The occurrence of Simpson's paradox if site-level effect was ignored in the TREAT Asia HIV Observational Database

TREAT Asia HIV Observational Database (TAHOD)

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Objectives In multisite human immunodeficiency virus (HIV) observational cohorts, clustering of observations often occurs within sites. Ignoring clustering may lead to “Simpson's paradox” (SP) where the trend observed in the aggregated data is reversed when the groups are separated. This study aimed to investigate the SP in an Asian HIV cohort and the effects of site-level adjustment through various Cox regression models. Study Design and Setting Survival time from combination antiretroviral therapy (cART) initiation was analyzed using four Cox models: (1) no site adjustment; (2) site as a fixed effect; (3) stratification through site; and (4) shared frailty on site. Results A total of 6,454 patients were included from 23 sites in Asia. SP was evident in the year of cART initiation variable. Model (1) shows the hazard ratio (HR) for years 2010–2014 was higher than the HR for 2006–2009, compared to 2003–2005 (HR = 0.68 vs. 0.61). Models (2)–(4) consistently implied greater improvement in survival for those who initiated in 2010–2014 than 2006–2009 contrasting findings from model (1). The effects of other significant covariates on survival were similar across four models. Conclusions Ignoring site can lead to SP causing reversal of treatment effects. Greater emphasis should be made to include site in survival models when possible.

Original languageEnglish
Pages (from-to)183-192
Number of pages10
JournalJournal of Clinical Epidemiology
Volume76
DOIs
Publication statusPublished - 2016 Aug 1

Fingerprint

HIV
Databases
Survival
Proportional Hazards Models
Cluster Analysis
Cohort Effect
Therapeutics

All Science Journal Classification (ASJC) codes

  • Epidemiology

Cite this

@article{ec7f9b4854b2406cbe62a201dddf649b,
title = "The occurrence of Simpson's paradox if site-level effect was ignored in the TREAT Asia HIV Observational Database",
abstract = "Objectives In multisite human immunodeficiency virus (HIV) observational cohorts, clustering of observations often occurs within sites. Ignoring clustering may lead to “Simpson's paradox” (SP) where the trend observed in the aggregated data is reversed when the groups are separated. This study aimed to investigate the SP in an Asian HIV cohort and the effects of site-level adjustment through various Cox regression models. Study Design and Setting Survival time from combination antiretroviral therapy (cART) initiation was analyzed using four Cox models: (1) no site adjustment; (2) site as a fixed effect; (3) stratification through site; and (4) shared frailty on site. Results A total of 6,454 patients were included from 23 sites in Asia. SP was evident in the year of cART initiation variable. Model (1) shows the hazard ratio (HR) for years 2010–2014 was higher than the HR for 2006–2009, compared to 2003–2005 (HR = 0.68 vs. 0.61). Models (2)–(4) consistently implied greater improvement in survival for those who initiated in 2010–2014 than 2006–2009 contrasting findings from model (1). The effects of other significant covariates on survival were similar across four models. Conclusions Ignoring site can lead to SP causing reversal of treatment effects. Greater emphasis should be made to include site in survival models when possible.",
author = "{TREAT Asia HIV Observational Database (TAHOD)} and Awachana Jiamsakul and Kerr, {Stephen J.} and Ezhilarasi Chandrasekaran and Aizobelle Huelgas and Sineenart Taecharoenkul and Sirinya Teeraananchai and Gang Wan and Ly, {Penh Sun} and Sasisopin Kiertiburanakul and Matthew Law and Ly, {P. S.} and V. Khol and Zhang, {F. J.} and Zhao, {H. X.} and N. Han and Lee, {M. P.} and Li, {P. C.K.} and W. Lam and Chan, {Y. T.} and N. Kumarasamy and S. Saghayam and C. Ezhilarasi and S. Pujari and K. Joshi and S. Gaikwad and A. Chitalikar and Merati, {T. P.} and Wirawan, {D. N.} and F. Yuliana and E. Yunihastuti and D. Imran and A. Widhani and S. Oka and J. Tanuma and T. Nishijima and Choi, {J. Y.} and S. Na and Kim, {J. M.} and Sim, {B. L.H.} and Gani, {Y. M.} and R. David and A. Kamarulzaman and {Syed Omar}, {S. F.} and S. Ponnampalavanar and I. Azwa and M. Mustafa and N. Nordin and R. Ditangco and E. Uy and R. Bantique",
year = "2016",
month = "8",
day = "1",
doi = "10.1016/j.jclinepi.2016.01.030",
language = "English",
volume = "76",
pages = "183--192",
journal = "Journal of Clinical Epidemiology",
issn = "0895-4356",
publisher = "Elsevier USA",

}

The occurrence of Simpson's paradox if site-level effect was ignored in the TREAT Asia HIV Observational Database. / TREAT Asia HIV Observational Database (TAHOD).

