Do estimates of survival change decisions made?

Kidszun A, et al. Effect of Neonatal Outcome Estimates on Decision-Making Preferences of Mothers Facing Preterm Birth: A Randomized Clinical Trial. JAMA Pediatr. 2020.

This is a short report of an interesting idea, published as a research letter. The authors from Germany randomized pregnant women hospitalised for threatened preterm labour, but who had already reached at least 28 weeks gestation, to respond to a scenario of an extremely preterm birth, at 22 or 23 weeks gestation, with either a 60% or a 30% survival rate. They were then asked whether they thought active intensive care provision or comfort care were preferable.

The attitudes toward intensive care provision for the very preterm infant were not different between the 2 scenarios, 47% would have wanted active intervention for the baby with a 60% chance of survival, and 50% for the baby with a 30% chance of survival.

We are advised by professional societies to make shared decisions with parents, prior to extremely preterm delivery, after ensuring that parents are well informed. These recommendations are often accompanied by long lists of potential complications and percentage risks that we are supposed to ensure that the parents understand prior to delivery. This new small, limited, study suggests that all of our information giving doesn’t have much impact, at least between these 2 percentage survival figures; what matters to the decisions which are made is the underlying attitudes of the parents. The authors state, without explaining where their conclusion came from that “an attitude that mere survival is at least as important as quality of life was associated with a preference for life-sustaining treatments”.

I think that we should use the antenatal encounter with potential extremely preterm parents to investigate their values, as much as that is possible, rather than trying to transfer complex information.

Many studies have examined ways to ensure that parents are well informed, but how much that numerical information impacts the decisions that are made is not clear. Decision aids are promoted by various groups as ways of ensuring that information is transferred, and they may indeed increase the amount of information retained by parents in the short term, but do they change decisions? One study from last year examined whether Decisional Conflict was affected by the use of a decision aid. This was an RCT using a decision aid that the authors have previously published. (Guillen U, et al. Evaluating the Use of a Decision Aid for Parents Facing Extremely Premature Delivery: A Randomized Trial. J Pediatr. 2019;209:52-60 e1).

I have previously criticized this specific decision aid for the numbers that are used to describe long term outcomes; the range of blindness for example for babies between 22 weeks and 25 weeks 6 days is shown as being 1 to 15%, I have never seen long-term data from this century showing a 15% incidence of blindness, the majority of studies give an upper limit of serious visual impairment of about 2 to 3%. Similarly, the decision aid includes a section about the risk of “mental disability” which is apparently something that happens in 18 to 54% of former preterm babies of this gestational age. Even the use of this term shows up the limitations of such decision aids, the person presenting the decision aid will have to explain what that term is supposed to mean, and their own prejudices and beliefs and values will become part of the discussion, a decision aid such as this is not value-free! I don’t think you can actually create a value-free aid unless you stick to objective outcomes such as death, and even then, whether death is the worst outcome, or whether some survivors would have been “better off dead” (and which ones) is something that medical caregivers and our patents/parents often disagree about.

The primary outcome of the Guillen study was whether the mothers were definite about their decision or remained uncertain, measured on a decisional conflict scale. The trial showed no difference in that primary outcome. About 200 mothers were recruited between 2013 and 2017, but it was actually the counsellors who were randomized (92 0f them). 123 babies finally delivered before 26 weeks gestation.

A secondary outcome of this trial was: understanding of the complications of extreme prematurity, measured using a 47-question true/false knowledge test. I can’t find an example of the test anywhere, but I wonder how anyone could have a good understanding of the complications of extreme prematurity after being presented with this aid! If the true-false question was “the prevalence of blindness among survivors born before 26 weeks is 1 to 15%” then retaining the knowledge provided by the decision aid would give you the wrong answer!

Even though parents may be more able to recollect the information they are given if it is presented differently (with the decision aid), that may have little or no impact on the decisions that are made, or how definite parents are about their decisions.

A contrasting question was asked by Marlyse Haward a few years ago: whether framing the same data as either positive or negative affected potential decisions made. Their group compared wishes for intensive, compared to comfort, care after receiving identical scenarios, one group received a document which mentioned the chances of survival and being without disability, the other received a document which described death and handicap rates. The information was identical except for the following section: “25 out of 100 infants will survive if provided intensive care. Of those who survive, 15 out of the 25 infants will not have severe developmental disabilities.” The negative version was: “75 out of 100 infants will die even if provided intensive care. Of those who don’t die, 10 out of the 25 infants will have severe developmental disabilities.”

