his second submission.

Wyatt’s other piece of statistical jiggery-pokery is, of course, the use of statistics for neonates born between 22 and 25 weeks gestation, two weeks either side of the current limit and when you start to work the numbers something a little odd emerges.

If you take the national data for live births and look at the breakdown for gestational age, then the numbers indicate that the before/after split for 22-25 weeks is about 1:3, to be more precise around 8% of births in that group take place at 22 weeks, 15% at 23 weeks and the remainder (77%) are more or less evenly split between 24 and 25 weeks – the actual figures are 40% for 24 weeks and 37% for 23 weeks and the slightly higher number at 24 weeks could be a reflection of clinical practice as its know that survival chances improve considerable once you get over the 24 week barrier so where a woman begins to go in to labour at 22-23 weeks there is often an effort made to hold back the birth until 24 weeks has been reached, which may skew the numbers slightly. There is also a degree of imprecision in assessing gestational age to take into account, a ‘safety margin’ in the case of abortions of seven days.

Wyatt’s data, given above, splits with 12% of live births at 22 weeks, 19% at 23 weeks – both higher than the national average, 32% at 24 weeks and 40% at 24 weeks. Small though the numbers are, Wyatt’s are out in terms of the percentages you’d expect at each age and, crucially, vary to the greatest extent at the point of the upper limit for abortion, 24 weeks.

This is of critical importance, given that the margin of error in assessing and recording gestational age could be out by as much as seven days and if Wyatt’s staff are perhaps, being a little conservative with their assessments of gestational age at around the 24 week boundary this could easily skew the numbers, inflating both the number of live births and the number of survivors recorded as having been born before the 24 week limit.

Suppose, for example, that there is such an error in Wyatt’s data and suppose, in addition, that the error functions in such a way as to result in all the ‘time-shifted’ births being amongst the survivors. If that were to happen then that would leave Wyatt with either 5 births and one survivor at 22 weeks or 6 births and 2 survivors (the actual figure is 5.65 but we’ll assume that Solomon’s not on shift when neonate no.6 arrives) cutting his survival rate down to between 20% and 33% – still impressive when the national rate is only 5% but not the 1 in 2 that the data purports to show.

The same shift at 23 weeks would give around 10-11 live births and 3-4 survivors, about a 30-35% rate and so on down the line, although the shift would have less impact on the numbers at 24 and 25 weeks because we’re dealing with a larger group of neonates.

Applying those same proportions to the 20 year data set of the now published study gives the number of live neonates admitted to UCLH’s ICU at 22 weeks as around 28-29, just over one a year, and the number of survivors over the 20 years at 13 or so on Wyatt’s numbers and around 9 or 10 on the time-shifted figures, while for 23 weeks the numbers run to 54 admissions to ICU and either 24 or 18 survivors. One way or another, even at what is thought to be one of the best neonatal ICU’s in the UK, very , very few neonates born below the 24 week limit beat the odds and yet anti-abortion campaigners want people to believe that this is grounds for denying access to abortion services to more than 2000 women a year, the number that would be caught up in a reduction of the limit to 20 week, women who are generally amongst the most vulnerable of all who choose to have an abortion.

Dorries, as one might expect, thinks that the belated publication of Wyatt’s data in a peer-reviewed journal indicates that the S&TC report is now out of date, despite the fact that its apparent from the UCLH press released that none of the statistical issues raised by Ben Goldacre have been address, indeed, in the absence of parliamentary scrutiny Wyatt has reverted back to using data on neonates admitted to the ICU, which is fine as measure of the effectiveness of the unit but does nothing at all to clarify the issue of viability.

  • Dudley

    Fantastic and fascinating analysis – thank you.