Links 11/10/24
Recent health and methods readings
HPV vaccines’ possible risks, controversies overlooked
HPV vaccines could eliminate cervical cancer, Saloni Dattani and Veronika Samborska write at Our World in Data. The highlighted benefits, however, may be offset by harms — a risk on which the OWD coverage is silent.
This silence veers into inaccuracy when the authors say that Japan withdrew the HPV vaccine “because of inaccurate media reports that the vaccine was linked to pregnancy complications.” As I wrote recently, HPV vaccination during the periconceptional or gestational period may indeed substantially increase miscarriage risk according to a recent meta-analysis.
These safety concerns are not sui generis. Rather, they are part of a long-running, high-profile controversy regarding possible adverse HPV vaccine responses.
As part of this dialogue, leading medical methodologist Peter Gøtzsche has criticized the European Medicines Agency for spinning away concerns about rare but serious neurological harms stemming from possible autoimmune reactions to the vaccine — and alleging bias and nontransparency in evaluating evidence in favor of the vaccine manufacturers amid conflict of interest policy violations. Proponents invoked Wakefield’s MMR-autism fraud. Critics critiqued a Cochrane HPV vaccines review; Cochrane responded. Critics have continued to advocate for open, independent discussion of vaccine safety concerns, reiterate their concerns about EMA’s mishandling of the investigation, and highlight persistent uncertainty in the available data about the relationship between HPV vaccination and neurological disorders including postural orthostatic tachycardia syndrome (POTS).
***
Pressure for earlier universal American gestational diabetes screening — and evidence-based rejection of doing the screening entirely
In a Clinical Perspective, Colorado maternal-fetal medicine doctors Kent Heyborne and Linda Barbour call for early universal gestational diabetes screening. This surprised me: It was my impression that universal screening was already the norm in well-resource places. So if universal screening is ascendant but not the norm, what’s its status internationally?
Who thinks it’s hot:
The oral glucose tolerance test is universally recommended in Australia, though coverage for Aboriginals is much lower and proponents argue early HBA1C testing could reach more women, sooner. Glucose tolerance testing also generates a huge number of false positives.
It’s recommended in Japan in the third trimester.
And in China in well-resourced areas.
It’s recommended universally in the U.S.: The American Diabetes Association has long endorsed universal GD screening — along with the American College of Obstetricians and Gynecologists guidance and U.S. Preventive Services Task Force recommendations Keyborne and Barbour are criticizing. The difference is that ACOG follows USPSTF’s logic that the evidence doesn’t support screening before week 24. So the debate here is about early screening, not universal screening.
Who thinks it’s not:
It turns out the UK National Screening Committee doesn’t recommend universal GD screening due to insufficient evidence its benefits outweigh its harms, lack of a safe and accurate screening test (criticizing the oral glucose tolerance test on safety grounds), and uncertain evidence about how to act on results. This contrasts with the U.S. debate, which centers on when (not whether) to start screening everyone. The NHS offers screening for at-risk women.
This is obviously not a comprehensive global practice survey. It’s just interesting to note the variation in the discourse. (My own read of the evidence agrees with the NSC’s, and I would add that preventive diet and lifestyle education and support for everyone makes far more sense than either mass or targeted screening approaches — with gestational diabetes screening itself a case of both.)
***
Beyond “it’s only a cry for help”: Depression as costly signal of need for more resources
It’s often said of suicidality that it’s either serious (death is intended), or it’s just a cry for help (it’s not). But what if there’s no difference in that distinction? What if an attempt to exit does not necessarily conflict with an expression of voice, in Hirschmann’s parlance? Or, in evolutionary psych terms, what if depression evolved to function as a credible signal of a cooperative species member’s need for more resources? The credibility comes from the potentially lethal costs, which — like most things in this world — key into probabilistic rather than certain binary outcomes.
So argue Ed Hagen and Kristen Syme in their recent Oxford Handbook of Evolution and the Emotions chapter “Credible sadness, coercive sadness: Depression as a functional response to adversity and strife” (via X). Some highlights...
Interpersonal conflict is a common feature of depression, anxiety, and PTSD. For instance, Delalibera et al’s 2015 systematic review found that family conflict was a major risk factor for complicated grief. In this framework, the non-zero risk of self-harm underscores the perceived severity of the individual’s situation. It’s a way for the powerless to try to change their situation. In other words, someone is not being very nice to the depressed person, and changing that requires that someone’s (or those someones’) mercy, goodwill, generosity, or cooperation.
