Background: The Tafamidis in Transthyretin Cardiomyopathy Clinical Trial (ATTR-ACT) demonstrated the effectiveness of tafamidis for the treatment of patients with transthyretin amyloid cardiomyopathy (ATTR-CM). Tafamidis reduced mortality in all subgroups of patients studied. Tafamidis also reduced observed frequency of cardiovascular (CV)-related hospitalizations in all subgroups except those who were New York Heart Association (NYHA) class III at baseline who, paradoxically, had a higher frequency of CV-related hospitalizations than placebo. Given the greater mortality rate with placebo, this analysis assessed the impact of the confounding effect of death on the frequency of CV-related hospitalization in ATTR-ACT. Methods: In ATTR-ACT, patients with ATTR-CM were randomized to tafamidis (n = 264) or placebo (n = 177) for 30 months. Post hoc analyses first defined and compared the effect of tafamidis treatment in the subset of NYHA class III patients from each treatment arm alive at month 30. The impact of a potential survivor bias was then adjusted for using principal stratification, estimating the frequency of CV-related hospitalization in NYHA class III patients who would have survived regardless of assigned treatment group (defined as the survivor average causal effect [SACE]). Results: In the subset of NYHA class III patients alive at month 30, tafamidis reduced the relative risk of CV-related hospitalization versus placebo (relative risk: 0.95 [95% CI: 0.55–1.65]). In the principal stratification analyses of those patients who would survive to 30 months regardless of treatment, tafamidis treatment was associated with a 24% lower risk of CV-related hospitalization (relative risk: 0.76 [95% CI: 0.45–1.24]). Similarly, there was a larger reduction in CV-related hospitalization frequency with tafamidis in NYHA class I or II patients in the SACE than was initially observed in ATTR-ACT. Conclusions: Initial data from ATTR-ACT likely underestimated the effect of tafamidis on CV-related hospitalizations due to the confounding effect of death. When SACE was used to adjust for survivor bias, there was a 24% reduction in the frequency of CV-related hospitalization in NYHA class III patients treated with tafamidis.

Transthyretin amyloid cardiomyopathy (ATTR-CM) is an underdiagnosed, fatal disease caused by the deposition of transthyretin amyloid fibrils in the heart leading to heart failure [1]. Tafamidis was shown to be an effective treatment for patients with ATTR-CM in the international, multicenter, double-blind, placebo-controlled, randomized Tafamidis in Transthyretin Cardiomyopathy Clinical Trial (ATTR-ACT) [2, 3]. It has since been approved for the treatment of the cardiomyopathy of wild-type or hereditary transthyretin-mediated amyloidosis in adults to reduce cardiovascular mortality and cardiovascular-related hospitalization and is recommended in heart failure treatment guidelines [4-7].

The primary endpoint of ATTR-ACT was a hierarchical assessment of all-cause mortality followed by frequency of cardiovascular (CV)-related hospitalizations according to the Finkelstein-Schoenfeld method in which tafamidis was shown to be superior to placebo [3]. Treatment with tafamidis resulted in significant reductions in all-cause mortality (a 30% reduction) and in CV-related hospitalizations (relative risk ratio of 0.68) [3]. Tafamidis reduced mortality, compared with placebo, both in patients with less severe (NYHA class I or II) and more severe (NYHA class III) disease at baseline [3, 8].

In contrast, while the observed frequency of CV-related hospitalizations was reduced with tafamidis compared with placebo in the subgroup of patients who were NYHA class I or II at baseline, in the subgroup of patients who were NYHA class III at baseline, there was a greater incidence of CV-related hospitalization with tafamidis compared with placebo, with a relative risk of 1.41 (95% CI: 1.05–1.90) [3, 9]. This was unexpected as there was an observed reduction not only in mortality but also in the decline in functional capacity (assessed by 6-min walk test) and quality of life (assessed by Kansas City Cardiomyopathy Questionnaire Overall Summary score) with tafamidis compared with placebo in this same group of patients.