In that study, 3/4 of the respondents (volunteers who weren’t in that situation) overall preferred active intervention and the remaining 1/4 preferred comfort care. When the same data were posed as positive, more respondents preferred active intervention than when the data were presented as death and disability. The impact wasn’t huge, but it seemed to be there.

Putting all these results together, it seems that the actual percentages of good and bad outcomes presented have very little impact on decision-making. The 2 things that matter are the pre-existing underlying values of the parents and whether the person doing the counselling thinks that the outcomes are good or bad.

One illustration of the importance of that second factor is found in the studies from the NICHD neonatal network. In some centres, 100% of babies who deliver at 22 completed weeks gestation receive comfort care. In other centres, 100% of those born alive get active intervention. I am sure that physicians and other caregivers in each centre believe that they practice shared decision making. I wouldn’t be surprised to learn that the counselling professionals in centres with universal intervention emphasize survival and good quality of life, whereas in the other centres they emphasize mortality, suffering and “mental disability”. Even if both groups describe a possible 30% survival, for example. (Of course, some centres just say “we don’t resuscitate at 22 weeks”).

Efforts to improve information transfer probably only improve information transfer, but don’t change the decisions, nor even how certain parents are about the decisions that they make.

I think repeating the German study with more extreme values of survival would probably reveal different preferences, but then the relevance to the real world decisions that we make would become less. Presenting survival of over 90% at 25 weeks among females with good prognostic factors, compared to below 5% at 21 weeks and 0 days for a growth-restricted boy might well reveal that parents respond to those figures with different preferences, but, in the range of outcomes where counselling and decision-making usually occur, the pre-existing values and beliefs of the parents are probably much more important than outcome percentages or lists of complications. And even more important is how those outcomes are presented, as being either a chance of a good life or a probability of suffering, death or disability.


About Keith Barrington

I am a neonatologist and clinical researcher at Sainte Justine University Health Center in Montréal
This entry was posted in Neonatal Research and tagged . Bookmark the permalink.

1 Response to Do estimates of survival change decisions made?

  1. Hi Keith,

    Honestly I am surprised looking at the result of this study by Kidzun et a. It just emphasizes that counselling is about how the information is being delivered (positive vs negative outlook) and most definitely should be based on the centre’s experience and outcome data in managing extremely premature infants, which will have a huge impact on the shared decision-making process with parents.
    However, for this particular study, I am interested to hear your thoughts on how they decided to only recruit 64 participants. I had a look into the supplementary data specifically at the sample size justification which I have included below;

    Sample size justification
    Based on the current NICHD network data set, it can be estimated that at 22 + 6/7 weeks of gestation (scenario A) ~ 50% of all preterm babies receive life-sustaining treatments after birth (Rysavy et al., 2015). One week later, at 23 + 6/7 weeks of gestation (scenario B), ~ 80% of children receive life-sustaining treatments. The primary hypothesis that should be investigated is that the difference in maternal preferences between the two groups does not reflect this 30% difference. I.e. with p1 and p2, the rates of maternal preference for life- sustaining measures, the null hypothesis H0:{p2-p1>=30%} should be tested against the alternative hypothesis H1:{p2-p1<30%}. At least 64 subjects are needed to show with a power of 90% that the rate of active (life-sustaining) treatment decisions in scenario A is less than 30% lower than in scenario B. This is calculated assuming that 80% of subjects in both groups opt for life-sustaining treatments. If the preference rate in scenario A is only 70%, 116 subjects are needed for a power of 90%. Since no preliminary data on the preference rates is available, up to 116 patients will be recruited if feasible.
    In order to recruit 64 subjects, at least 85 patients must be contacted, assuming a rejection rate of 20% and a lack of evaluability for 5% of the subjects.

    Acknowledging my poor statistic knowledge, although a minimum number of 64 participants is needed to show the difference of less than 30% between the scenarios (with a power of 90%), this number is based on the assumption that 80% of subjects in both groups opt for life-sustaining treatments. However, their results showed only 46.9% and 50% (in the 60% and 30% survival group respectively) opted for active care in scenario A which is less than 70%. Does this mean the study is underpowered?

    Looking forward to hear your thoughts!

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.