Limits...
The Impact of Different CD4 Cell-Count Monitoring and Switching Strategies on Mortality in HIV-Infected African Adults on Antiretroviral Therapy: An Application of Dynamic Marginal Structural Models.

Ford D, Robins JM, Petersen ML, Gibb DM, Gilks CF, Mugyenyi P, Grosskurth H, Hakim J, Katabira E, Babiker AG, Walker AS, DART Trial Te - Am. J. Epidemiol. (2015)

Bottom Line: Assumptions included no direct effect of randomized group on mortality or confounders and no unmeasured confounders which influenced treatment switch and mortality or treatment switch and time-dependent covariates.After 48 weeks of first-line ART, 2,946 individuals contributed 11,351 person-years of follow-up, 625 switches, and 179 deaths.These findings support a benefit from identifying patients immunologically failing first-line ART at 48 weeks.

View Article: PubMed Central - PubMed

Show MeSH

Related in: MedlinePlus

Directced acyclic graph illustrating associations between randomized trial group (R), time-dependent covariates at time t (Ct, e.g., CD4 cell count), switch to second-line antiretroviral therapy (ART) before/at time t (Et), death before/at time t (Dt), and unmeasured common causes of C and D (U) among human immunodeficiency virus–positive patients on ART. Arrows represent direct causal relationships between variables. Time-dependent covariates (C) at a given time point influence whether treatment is switched to second-line ART at that time point or subsequently (E) and influence time-dependent covariates (C) at later time points and mortality (D). Switching treatment regimens (E) influences time-dependent covariates (C), switching (E), and mortality at later time points (D). The following assumptions are made: R has no effect on C other than via E; R has no effect on D other than via E; there are no unmeasured common causes of E and C or E and D; and R is randomized. Different line styles and colors are used only to distinguish the effects of randomized group, different covariates, exposures, and death: Effects of R are shown by solid black lines; effects of C0 by dashed gray lines; effects of E0 by dashed black lines; effects of D1 by solid gray lines; effects of C1 by dotted black lines; effects of E1 by dashed-dotted gray lines; and effects of U by dotted gray lines.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4581589&req=5

KWV083F1: Directced acyclic graph illustrating associations between randomized trial group (R), time-dependent covariates at time t (Ct, e.g., CD4 cell count), switch to second-line antiretroviral therapy (ART) before/at time t (Et), death before/at time t (Dt), and unmeasured common causes of C and D (U) among human immunodeficiency virus–positive patients on ART. Arrows represent direct causal relationships between variables. Time-dependent covariates (C) at a given time point influence whether treatment is switched to second-line ART at that time point or subsequently (E) and influence time-dependent covariates (C) at later time points and mortality (D). Switching treatment regimens (E) influences time-dependent covariates (C), switching (E), and mortality at later time points (D). The following assumptions are made: R has no effect on C other than via E; R has no effect on D other than via E; there are no unmeasured common causes of E and C or E and D; and R is randomized. Different line styles and colors are used only to distinguish the effects of randomized group, different covariates, exposures, and death: Effects of R are shown by solid black lines; effects of C0 by dashed gray lines; effects of E0 by dashed black lines; effects of D1 by solid gray lines; effects of C1 by dotted black lines; effects of E1 by dashed-dotted gray lines; and effects of U by dotted gray lines.

Mentions: Unless otherwise stated, we pooled participants from both randomized groups (LCM/CDM) under the assumptions that 1) participants were comparable at baseline (48 weeks on first-line ART); 2) there was no direct effect of randomized group on mortality or any confounders—that is, any effect on mortality or time-dependent covariates (e.g., CD4 count, WHO 4 events) of access to CD4 test results in the LCM group or lack of access to results in the CDM group was indirect and occured through switching; and 3) there were no unmeasured confounders which influenced treatment switch and mortality or treatment switch and time-dependent covariates (Figure 1). The DART Trial found no difference between randomized groups in any toxicity outcome (2), suggesting a lack of impact of other (non-CD4) laboratory tests. Switching and outcome models were fitted without including randomized group. The rationale for pooling groups and related assumptions are discussed further in Web Appendix 1 (which includes Web Figure 1 and Web Table 1), available at http://aje.oxfordjournals.org/.Figure 1.