The evaluation of the causal effect of treatment on CV-related hospitalization is confounded by the occurrence of death post randomization. Death precludes further hospitalization, limiting the evaluation of CV-related hospitalization rates between groups, and introduces selection bias, potentially confounding the results. In ATTR-ACT, this confounding effect of death is particularly relevant in NYHA class III patients who had notably higher mortality rates than NYHA class I or II patients [3, 8]. A recent post hoc analysis from ATTR-ACT examined mortality and CV-related hospitalizations in patients categorized into quartiles by baseline functional capacity (as measured by 6-min walk test) [9]. In that analysis, mortality increased with baseline disease severity (while being lower with tafamidis) in each quartile with both tafamidis and placebo [9]. CV-related hospitalization frequency increased in each quartile with tafamidis; however, with placebo, there was an increase from the first to the third quartile, but a decrease (from 0.92 to 0.73) from the third to the fourth (most severe) quartile (online suppl. Fig. 1; see www.karger.com/doi/10.1159/000525883 for all online suppl. material) [9]. The decrease in CV-related hospitalizations together with greater mortality observed in the placebo group in the fourth quartile suggests that the relatively greater CV-related hospitalization frequency with tafamidis compared with placebo in more severe patients was a consequence of their comparably lower mortality rate [9].

These data suggested that the difference in mortality rate with tafamidis compared with placebo had an impact on the evaluation of CV-related hospitalization frequency in patients with more severe disease. The aim of the additional analyses presented here was to assess the impact of the confounding effect of death on the rate of CV-related hospitalization in ATTR-ACT in NYHA class III patients.

Source Data

Data are from ATTR-ACT, a phase 3, multicenter, international, 3-arm, parallel-design, placebo-controlled, double-blind, randomized study, which has been described previously (NCT01994889) [2, 3]. Briefly, adult patients with ATTR-CM were randomized to tafamidis 80 or 20 mg once daily or matching placebo in a 2:1:2 ratio for 30 months’ treatment. Data from each tafamidis dose were pooled for the primary analysis and for this analysis. Separately, the effect of tafamidis 80 mg alone (the approved dose [6]) was also assessed. Different methods were applied to assess the impact of the confounding effect of death.

ATTR-ACT was approved by the Independent Review Boards or Ethics Committee at each participating site and was conducted according to the provisions of the Declaration of Helsinki and the International Council for Harmonisation (ICH) Good Clinical Practice guidelines. All patients provided written informed consent.

Survival Bias in the Evaluation of CV-Related Hospitalization Frequency

The potential compounding effect of death in the evaluation of CV-related hospitalization in patients with NYHA class III in ATTR-ACT is illustrated in online supplementary Figure 2. The potential for survivor bias can be seen when considering patient C1; as death occurs relatively early in the study, there is no further opportunity for this patient to experience hospitalization. In contrast, a patient living longer will have more opportunity to experience hospitalizations. Given that tafamidis reduced mortality in this population compared with placebo [3], if the effect of death is not accounted for in the analysis, then the frequency of hospitalization can appear higher with tafamidis than with placebo.

CV-Related Hospitalizations in Surviving Patients

In a post hoc analysis, the effect of tafamidis treatment on CV-related hospitalizations was defined and compared in the subset of NYHA class III patients from each treatment arm who were alive at the end of the study (that is, at month 30) using a Poisson regression model for frequency, with treatment and transthyretin genotype as factors, and adjusted for treatment duration. However, the subset of patients who survived on tafamidis would be expected to have different characteristics from the subset of patients who survived on the control arm (placebo), and thus, the analysis on the naïve subset does not estimate the causal effect [10, 11]. To address this potential survivor bias, an additional post hoc principal stratification analysis was conducted. In addition to the analysis in patients who were NYHA class III at baseline, the principal stratification analysis was also conducted in patients who were NYHA class I or II at baseline.

Principal Stratification Analyses and Survivor Average Causal Effect

Principal stratification has become an increasingly popular tool to address certain causal inference questions, particularly in dealing with post randomization factors in randomized trials [12]. Principal stratification is a statistical method to adjust for posttreatment events that may be related to treatment and outcome, in this case mortality, by classifying patients according to their potential to survive under each treatment arm [13, 14]. As principal stratification is not affected by treatment, principal effects are always causal effects and can therefore be used as a pretreatment covariate [15].

