1. Confounding factors and choice of controls in studies of immune activation and inflammation
Caroline Sabin

2. Potential conflicts of interest
Over the past year, I have received funding for membership of Data Safety and Monitoring Boards, Advisory Boards and for the preparation of educational materials from the following companies:
Gilead Sciences
ViiV Healthcare
Janssen-Cilag

3. Age of UK CHIC participants
UK CHIC dataset (S Jose, C Sabin)

4. Age-associated co-morbidities
Literature abundant with studies reporting that HIV causes ‘premature ageing’ or that co- morbidities occur at an earlier age in PLWH

5. Age-associated co-morbidities
Literature abundant with studies reporting that HIV causes ‘premature ageing’ or that co- morbidities occur at an earlier age in PLWH

6. Age-associated co-morbidities
Literature abundant with studies reporting that HIV causes ‘premature ageing’ or that co- morbidities occur at an earlier age in PLWH

7. Age-associated co-morbidities
Literature abundant with studies reporting that HIV causes ‘premature ageing’ or that co- morbidities occur at an earlier age in PLWH
Search continues for biological mechanisms that drive this apparent increased risk
‘Inflammageing’?
Altered gut microbiota?
Mitochondrial dysfunction?
Immunosenescence?
But in our rush to establish mechanisms, have we forgotten the basic rules of epidemiology?

8. Bias due to confounding
Occurs when a spurious association arises (or is hidden) due to a failure to fully adjust for factors related to both the risk factor and outcome ?
Factor of
interest
Outcome

9. Bias due to confounding
Occurs when a spurious association arises (or is hidden) due to a failure to fully adjust for factors related to both the risk factor and outcome
Confounding
factor ?
Factor of
interest
Outcome

10. Bias due to confounding
Occurs when a spurious association arises (or is hidden) due to a failure to fully adjust for factors related to both the risk factor and outcome
Peanuts,
crisps, cheese ?
Drinking wine
Weight gain

11. Confounding
PLWH have very different characteristics to the general population, including increased risk of:
sexually transmitted infections
viral coinfections
smoking
recreational drug use, etc.
Could these other factors confound associations with co-morbidities and/or bio-markers?

12. Co-morbidities are often multi-factorial

13. Confounding
PLWH have very different characteristics to the general population, including increased risk of:
sexually transmitted infections
viral coinfections
smoking
recreational drug use, etc.
Could these other factors confound associations with co-morbidities and/or bio-markers?
Is it true that HIV causes premature ageing – or is this finding simply a result of unmeasured confounding due to the selection of inappropriate control groups?

14. Typical ‘control’ groups
General population
‘Healthy controls’ (including those from lab, research team, etc.)
Blood bank donors

15. Age at onset of co-morbidity
Median age at diagnosis = 67.5 years







 20 30 40 50 60 70 80
Age
Sabin CA, Reiss P. AIDS 2017 ; 31(Suppl 2): S121-S128.

16. Age at onset of co-morbidity
Median age at diagnosis = 57.5 years







 20 30 40 50 60 70 80
Age
Sabin CA, Reiss P. AIDS 2017 ; 31(Suppl 2): S121-S128.

17. Age at onset of co-morbidity
VACS VC – Adjusted difference in mean at diagnosis
Event
No. events
Mean age diagnosis
Crude diff.
Adjusted diff.
95% CI MI
HIV-ve
308
56.0
0.2
-0.11
-0.59, +0.37
HIV+ve
291
56.2
End-stage renal disease
688
59.4
-3.4
-0.46
-0.86, -0.07
447
NADC
2708
58.9
-1.1
-0.10
-0.30, 0.10
1471
57.8
HIV-associated cancers
826
58.6
-2.0
-0.22
-0.52, 0.08
732
56.6
Althoff KN et al. Clin Infect Dis 2015;60:

18. The Co-morBidity in Relation to Aids (COBRA) Collaboration
Formed from sub-studies of POPPY and AGEhIV:
To investigate association between HIV infection and AANCC
To elucidate possible causative links between HIV and AANCC
To clarify potential pathogenic mechanisms underlying any links identified
To identify biomarkers that may be used for better prevention, treatment and management of AANCC

19. Inclusion criteria

20. CD4 and CD8 T cell senescence
HIV-positive
Blood-bank donors N 40 35
Age (yrs), median (IQR)
58 (53-63)
58 (52-65)
Male sex, % 90
51.4
African origin, %
12.5
n/a
MSM, %
80.0
CMV+ve, %
95.0
22.9
Anti-CMV IgG
50.9 ( )
11.3 ( )
High avidity anti-CMV IgG
30.7 ( )
10.7 ( )
Booiman T et al. (Submitted)

21. CD4 and CD8 T cell senescence
Booiman T et al. (Submitted)

22. CD4 and CD8 T cell senescence
HIV-positive
HIV-negative
Blood-bank donors N 40 35
Age (yrs), median (IQR)
58 (53-63)
59 (53-64)
58 (52-65)
Male sex, % 90
92.5
51.4
African origin, %
12.5
2.5
n/a
MSM, %
80.0
75.0
CMV+ve, %
95.0
77.5
22.9
Anti-CMV IgG
50.9 ( )
23.9 ( )
11.3 ( )
High avidity anti-CMV IgG
30.7 ( )
13.3 ( )
10.7 ( )
Booiman T et al. (Submitted)

23. CD4 and CD8 T cell senescence
Booiman T et al. (Submitted)
Booiman T et al. (Submitted)

24. CD4 and CD8 T cell senescence
HIV-positive
HIV-negative
Blood-bank donors N 40 35
Age (yrs), median (IQR)
58 (53-63)
59 (53-64)
58 (52-65)
Male sex, % 90
92.5
51.4
African origin, %
12.5
2.5
n/a
MSM, %
80.0
75.0
CMV+ve, %
95.0
77.5
22.9
Anti-CMV IgG
50.9 ( )
23.9 ( )
11.3 ( )
High avidity anti-CMV IgG
30.7 ( )
13.3 ( )
10.7 ( )
Booiman T et al. (Submitted)

25. CD4 and CD8 T cell senescence
Booiman T et al. (Submitted)

26. Bias due to confounding
Occurs when a spurious association arises (or is hidden) due to a failure to fully adjust for factors related to both the risk factor and outcome
CMV infection ?
Activation
markers
HIV

27. Age advancement
Biological age derived using set of 10 biomarkers identified through EU FP7 MARK-AGE project (
Age advancement: biological - chronological age
De Francesco, et al. Poster TULBPEB19, IAS 2017

28. Age advancement HIV-positive HIV-negative Blood-bank donors N 134 7935
Age (yrs), median (IQR)
56 (51-62)
57 (52-65)
59 (52-65)
Male sex, % 93 92 51
African origin, % 12 3
n/a
MSM, % 78 75
Recreational drug use, % 33 23
Chronic HBV, %
Chronic HCV, % 2
CMV infection, % 98 80
De Francesco, et al. Poster TULBPEB19, IAS 2017

29. Age advancement
De Francesco, et al. Poster TULBPEB19, IAS 2017

30. Are we comparing apples to oranges?

31. Summary
Choice of control group can affect our interpretation of studies of the impact of HIV on co-morbidities and/or bio-markers
PLWH often exhibit different lifestyles, and it be a consequence of these, rather than HIV itself, that drives apparent ageing
Whilst the association with HIV may be weaker than previously reported, this may actually simplify our search for mechanisms

32. The CoBRA Collaboration
Particular thanks to Peter Reiss, Neeltje Kootra and Davide de Francesco for provision of slides