Epidemiology as driver of personalised transplantation care

Dr. Roemer J. Janse

R.J.Janse-5@umcutrecht.nl

UMC Utrecht

Kidney transplantation and cardiovascular disease

Post-KTx risks

Post-KTx risks

Post-KTx risks

Post-KTx risks

Jardine et al., 2011 Lancet; DOI: 10.1016/s0140-6736(11)61334-2

Post-KTx risks

Jardine et al., 2011 Lancet; DOI: 10.1016/s0140-6736(11)61334-2

Ying et al., 2020 J Am Soc Nephrol; DOI: 10.1681/ASN.2020050566

Post-KTx risks

Johansen et al., 2026, Am J Kidney Dis; DOI: 10.1053/j.ajkd.2026.02.635

Cardiovascular prevention

Risk factor
Hypertensiona,b
Diabetes mellitusa,b
Smoking
Dyslipidaemiaa,b
Obesitya

^a Exacerbated by steroids
^b Exacerbated by calcineurin inhibitors

Cardiovascular prevention

Risk factor	Pharmaceutical treatment
Hypertensiona,b	ACEis, ARBs, CCBs
Diabetes mellitusa,b	Metformin, SU derivates, SGLT2is, GLP1-RAs, Insulin
Smoking
Dyslipidaemiaa,b	Statins
Obesitya	Metformin, GLP1-RAs

^a Exacerbated by steroids
^b Exacerbated by calcineurin inhibitors

Underlying evidence

Colombijn et al., 2024 JAMA Netw Open; DOI: 10.1001/jamanetworkopen.2024.0427

RCTs for evidence-based medicine

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") 

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    

Average treatment effect

The Average treatment effect (ATE) is the average effect in the total population

• The main analysis of our RCT on drug A vs. placebo yields a risk reduction of 20% for cardiovascular events
• If we prescribe drug A to this population, the population will experience 20% less cardiovascular events
• An individual may have a different risk reduction: 10%, 30%, -5%
• But the average of all those individuals will be 20%
• This is useful for people who treat populations (policy makers)
• This is less useful for people who treat individuals (clinicians)

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    

Conditional average treatment effect

The conditional average treatment effect (CATE) is the average treatment effect in a subgroup of the population

• We perform a subgroup analysis on biological sex, which yields a risk reduction of 30% for women and 12% for men
• If we prescribe drug A to this population, the population will experience 20% less cardiovascular events
• But the population of women will experience 30% less cardiovascular events
• And the population of men will experience 12% less cardiovascular events
• An individual woman or an individual man may still have a different risk reduction: 10%, 30%, -5%
• This is slightly useful for people who treat populations (policy makers)
• This is slightly useful for people who treat individuals (clinicians)

Conditional average treatment effect

Conditional average treatment effect

Conditional average treatment effect

Conditional average treatment effect

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")

Individualised treatment effects

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> magic["?"]

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> magic["?"]
    magic --> ite(["Individual treatment effect"])

Individual treatment effect

The individual treatment effect is the treatment effect for an individual

• We assign drug A to an individual
• We open an alternative universe where everything is exactly the same, but we do not assign drug A to that individual
• The difference in their outcomes is the individual treatment effect

Individualised treatment effect

The individualised treatment effect (ITE) is a highly specific conditional treatment effect

• We calculate the effect (CATE) based on many variables at once
• The CATE becomes so specific, that we consider it individualised
• For instance, drug A will give a 53-year old woman with diabetes, no cardiovascular history, a kidney function of 55 mL/min/1.73m², with an increased lymphocyte count and who is using tacrolimus, mycophenolate, and basiliximab, as well as a renin-angiotensin-system inhibitor, a risk-reduction of 35%
• This is useful for people who treat individuals (clinicians)

Individualised treatment effect

To calculate the ITE, we can use two methods:

Risk modelling:
- Using a (valid) existing prediction model, we perform a subgroup analysis by predicted outcome risk
Because we rarely make many subgroups, I would not consider this an ITE

Individualised treatment effect

To calculate the ITE, we can use two methods:

Effect modelling:

• Directly predict the outcome if an individual was treated
• Directly predict the outcome if an individual was not treated
• Subtracting these predictions gives the ITE

Individualised treatment effect

Important

Both methods assume that there is no confounding left between the groups

If there is no confounding, we can use untreated individuals to predict what would have happened to the treated individuals, and vice versa

Using sophisticated study designs/statistical techniques, we can remove confounding in our study population

Individualised treatment effect

Important

The ITE is a prediction: it is what we expect, but we do not know for sure if it is true (and never will)

We thus need to validate the prediction, using special methods that take into account our limited knowledge

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind("Individual people")

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> path["Risk/effect modelling"]
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind("Individual people")

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> path["Risk/effect modelling"]
    path --> ite(["Individualised treatment <br> effect (ITE)"])
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind("Individual people")

RCT-based evidence

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> path["Risk/effect modelling"]
    path --> ite(["Individualised treatment <br> effect (ITE)"])
    ite --> ind("Individual people")
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind

A final note

Sample size struggles

flowchart TB
    rct("Randomised controlled trial") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> path["Risk/effect modelling"]
    path --> ite(["Individualised treatment <br> effect (ITE)"])
    ite --> ind("Individual people")
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind
    
%% Custom style for node CATE and ITE
style cate fill:#F9A03F,stroke:#C36F09
style ite fill:#E79E9C,stroke:#6F1D1B

Sample size struggles

flowchart TB
    rct("Pooled RCTs/<br>observational data") --> ma["Main analysis"]
    ma --> ate(["Average treatment effect (ATE)"])
    ate --> pop("Population")
    rct --> sa["Subgroup analysis"]
    sa --> cate(["Conditional average <br> treatment effect (CATE)"])
    cate --> subpop("Subgroups in population")
    rct --> path["Risk/effect modelling"]
    path --> ite(["Individualised treatment <br> effect (ITE)"])
    ite --> ind("Individual people")
    rct --> magic["?"]
    magic --> more_magic(["Individual treatment effect"])
    more_magic --> ind

Thank you!

• Web: rjjanse.github.io
• Contact: r.j.janse-5@umcutrecht.nl
• Slides: rjjanse.github.io/talks