Looking to research on other primates, Hagen and Syme frame depression as an involuntary strategy of submission and withdrawal after social defeat, meant to de-escalate. If you could escalate instead, you would, Hagen suggests based on his research (2003) correlating lower grip strength with depression. (I gotta get a Powerball.)
Does it work? Sometimes — when depression as a credible signal of need elicits more short-term aid, leading to lasting life changes. Neeleman et al observe positive life events and “fresh starts” track with depression recovery — with stronger social networks correlating with more positive life changes.
So what’s pathological about credibly signaling that you need more support when you do? Nothing, Hagen and Syme argue. Just because people dislike depression — it’s alienating to be around someone who ruminates all the time — doesn’t mean it’s a disorder. Sometimes it works to drum up essential support to change your position, in spite of other people not wanting to pay higher costs of cooperation to help you change it.
And then you get a reset — a new shot at a life you want to live. On better terms. Because the old ones were too bad to live with. But you couldn’t change them yourself.
***
Reducing antibiotic overuse through tech-assisted prescribing decisions
Michael Ribers and Hannes Ullrich showed AI-assisted prescribing lowered antibiotic overuse in service of reducing antibiotic resistance (“Complementarities between algorithmic and human decision-making: The case of antibiotic prescribing,” Quantitative Marketing and Economics, 2024, via the Hertie School Data Science Brown Bag). This raises the question of whether (and then when, why, how, and among whom) such complementary uses of technology-assisted decision-making can overcome the accuracy-error trade-off (see previous analysis and slides). Usually, statisticians agree there is no exit from it, practitioners disagree, and a pervasive lack of accessible efficacy data precludes independent analysts from answering the question.
This is a very special case study — antibiotic prescribing for UTIs in a Danish primary case setting — because we get to observe docs in the ROC space. They see a patient, write a script or don’t, and three days later get back the gold standard lab results indicating whether their initial decision to treat was right or wrong. Usually the world is messier (e.g., in security cases — where the validation problem, in which we struggle to establish what exactly to infer from a particular test or intervention, tends to persist). But what a great idea to find and build knowledge from cases like this.
Ribers and Ullrich found a hybrid approach worked best, with the algorithm better classifying hard and easy cases, and doctors achieving better accuracy in intermediate-difficulty cases (which constituted over 50% of decisions). That approach reduced overprescribing by around 1/5 (95% CI 15.1-20.8%).
We need research on how people will actually use the tech.
***
Political bias in fluoridated water coverage
Fluoride has been in the news a few times lately, and the coverage has changed to reflect changing political associations. Back in August, fluoride was under suspicion: the U.S. National Toxicology Program had suddenly gotten the memo that it’s a developmental neurotoxicant, and went on the press circuit before the publication of its meta-analysis on high fluoride and child IQ. The AP was on board. In October, a Cochrane review suggested fluoridating water might slightly improve children’s oral health by preventing dental caries; but effects have probably attenuated (e.g., since we now know to brush with fluoridated toothpaste), and the evidence doesn’t establish net benefit. The Guardian was on board.
Then Trump won the 2024 U.S. Presidential election, and the fluoride discussion took on new political significance. Because RFK Jr. might play an influential role in health policy under Trump, both are controversial figures, and RFK Jr. has come out against fluoridating water in the context of his Trump Administration affiliation. Now the Guardian coverage skews negative on removing fluoride from water, e.g., highlighting research that “found no association between drinking fluoridated water before age five and changes in behavioural development or executive functioning” — instead of a recent review by leading environmental toxicologist Philippe Grandjean finding “recent epidemiological results support the notion that elevated fluoride intake during early development can result in IQ deficits that may be considerable.”
***
Misoprostol and preterm birth risk
Joining a large literature of oft-neglected results suggesting possible future pregnancy risks of abortion, a recent article finds missed abortions that are medically treated with misopristol (one of the usual two abortion drugs) correlate with increased risk of preterm birth, C-section, and postpartum hemorrhage — but only when women didn’t space pregnancies 18 months apart (“Interpregnancy interval following missed abortion and the risk for preterm birth,” Bachar et al, Eur J Obstet Gynecol Reprod Biol, 2024 Sep:300:202-205).