The model stratified patients as belonging to 1 of 4 groups, where each patient has 2 sets of potential outcomes – (1) outcome on placebo; and (2) outcome on tafamidis, shown in Table 1 – only 1 of which is observed [13]. These groups are: survival, patients who would survive to the end of planned follow-up, regardless of treatment assignment; death, patients who would die before the end of planned follow-up, regardless of treatment assignment; benefit, patients for whom death before the end of planned follow-up would only occur if the patient were to receive placebo; and detriment, patients for whom death before the end of planned follow-up would only occur if the patient were to receive tafamidis.

Table 1.

Principal stratification and potential outcomes

Principal stratification and potential outcomes
Principal stratification and potential outcomes

The estimand of interest was the frequency of CV-related hospitalization in the principal stratum of “survival,” in which patients would have survived regardless of which treatment group they were assigned. This was defined as the survivor average causal effect (SACE) of the NYHA class III patients. As patients are only assigned to 1 treatment but have 2 sets of potential outcomes (i.e., the strata they are in is not directly observable), assumptions are required to identify the proportion of patients in the “survival” stratum [16, 17]. These assumptions are: the stable unit treatment value assumption, that treating 1 patient does not impact another patient; and the monotonicity assumption, in which it is assumed that patients who did not die while receiving placebo would also not have died if they had received active treatment, and patients who died while receiving active treatment would also have died if they had received placebo. To assess sensitivity to departures from monotonicity, Bayesian inference was utilized to assess the causal estimand under different prior distributions, including a weaker assumption than monotonicity (i.e., weak monotonicity) and without the monotonicity assumption (i.e., no monotonicity) [12, 15, 17]. Statistical details on the Bayesian model for estimating principal causal effect are given in the online supplementary material.

CV-Related Hospitalization Frequency in Patients Who Survived to Month 30

A total of 141 patients in ATTR-ACT were NYHA class III at baseline (78 treated with tafamidis; 63 treated with placebo). In the naïve subset of NYHA class III patients alive at month 30 (35 treated with tafamidis; 24 treated with placebo), the relative risk for CV-related hospitalization with tafamidis compared with placebo was 0.95 (95% CI: 0.55–1.65).

Survivor Average Causal Effect (SACE) Analysis on CV-Related Hospitalization Frequency

Patients who were NYHA class III at baseline were stratified into estimated principal strata proportions, with 33% of patients in the survival strata, 50% in the death strata, and 16% in the benefit strata (Fig. 1); under the monotonicity assumption, there are no detriment patients. The effects of relaxing the monotonicity assumptions on estimated principal strata proportions are shown in Figure 1. As the monotonicity assumption is relaxed, the probability of belonging to the survival stratum remains similar. The probability of belonging to the death stratum is the largest among all strata, which is consistent with what would be expected for the NYHA class III patient population.

Fig. 1.

Estimated principal strata proportions. Strata proportions were estimated using a Bayesian model and sensitivity analysis, under a monotonicity assumption (that patients who did not die on placebo would also not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak and no monotonicity. Survival: patients who would survive to the end of planned follow-up, regardless of treatment assignment. Death: patients who would die before the end of planned follow-up, regardless of treatment assignment. Benefit: patients for whom death before the end of planned follow-up would only occur if the patient were to receive placebo. Detriment: patients for whom death before the end of planned follow-up would only occur if the patient were to receive tafamidis. CI, confidence interval.

Fig. 1.

Estimated principal strata proportions. Strata proportions were estimated using a Bayesian model and sensitivity analysis, under a monotonicity assumption (that patients who did not die on placebo would also not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak and no monotonicity. Survival: patients who would survive to the end of planned follow-up, regardless of treatment assignment. Death: patients who would die before the end of planned follow-up, regardless of treatment assignment. Benefit: patients for whom death before the end of planned follow-up would only occur if the patient were to receive placebo. Detriment: patients for whom death before the end of planned follow-up would only occur if the patient were to receive tafamidis. CI, confidence interval.

Close modal

Under the monotonicity assumption, there was a lower risk of CV-related hospitalization with tafamidis compared with placebo, with a 24% reduction in frequency (relative risk 0.76 [95% CI: 0.45–1.24]) (Fig. 2). With the monotonicity assumption relaxed in the sensitivity analyses, the same trend of lower CV-related hospitalization frequency with tafamidis compared with placebo was observed: a 15% reduction with tafamidis with weak monotonicity and a 14% reduction with tafamidis with no monotonicity (Fig. 2).