In: Journal of Clinical Epidemiology, Vol. 76, 01.08.2016, p. 183-192.

Research output: Contribution to journalArticle

TY - JOUR

T1 - The occurrence of Simpson's paradox if site-level effect was ignored in the TREAT Asia HIV Observational Database

AU - TREAT Asia HIV Observational Database (TAHOD)

AU - Jiamsakul, Awachana

AU - Kerr, Stephen J.

AU - Chandrasekaran, Ezhilarasi

AU - Huelgas, Aizobelle

AU - Taecharoenkul, Sineenart

AU - Teeraananchai, Sirinya

AU - Wan, Gang

AU - Ly, Penh Sun

AU - Kiertiburanakul, Sasisopin

AU - Law, Matthew

AU - Ly, P. S.

AU - Khol, V.

AU - Zhang, F. J.

AU - Zhao, H. X.

AU - Han, N.

AU - Lee, M. P.

AU - Li, P. C.K.

AU - Lam, W.

AU - Chan, Y. T.

AU - Kumarasamy, N.

AU - Saghayam, S.

AU - Ezhilarasi, C.

AU - Pujari, S.

AU - Joshi, K.

AU - Gaikwad, S.

AU - Chitalikar, A.

AU - Merati, T. P.

AU - Wirawan, D. N.

AU - Yuliana, F.

AU - Yunihastuti, E.

AU - Imran, D.

AU - Widhani, A.

AU - Oka, S.

AU - Tanuma, J.

AU - Nishijima, T.

AU - Choi, J. Y.

AU - Na, S.

AU - Kim, J. M.

AU - Sim, B. L.H.

AU - Gani, Y. M.

AU - David, R.

AU - Kamarulzaman, A.

AU - Syed Omar, S. F.

AU - Ponnampalavanar, S.

AU - Azwa, I.

AU - Mustafa, M.

AU - Nordin, N.

AU - Ditangco, R.

AU - Uy, E.

AU - Bantique, R.

PY - 2016/8/1

Y1 - 2016/8/1

N2 - Objectives In multisite human immunodeficiency virus (HIV) observational cohorts, clustering of observations often occurs within sites. Ignoring clustering may lead to “Simpson's paradox” (SP) where the trend observed in the aggregated data is reversed when the groups are separated. This study aimed to investigate the SP in an Asian HIV cohort and the effects of site-level adjustment through various Cox regression models. Study Design and Setting Survival time from combination antiretroviral therapy (cART) initiation was analyzed using four Cox models: (1) no site adjustment; (2) site as a fixed effect; (3) stratification through site; and (4) shared frailty on site. Results A total of 6,454 patients were included from 23 sites in Asia. SP was evident in the year of cART initiation variable. Model (1) shows the hazard ratio (HR) for years 2010–2014 was higher than the HR for 2006–2009, compared to 2003–2005 (HR = 0.68 vs. 0.61). Models (2)–(4) consistently implied greater improvement in survival for those who initiated in 2010–2014 than 2006–2009 contrasting findings from model (1). The effects of other significant covariates on survival were similar across four models. Conclusions Ignoring site can lead to SP causing reversal of treatment effects. Greater emphasis should be made to include site in survival models when possible.

AB - Objectives In multisite human immunodeficiency virus (HIV) observational cohorts, clustering of observations often occurs within sites. Ignoring clustering may lead to “Simpson's paradox” (SP) where the trend observed in the aggregated data is reversed when the groups are separated. This study aimed to investigate the SP in an Asian HIV cohort and the effects of site-level adjustment through various Cox regression models. Study Design and Setting Survival time from combination antiretroviral therapy (cART) initiation was analyzed using four Cox models: (1) no site adjustment; (2) site as a fixed effect; (3) stratification through site; and (4) shared frailty on site. Results A total of 6,454 patients were included from 23 sites in Asia. SP was evident in the year of cART initiation variable. Model (1) shows the hazard ratio (HR) for years 2010–2014 was higher than the HR for 2006–2009, compared to 2003–2005 (HR = 0.68 vs. 0.61). Models (2)–(4) consistently implied greater improvement in survival for those who initiated in 2010–2014 than 2006–2009 contrasting findings from model (1). The effects of other significant covariates on survival were similar across four models. Conclusions Ignoring site can lead to SP causing reversal of treatment effects. Greater emphasis should be made to include site in survival models when possible.

UR - http://www.scopus.com/inward/record.url?scp=84959216697&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84959216697&partnerID=8YFLogxK

U2 - 10.1016/j.jclinepi.2016.01.030

DO - 10.1016/j.jclinepi.2016.01.030

M3 - Article

C2 - 26854260

AN - SCOPUS:84959216697

VL - 76

SP - 183

EP - 192

JO - Journal of Clinical Epidemiology

JF - Journal of Clinical Epidemiology

SN - 0895-4356

ER -