David M Kent

Simple country doctor, interested in prediction. Director, Tufts MC PACE Center (=Predictive Analytics & Comparative Effectiveness);

[David M Kent @Tufts_PACE]

In @JAMA_current we present a 6-trial IPDMA showing that a causal classification system—based on 8 easily obtainable variables—turns the effect of device closure off and on in pts w/ PFO-associated stroke.
Thread summarizing ~15 yrs of research. 1/24
https://t.co/6PYqF5FWZS

[JAMA @JAMA_current]

💬 Viewpoint: Adversarial collaboration is designed to move beyond traditional replication and consensus-building by engaging credible experts from opposing viewpoints to jointly resolve disputes in #BiomedicalScience.

https://t.co/ZK51kt3Vgh

[JAMA @JAMA_current]

Patent foramen ovale (PFO) is present in approximately 25% of all adults and is a common cause of stroke in young and middle-aged patients.

📝 This Review discusses PFO-associated stroke management.

https://t.co/yr4xTLbuXh

[David M Kent @Tufts_PACE]

@VickersBiostats congrats!

[The Greenwall Foundation @GreenwallFdn]

Our December newsletter is out! 📰

Check your inbox for updates featuring @Tufts_PACE and @KerenLadin in @npjDigitalMed, @neva9257 in @J_Law_Biosci, @EthicsdoctorP in @Medscape, and so much more.
https://t.co/OpJLne8pz8

[Annals of Int Med @AnnalsofIM]

An analysis found that the clinical net benefit of using race-aware models over race-unaware models for disease risk prediction was smaller than expected. https://t.co/C56xMD1adh

@madisoncoots @HarvardMed @Soroush_Saghaf @Tufts_PACE @5harad

[Madison Coots @madisoncoots]

The use of race in clinical risk models is heavily debated. While race-aware models can be more accurate, some are concerned about reinforcing racialized views of medicine. In our paper in @AnnalsofIM, we offer a new perspective on this debate. 🧵👇 https://t.co/1e8tJ75wVg

[David M Kent @Tufts_PACE]

@ehlJAMA Thank you for this comment Ed. I will try to take a look when I come off clinical service.

[David M Kent @Tufts_PACE]

A very reasonable take by Frank.

It is inconceivable to me, a simple country doc, that treatment effects will generally be homogeneous across patients, who are so heterogeneous, including in their outcome mechanisms.

Yet reliably uncovering HTE in RCTs powered for their main effect, without strong prior information, can seem a fool’s errand; it is surprisingly difficult.

Nevertheless, that seems a poor excuse to settle for ATEs when our patients expect more.

Let’s respect both terms of this important dilemma.

1/x

[Frank Harrell @f2harrell]

@DanMarkMD One size does not fit all. But to the resolution of available data and given that RCT sample sizes are barely large enough to estimate simple average effects, HTE might as well not exist for most situations.

[David M Kent @Tufts_PACE]

@doc_BLocke maybe too complicated for twitter, but yes specifics definitely matter...

[David M Kent @Tufts_PACE]

@DanMarkMD I agree with this that this represents not a lack of HTE, but a failure of science.

[David M Kent @Tufts_PACE]

@drjohnm I am not sure why my prior tweet did not link to the paper, but let's try again.

https://t.co/1SdHlOIOk1

[David M Kent @Tufts_PACE]

We found 14 re-analyses that delivered evidence that we considered both credible and clinically important. Many of these were IPDMA, combining databases.

Among these 14, there were 10 simple ‘risk modeling’ studies. Interestingly, >half of these showed variation in both absolute AND relative effects across the different risk strata.

There were also 4 ‘effect models’, that more flexibly explored treatment interactions with individual variables for subgroup identification, with credible and clinically important results. These 4 have been validated externally. Most 'effect models', however, yielded 'exploratory' results awaiting confirmation--or more likely refutation.

3/x

[David M Kent @Tufts_PACE]

We’ve promoted a ‘poor man’s’ approach to predicting more patient-specific effects, just stratifying a trial by risk (i.e. ‘risk modeling’), in the PATH statement. W/ @jvselby, we’ve recently review >40 predictive HTE analyses of RCTs that cited PATH. Our work-in-progress is posted here on medrxiv (comments welcomed):
Impact of the PATH Statement on Analysis and Reporting of Heterogeneity of Treatment Effect in Clinical Trials: A Scoping Review | medRxiv
2/x

[David M Kent @Tufts_PACE]

@haroldpollack @thehowie astounding how effective that word genocide has been.

a good german friend, lovely person, said to me (sic): "its so sad to see the jews now doing to others what has been done to them."

me: "ah... no."

[David M Kent @Tufts_PACE]

@f2harrell @HJHansford @statsepi Your future in marketing is not very promising…

[David M Kent @Tufts_PACE]

@f2harrell @drjohnm we are incentivized by our funders to be optimists, both with regards to the effects size and recruitment

[David M Kent @Tufts_PACE]

It was great fun to host @VickersBiostats for our annual CTS Symposium, where he connected with @jpkassirer whose work 50 years ago with Steve Pauker on the threshold approach to decision making laid the foundation for Decision Curve Analysis.

Unfortunately, Charles Pierce was unavailable to join the photo...

[David M Kent @Tufts_PACE]

In which @VickersBiostats attempts to explain the nuances of decision curve analysis to gastroenterologists, hospitalists, hematologists, etc

Not to be missed!

[Karen Freund MD @KarenFreundMD]

This year's @tuftsctsi GR sure to generate discussion, with @VickersBiostats from @MSKCancerCenter @WeillCornell.
@Tufts_PACE @TuftsMedicalCtr @TuftsMedSchool

[David M Kent @Tufts_PACE]

@haroldpollack Please stop validating pernicious cultural stereotypes…

[David M Kent @Tufts_PACE]

@ZaidJilani @abbydphillip 61% of young (18 to 34) Jews support US military aid to Israel; 26% oppose.
However, the from-the-river-to-the-sea ‘asajews’ are a small fringe, correctly characterized by the guest on cnn

https://t.co/d3UBuI0j6Z(89%25%20vs.