Fig. 2.

Estimation of risk ratio in “survival” strata for the NYHA class III patient population. Risk ratios were estimated using Bayesian model-derived strata proportions; the results from the monotonicity assumption (that patients who did not die on placebo would not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak monotonicity and no monotonicity. CI, confidence interval; NYHA, New York Heart Association; SACE, survivor average causal effect.

Fig. 2.

Estimation of risk ratio in “survival” strata for the NYHA class III patient population. Risk ratios were estimated using Bayesian model-derived strata proportions; the results from the monotonicity assumption (that patients who did not die on placebo would not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak monotonicity and no monotonicity. CI, confidence interval; NYHA, New York Heart Association; SACE, survivor average causal effect.

Close modal

As for the analysis in NYHA class III patients, the SACE estimate in all patients and in other subgroups demonstrated a larger reduction in CV-related hospitalization frequency with tafamidis when compared with the initial analysis [3], which did not adjust for survival. In all patients, the reduction in CV-related hospitalization with tafamidis in the SACE estimate (0.39 [0.27–0.56]) (Fig. 3) was larger than the observed reduction in ATTR-ACT (0.68 [0.56–0.81]) [3]. The reduction in the frequency of CV-related hospitalization with tafamidis was larger in NYHA class I/II patients (relative risk 0.36 [95% CI: 0.24–0.50]) than in NYHA class III patients (relative risk: 0.76 [95% CI: 0.45–1.24]) (Fig. 3). There was a similar reduction in the frequency of CV-related hospitalization with tafamidis 80 mg alone as for pooled tafamidis (Fig. 3).

Fig. 3.

SACE estimand on frequency of CV-related hospitalization risk ratio for all patients and by NYHA baseline classification and for pooled tafamidis and tafamidis 80 mg. Risk ratios were estimated using Bayesian model-derived strata proportions, resulting from the monotonicity assumption (that patients who did not die on placebo would not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak monotonicity and no monotonicity. CI, confidence interval; NYHA, New York Heart Association.

Fig. 3.

SACE estimand on frequency of CV-related hospitalization risk ratio for all patients and by NYHA baseline classification and for pooled tafamidis and tafamidis 80 mg. Risk ratios were estimated using Bayesian model-derived strata proportions, resulting from the monotonicity assumption (that patients who did not die on placebo would not have died on tafamidis, and patients who died on tafamidis would have died on placebo). The results for the sensitivity analyses, under which the monotonicity assumption was relaxed, are shown for weak monotonicity and no monotonicity. CI, confidence interval; NYHA, New York Heart Association.

Close modal

In this post hoc analysis, patients in ATTR-ACT who were NYHA class III at baseline and alive at month 30 had a 5% reduced risk of CV-related hospitalization with tafamidis compared with placebo (relative risk 0.95 [95% CI: 0.55–1.65]). This result was in contrast to the observed frequency of CV-related hospitalization in NYHA class III patients in ATTR-ACT, where there was a 41% greater incidence of CV-related hospitalization with tafamidis compared with placebo (relative risk: 1.41 [95% CI: 1.05–1.90]) [3, 9]. When principal stratification was used as a means of adjusting for survival bias, there was a 24% reduction in the frequency of CV-related hospitalization with tafamidis compared with placebo in NYHA class III patients. These results suggest improved outcomes with tafamidis regardless of NYHA class and with both pooled tafamidis and tafamidis 80 mg alone. It is not surprising that the largest difference (showing a greater reduction with treatment) from the observed frequency of hospitalization was seen in NYHA class III patients, as these patients have the highest burden of disease and mortality. This analysis supports the hypothesis that initial data reported from ATTR-ACT [3] likely underestimated the true effect of tafamidis on CV-related hospitalizations due to the confounding effect of death.

The ICH E9(R1) guidelines [18] (which provide guidance on the conduct, analysis, and evaluation of clinical trials) define events occurring post randomization and treatment initiation that potentially affect outcomes measured (such as death) as “intercurrent events” [17]. In situations where the intercurrent event is a consequence of treatment, meaningful estimation of a treatment effect cannot be achieved using randomization alone. Accurate measurement of the impact of treatment on outcomes of interest that can be truncated by death can become challenging, since the patient has died before the outcome could be measured [10]. As such, it was likely that the effect of tafamidis in reducing mortality in patients with ATTR-CM [3] would impact the incidence of CV-related hospitalizations, particularly in patients with more severe ATTR-CM (NYHA class III), since patients with more advanced disease receiving tafamidis were likely to live longer, and therefore have more opportunity to experience CV-related hospitalizations.

