<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Joe Faith on Medium]]></title>
        <description><![CDATA[Stories by Joe Faith on Medium]]></description>
        <link>https://medium.com/@joedotfaith?source=rss-1405ed84f15f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*L7BfVfGBDZWvcFZ2.jpeg</url>
            <title>Stories by Joe Faith on Medium</title>
            <link>https://medium.com/@joedotfaith?source=rss-1405ed84f15f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 14:37:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@joedotfaith/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Data Science in the Real World]]></title>
            <link>https://medium.com/@joedotfaith/data-science-in-the-real-world-dbbb05eb59a8?source=rss-1405ed84f15f------2</link>
            <guid isPermaLink="false">https://medium.com/p/dbbb05eb59a8</guid>
            <category><![CDATA[ds-in-the-real-world]]></category>
            <category><![CDATA[causal-inference]]></category>
            <dc:creator><![CDATA[Joe Faith]]></dc:creator>
            <pubDate>Fri, 17 Jan 2020 09:43:56 GMT</pubDate>
            <atom:updated>2020-01-17T15:08:37.245Z</atom:updated>
            <content:encoded><![CDATA[<p><a href="https://towardsdatascience.com/data-science-in-the-real-world/home">Data Science in the Real World</a></p><h3>Causal inference and the woman who saved the US from the Thalidomide disaster</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/307/1*kIoh6KI8FVxm07k3OcMXQA.png" /><figcaption>Frances Oldham Kelsey. Hero.</figcaption></figure><p><strong>In the US, Thalidomide is known, if at all, as a treatment for leprosy. But in Europe and Australia it is a byword for medical disaster.</strong></p><p>Thalidomide was discovered in 1953 by Grünenthal, the German pharmaceutical company, and marketed as an effective and safe sedative for treating nausea, vomiting and insomnia. And in 1957 it started to be prescribed for morning sickness in pregnant women in Europe, Australia, and New Zealand. The results were catastrophic because, as it turned out, the drug interferes with angiogenesis, the production of new blood vessels. This is not a significant problem for adults with mature vascular systems, but is critical for developing fetuses.</p><p>Thousands of babies were born with severe abnormalities: truncated, flipper-like limbs (phocomelia), or damage to the brain, eyes, ears, or face. Around half died within a few months. Estimates of the number of victims range from 10 to 20,000, though the real number could be much higher if those who never reached full term are included.</p><p>The survivors and families struggled. First, with the physical pain and disability, but then to get recognition and compensation. In Germany, it required a class action suit for negligent homicide to get compensation. In the UK it required many years campaigning by the <em>Insight</em> team of the <em>Sunday Times</em> newspaper (of which I’m proud to say <a href="https://www.theguardian.com/education/2018/oct/04/nicholas-faith-obituary">my father</a> was a member).</p><p>Even into the 1980s the term ‘flid’ — short for ‘thalidomide’ — was a common playground insult in the UK for someone clumsy or incapable.</p><p>The US was spared from the Thalidomide disaster by the brave — and insightful — actions of Frances Oldham Kelsey. Despite intense and repeated pressure from the US distributors of the drug, Kelsey, then a medical reviewer for the FDA, refused to grant a license.</p><p><strong>Why did Oldham Kelsey refuse a license? What did she understand that the European regulators did not?</strong></p><p>Frances Oldham Kelsey initially had concerns that the safety trials were rushed by the distributors to get the drug, branded as Kevadon in the US, to market before the 1960 Christmas season when demand for sedatives soar. (The stress of Christmas means it is the peak period for both suicides and divorce.) “The clinical reports were more on the nature of testimonials,” <a href="https://permanent.access.gpo.gov/lps1609/www.fda.gov/fdac/features/2001/201_kelsey.html">said Kelsey later</a> “rather than the results of well-designed, well-executed studies.” She repeatedly asked for the safety trials to be repeated.</p><p>But two other insights hardened her reluctance into outright rejection. The first was that she noticed that Thalidomide was not very well absorbed in most patients. This suggested that the dosage recorded in the trials would not necessarily reflect how much of the drug the patient was actually exposed to — so the reported toxicity would not be reliable. Some conditions could change the absorption rate, which could then produce toxic effects.</p><p>The second was due to a letter in the British Medical Journal from a physician who reported patients who had taken Thalidomide for a long time experiencing painful tingling in their hands and feet. This suggested to Oldham Kelsey that Thalidomide could be damaging nerves — which meant it could also damage a developing fetus. The license application was rejected. And by November 1961, reports from Europe confirmed that Thalidomide was having a devastating effect, but the US was spared.</p><p>In 1962 President John F. Kennedy presented Oldham Kelsey with the medal for Distinguished Federal Civilian Service — the highest honorary award that the Federal Government can grant a career civilian employee.</p><p><strong>For me, the story of Thalidomide in the US is a great example of the importance of a causal understanding of a problem in order to make sense of the data we observe. The clinical trial data showed no association between Thalidomide and congenital birth defects, but Oldham Kelsey used her understanding of the causal processes of neuralgia and drug adsorption to argue that the data could not be relied on.</strong></p><p>Determining the safety of drugs during pregnancy is hard, since most randomised controlled drug trials — the ‘gold standard’ of drug testing — deliberately exclude pregnant women, and results from animal studies do not necessarily translate to humans. And so the <a href="https://www.fda.gov/media/71368/download">FDA recommends</a> that observational and case studies are used instead. These involve pregnant women who are exposed to drugs, either because they did not realise they were pregnant, or they became pregnant with a condition that required continued treatment. But the problem with observational and case studies is that it is hard to control for <em>confounders </em>— factors that affect both the treatment and the effect we want to test for. In this case, the only pregnant women likely to receive a drug are those who have the condition being treated, in which case we won’t know if the outcomes — such as harm to the baby — is due to the drug or the condition it was given to treat. If a women has the condition then there is no additional in her continuing to take the drug.</p><p>So eliminating confounding effects is a key problem for determining the effectiveness and safety of drugs for pregnant women, and ensuring we avoid another Thalidomide. Until relatively recently it wasn’t thought possible to robustly disentangle the effect of confounders from observational studies — this was one of the problems that plagued the attempt to prove that smoking caused lung cancer, for example. But the invention of the statistical techniques of causal inference, particularly the work of <a href="http://ftp.cs.ucla.edu/pub/stat_ser/r350.pdf">Judea Pearl</a> from the 1990s onwards, showed how the effects of causation could be systematically distinguished from correlations produced by confounders. (Pearl’s <a href="http://bayes.cs.ucla.edu/WHY/"><em>The Book of Why</em></a> is a great introduction). And a recent article in the journal of <a href="https://onlinelibrary.wiley.com/doi/full/10.1002/pds.4336"><em>Pharmacoepidemiology and Drug Safety</em></a> calls out how causal inference is now vital for robust pregnancy medication safety studies.</p><p>For example, up to 10% of pregnant women are exposed to SSRI anti-depressants during pregnancy. Earlier studies had shown no association between the drug and lower birth weight or premature births, but <a href="https://academic.oup.com/ije/article/45/6/2018/2617170">an observational study of 30,000 Norwegian mothers</a>, including just 194 who had been exposed to SSRIs, analysed the effect of a wider range of confounding factors — from the amount of sick leave the mothers took, to their histories of depression — and found a small, but clear, effect.</p><p><strong>Whenever I read reports like this I like to think of Frances Kelsey nodding in approval, quietly but persistently reminding us of the importance of understanding what is going on behind the data.</strong></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dbbb05eb59a8" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>