The Impact of Different CD4 Cell-Count Monitoring and Switching Strategies on Mortality in HIV-Infected African Adults on Antiretroviral Therapy: An Application of Dynamic Marginal Structural Models.

Ford D, Robins JM, Petersen ML, Gibb DM, Gilks CF, Mugyenyi P, Grosskurth H, Hakim J, Katabira E, Babiker AG, Walker AS, DART Trial Te - Am. J. Epidemiol. (2015)

Directced acyclic graph illustrating associations between randomized trial group (R), time-dependent covariates at time t (Ct, e.g., CD4 cell count), switch to second-line antiretroviral therapy (ART) before/at time t (Et), death before/at time t (Dt), and unmeasured common causes of C and D (U) among human immunodeficiency virus–positive patients on ART. Arrows represent direct causal relationships between variables. Time-dependent covariates (C) at a given time point influence whether treatment is switched to second-line ART at that time point or subsequently (E) and influence time-dependent covariates (C) at later time points and mortality (D). Switching treatment regimens (E) influences time-dependent covariates (C), switching (E), and mortality at later time points (D). The following assumptions are made: R has no effect on C other than via E; R has no effect on D other than via E; there are no unmeasured common causes of E and C or E and D; and R is randomized. Different line styles and colors are used only to distinguish the effects of randomized group, different covariates, exposures, and death: Effects of R are shown by solid black lines; effects of C0 by dashed gray lines; effects of E0 by dashed black lines; effects of D1 by solid gray lines; effects of C1 by dotted black lines; effects of E1 by dashed-dotted gray lines; and effects of U by dotted gray lines.
© Copyright Policy - creative-commons
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4581589&req=5

KWV083F1: Directced acyclic graph illustrating associations between randomized trial group (R), time-dependent covariates at time t (Ct, e.g., CD4 cell count), switch to second-line antiretroviral therapy (ART) before/at time t (Et), death before/at time t (Dt), and unmeasured common causes of C and D (U) among human immunodeficiency virus–positive patients on ART. Arrows represent direct causal relationships between variables. Time-dependent covariates (C) at a given time point influence whether treatment is switched to second-line ART at that time point or subsequently (E) and influence time-dependent covariates (C) at later time points and mortality (D). Switching treatment regimens (E) influences time-dependent covariates (C), switching (E), and mortality at later time points (D). The following assumptions are made: R has no effect on C other than via E; R has no effect on D other than via E; there are no unmeasured common causes of E and C or E and D; and R is randomized. Different line styles and colors are used only to distinguish the effects of randomized group, different covariates, exposures, and death: Effects of R are shown by solid black lines; effects of C0 by dashed gray lines; effects of E0 by dashed black lines; effects of D1 by solid gray lines; effects of C1 by dotted black lines; effects of E1 by dashed-dotted gray lines; and effects of U by dotted gray lines.
Mentions: Unless otherwise stated, we pooled participants from both randomized groups (LCM/CDM) under the assumptions that 1) participants were comparable at baseline (48 weeks on first-line ART); 2) there was no direct effect of randomized group on mortality or any confounders—that is, any effect on mortality or time-dependent covariates (e.g., CD4 count, WHO 4 events) of access to CD4 test results in the LCM group or lack of access to results in the CDM group was indirect and occured through switching; and 3) there were no unmeasured confounders which influenced treatment switch and mortality or treatment switch and time-dependent covariates (Figure 1). The DART Trial found no difference between randomized groups in any toxicity outcome (2), suggesting a lack of impact of other (non-CD4) laboratory tests. Switching and outcome models were fitted without including randomized group. The rationale for pooling groups and related assumptions are discussed further in Web Appendix 1 (which includes Web Figure 1 and Web Table 1), available at http://aje.oxfordjournals.org/.Figure 1.

Bottom Line: Assumptions included no direct effect of randomized group on mortality or confounders and no unmeasured confounders which influenced treatment switch and mortality or treatment switch and time-dependent covariates.After 48 weeks of first-line ART, 2,946 individuals contributed 11,351 person-years of follow-up, 625 switches, and 179 deaths.These findings support a benefit from identifying patients immunologically failing first-line ART at 48 weeks.

View Article: PubMed Central - PubMed

Show MeSH
Related in: MedlinePlus