As those who survive with tafamidis treatment may differ in important ways from those who survive with placebo, a simple comparison of CV-related hospitalization frequency between the tafamidis and placebo groups may generate estimates of effects of intervention that do not have a causal interpretation. To address the complication that arose from post randomization selection bias, we framed the evaluation of treatment effect with the principal stratification framework and applied SACE to account for the confounding effect of death. Principal stratification is 1 of 5 strategies described in the ICH E9(R1) guidelines for addressing intercurrent events when defining the clinical questions of interest [13]. A crucial aspect of principal stratification is that membership to each principal strata is not affected by treatment assignment [19]. As such, causal effects may be attained when individuals who have the same potential values as the posttreatment variable to be adjusted for under both assigned treatment groups are compared [19]. As a result, unobserved confounders that correlate with treatment, posttreatment variables, and the outcome can be controlled for by focusing on principal strata [19]. Principal stratification handles truncation by death by defining causal contrast to be restricted among the group that would have survived whether they were allocated to the intervention or control group. Principal stratification and analysis of SACE have been proposed as an approach to assess exposure-outcome relationships that are likely to be affected by survival bias [15, 17, 20]. Therefore, this method is the most appropriate for taking mortality into account when examining CV-related hospitalization in ATTR-ACT.

Principal stratification has been used in previous studies, for example, in the EXPAND trial in patients with multiple sclerosis, where the treatment effect was considered in patients who would not relapse when taking placebo or siponimod [12]. The method was also used in an analysis of antidrug antibodies to targeted oncology therapies, where an exploratory reanalysis of data from drug trials by adenosine deaminase (ADA) status (defined as patient ever having been ADA+ during the observation period) was performed [21]. The principal stratification approach has also been utilized in preventative vaccine efficacy trials, where using the standard noncausal estimand comparing the HIV-infected HIV vaccine group with the HIV-infected placebo group could lead to a safe vaccine appearing to harmfully increase viral load [22]. In an example of a more implicit principal stratification, following results suggesting a shorter overall survival duration in patients with gastric cancer with the lowest exposure to trastuzumab, an analysis that matched patients with the lowest quartile trough concentrations of trastuzumab in cycle 1 with patients in the chemotherapy arm was performed [23]. The consequences of analyzing results without accounting for intercurrent events have been illustrated in prostate cancer, where an initial analysis provided generally positive results for finasteride [24]. However, the analysis also indicated that patients randomized to this treatment had a statistically higher risk of high-grade prostate cancer [24]. Subsequent principal stratification sensitivity analyses in conjunction with a long-term follow-up of patients in the trial led to the conclusion that these initial concerns of higher risk of high-grade prostate cancer in patients receiving finasteride “had not been borne out” [25, 26].

It is important to ensure that survival bias is adjusted for in the principal stratification. For example, if the treatment examined reduces mortality, those with more severe disease (e.g., NYHA class III in the present study) will be less likely to die, but more likely to experience the outcome in question (CV-related hospitalization) [16]. As such, the treatment effect will be underestimated in a survivor analysis if it is based on a heterogeneous group of survivors including patients with less, and those with more severe disease at baseline [16]. The applied methodology (SACE analysis) is most useful when the outcome being examined is strongly associated with survival [20]. In a high mortality setting, such as in patients with severe ATTR-CM, a meaningful treatment effect is well defined in the principal stratum of patients who would have survived regardless of treatment [13]. Although every patient has 2 potential outcomes when using the principal stratification, only 1 potential outcome is observed per patient, meaning that this is considered a fixed attribute, allowing estimation of causal effects [12].