Specifically:
The rates of spontaneous PTB <37 and <34 weeks of gestation were significantly higher in the short vs. long IPI cohort (16.28 % vs. 7.06 % and 6.74 % vs. 5.0 %, respectively, p < 0.05). These patients also had a higher risk for Cesarean delivery (31.63 % vs. 23.34 %, p = 0.005) and postpartum hemorrhage (4.42 % vs. 2.06 %, p = 0.029) compared to patients with IPI >18 months. The observed differences remained statistically significant even after adjusting for potential confounding variables using multiple regression analysis.
It’s worth noting, though, that the observed effect sizes here are small — single-digit.
Still, this conflicts with the current (mixed) consensus on miscarriage and pregnancy spacing, which says women don’t need to wait to try again. The disconnect could be explained by the misopristol factor, if the use of that agent heightens risks of future adverse pregnancy outcomes. Or it could reflect sampling variance or another limitation of these particular evidentiary slices of the truth as a whole.
Another recent study, also in a sample of women who had experienced natural pregnancy losses, similarly found possible future pregnancy risks associated with D&C (the typical surgical abortion method that was dominant before the pill).
This pattern of possibly concerning findings fits in the context of a larger pattern of abortion myths in which proponents claim there are no risks or even net benefits for women’s health, while opponents claim there are substantial risks — particularly for mental health. The reality is that the available evidence doesn’t establish either no risks or a causal relationship between abortion and worse outcomes. And that — given the deeply confounded nature of observational data, and the ethically fraught nature of conducting experimental research on this topic — it probably never will.
***
UK NSC on risk stratified breast cancer screening models
Mass mammography screening for early breast cancer detection doesn’t yield a proven net mortality benefit in trade for generating high false positives with substantial costs such as disfiguring, risky, and unnecessary interventions. A lot of medical organizations recommend it anyway, sometimes for political reasons — much to the chagrin of leading evidence-based medicine experts. But what if we could improve the cost-benefits balance by making the screening more targeted?
That’s the question addressed in a recent UK National Screening Committee workshop, now reported in a peer-reviewed publication. The current screening is age-based. Experts reviewed recent alternative risk prediction and cost-benefits models. Among the published models, none appears to tackle the problem with age-based stratification alone: causation.
At a general methodological level, the models discussed don’t appear to account for causality. At a field-specific level, this is a problem because causal diagramming (DAGs) would reveal that we have to specify if we care about cases or deaths. If we care about deaths, we care about metastases. And one of the main risk factors for breast cancer metastases is parity.
Postpartum breast cancer metastasizes at much higher rates than other types. The effect probably doesn’t cross over from positive to negative for around 26 years (so it affects many of the 50-70 year olds in the current UK mass screening protocol). And yet, we can’t neatly code age and parity as risk factors, throw them in a model, and have a nice day without making further assumptions about how the two causally relate.
The reason is that they manifest opposing base rate trends: Mothers face heightened risks of deadly breast cancer that rise with age at first birth — older first-time mothers risk more postpartum breast cancer cases. But younger mothers diagnosed with postpartum breast cancer suffer higher fatality likelihood by case. The usually posited causal story here is that cancer risks increase with age anyway, pregnancy/lactation differentiates breast tissue and differentiated tissue is less dangerous vis-a-vis cancer, but the remodeling process that differentiation entails is also quite risky; and probably the younger you are when you have a detected cancer, the more virulent it was to overcome comparatively strong homeostasis to be detected.
So you have to diagram causality to stratify breast cancer screening in relation to age and parity, but this doesn’t appear to have been done. The committee also doesn’t appear to have addressed concerns that screenings may unintentionally, causally contribute to cancer cases and/or cancer progression (e.g., through mammography compression or follow-up needle biopsy circulating cancer cells to other tissues they can then invade, or through radiation exposure — particularly from repeated screenings). This possible iatrogenesis complicates calculations of net costs and benefits.
Anyway, the meeting report concludes:
The workshop was not intended to present a balanced evaluation of how to deliver screening in future. Areas for further work identified included methods for comparing models to assess accuracy, the optimum risk assessment tools, the digital screening infrastructure, acceptability of stratification, choice of screening test and reducing inequalities.
***
Sciencing, libertarian-style
A large proportion of scientists self-experiment, many seriously, but institutional cultures tend to discourage it. (H/t Nature coverage of this case report of a scientists self-treating breast cancer recurrence by injecting viruses to stimulate immune response, critiqued here by Vinay Prasad.)
***
Sciencing, communist-style
Sci-Hub founder Alexandra Elbakyan reports she’s working on an article about reconceptualizing “black open access” to reflect the heterogeneity in free science since 2001.