Limitations

There are limitations associated with analyzing data using the principal stratum strategy, which includes assumptions that cannot be tested using the observed data. For example, it was assumed that that there were no patients who died receiving active treatment who would have survived if receiving placebo [16, 17]. It is therefore crucial to assess the robustness of results using sensitivity analyses [13, 16, 17]. In this analysis, the same trend was observed when the monotonicity assumption was relaxed in the sensitivity analyses. In addition, the analysis in NYHA class III patients does not include all randomized patients, meaning that it does not conform to the “intention-to-treat” principle for randomized trials [16, 17]. Nevertheless, considering the bias that may be introduced if confounding factors such as survival bias are not taken into account, the approach is useful as a post hoc analysis to determine the effects, if any, of these potentially confounding factors.

The initial data from ATTR-ACT likely underestimated the effect of tafamidis on CV-related hospitalizations (reporting a 41% increase in NYHA class III patients with tafamidis compared with placebo), due to the confounding effect of death. Though ATTR-ACT was not originally designed for subgroup analyses, post hoc evaluation of treatment group among observed survivors estimated a smaller relative risk of CV-related hospitalization for NYHA class III patients; however, that analysis did not fully address the potential survivor bias. Despite some limitations, the principal stratification approach used here demonstrated that, when deaths were accounted for as an intercurrent event using the SACE to adjust for the survivor bias, there was a 24% reduction in the frequency of CV-related hospitalization in NYHA class III patients treated with tafamidis. These findings suggest that there are improved outcomes with tafamidis in patients with ATTR-CM, regardless of NYHA class.

We thank Anqi Yin, Georgetown University, Washington D.C., USA, for assistance with developing the programming codes for this analysis. Medical writing support was provided by Joshua Fink, PhD, of Engage Scientific Solutions, and funded by Pfizer.

ATTR-ACT was approved by the Independent Review Boards or Ethics Committee at each participating site [3]. All patients provided written informed consent. This post hoc study did not require additional local ethical approval in accordance with local/national guidelines.

Huihua Li, Mark Rozenbaum, Michelle Casey, and Marla B. Sultan are employees of Pfizer and hold stock options with Pfizer.

This study was sponsored by Pfizer. The sponsor contributed to the study design, management, and collection of data. In their role as authors, employees of Pfizer were involved in the interpretation of data, preparation, review, and approval of the manuscript, and the decision to submit for publication, along with their coauthors. The study sponsor approved the manuscript from an intellectual property perspective but had no right to veto the publication.

All authors fulfill the ICJME criteria for authorship. All authors contributed to the writing, conception, design, and revision of the manuscript. Huihua Li and Michelle Casey were responsible for statistical analysis. Huihua Li, Mark Rozenbaum, Michelle Casey, and Marla B. Sultan read and approved the final version of this work.

Upon request, and subject to review, Pfizer will provide the data that support the findings of this study. Subject to certain criteria, conditions and exceptions, Pfizer may also provide access to the related individual deidentified participant data. See https://www.pfizer.com/science/clinical-trials/trial-data-and-results for more information.

1.
Ruberg
FL
,
Grogan
M
,
Hanna
M
,
Kelly
JW
,
Maurer
MS
.
Transthyretin amyloid cardiomyopathy: JACC state-of-the-art review
.
J Am Coll Cardiol
.
2019
;
73
(
22
):
2872
91
. .
2.
Maurer
MS
,
Elliott
P
,
Merlini
G
,
Shah
SJ
,
Cruz
MW
,
Flynn
A
,
Design and rationale of the phase 3 ATTR-ACT clinical trial (Tafamidis in Transthyretin Cardiomyopathy Clinical Trial)
.
Circ Heart Fail
.
2017
;
10
(
6
):
e003815
. .
3.
Maurer
MS
,
Schwartz
JH
,
Gundapaneni
B
,
Elliott
PM
,
Merlini
G
,
Waddington-Cruz
M
,
Tafamidis treatment for patients with transthyretin amyloid cardiomyopathy
.
N Engl J Med
.
2018
;
379
(
11
):
1007
16
. .
4.
McDonagh
TA
,
Metra
M
,
Adamo
M
,
Gardner
RS
,
Baumbach
A
,
Böhm
M
,
2021 ESC guidelines for the diagnosis and treatment of acute and chronic heart failure: developed by the task force for the diagnosis and treatment of acute and chronic heart failure of the European Society of Cardiology (ESC) With the special contribution of the Heart Failure Association (HFA) of the ESC
.
Eur Heart J
.
2021
;
42
(
36
):
3599
726
. .
5.
Heidenreich
PA
,
Bozkurt
B
,
Aguilar
D
,
Allen
LA
,
Byun
JJ
,
Colvin
MM
,
AHA/ACC/HFSA guideline for the management of heart failure
.
J Am Coll Cardiol
.
2022
;
0
(
0
).
6.
Pfizer Labs. New York: VYNDAQEL and VYNDAMAX highlights of prescribing information. [cited 2021 Nov 5]. Available from: https://www.fda.gov/media/126283/download.
7.
European Medicines Agency
:
Vyndaqel (tafamidis) summary of product characteristics
.
Available from:
https://www.ema.europa.eu/en/medicines/human/EPAR/vyndaqel.
8.
Rapezzi
C
,
Elliott
P
,
Damy
T
,
Nativi-Nicolau
J
,
Berk
JL
,
Velazquez
EJ
,
Efficacy of tafamidis in patients with hereditary and wild-type transthyretin amyloid cardiomyopathy: further analyses from ATTR-ACT
.
JACC Heart Fail
.
2021
;
9
(
2
):
115
23
. .
9.
Rapezzi
C
,
Kristen
AV
,
Gundapaneni
B
,
Sultan
MB
,
Hanna
M
.
Benefits of tafamidis in patients with advanced transthyretin amyloid cardiomyopathy.
Eur Heart J
.
2020
;
41
(
2
):
ehaa946.2115
. .
10.
Wang
L
,
Zhou
XH
,
Richardson
TS
.
Identification and estimation of causal effects with outcomes truncated by death
.
Biometrika
.
2017
;
104
(
3
):
597
612
. .
11.
Rubin
DB
.
Causal inference through potential outcomes and principal stratification: application to studies with “censoring” due to death
.
Stat Sci
.
2006
;
21
(
3
):
299
309
. .
12.
Magnusson
BP
,
Schmidli
H
,
Rouyrre
N
,
Scharfstein
DO
.
Bayesian inference for a principal stratum estimand to assess the treatment effect in a subgroup characterized by postrandomization event occurrence
.
Stat Med
.
2019
;
38
(
23
):
4761
71
. .
13.
Mallinckrodt
CH
,
Bell
J
,
Liu
G
,
Ratitch
B
,
O'Kelly
M
,
Lipkovich
I
,
Aligning estimators with estimands in clinical trials: putting the ICH E9(R1) guidelines into practice
.
Ther Innov Regul Sci
.
2020
;
54
(
2
):
353
64
. .
14.
Zhang
JL
,
Rubin
DB
.
Estimation of causal effects via principal stratification when some outcomes are truncated by “death”
.
J Educ Behav Stat
.
2003
;
28
(
4
):
353
68
. .
15.
Frangakis
CE
,
Rubin
DB
.
Principal stratification in causal inference
.
Biometrics
.
2002
;
58
(
1
):
21
9
. .
16.
Colantuoni
E
,
Scharfstein
DO
,
Wang
C
,
Hashem
MD
,
Leroux
A
,
Needham
DM
,
Statistical methods to compare functional outcomes in randomized controlled trials with high mortality
.
BMJ
.
2018
;
360
:
j5748
. .
17.
Bornkamp
B
,
Rufibach
K
,
Lin
J
,
Liu
Y
,
Mehrotra
DV
,
Roychoudhury
S
,
Principal stratum strategy: potential role in drug development
.
Pharm Stat
.
2021
;
20
(
4
):
737
51
. .
18.
European Medicines Agency. Amsterdam: ICH E9 statistical principles for clinical trials (R1) Addendum [cited 2021 Nov 5]. Available from: https://www.ema.europa.eu/en/ich-e9-statistical-principles-clinical-trials.
19.
Blanco
G
,
Chen
X
,
Flores
CA
,
Flores-Lagunes
A
.
Bounds on average and quantile treatment effects on duration outcomes under censoring, selection, and noncompliance
.
J Bus Econ Stat
.
2020
;
38
(
4
):
901
20
. .
20.
McGuinness
MB
,
Kasza
J
,
Karahalios
A
,
Guymer
RH
,
Finger
RP
,
Simpson
JA
.
A comparison of methods to estimate the survivor average causal effect in the presence of missing data: a simulation study
.
BMC Med Res Methodol
.
2019
;
19
(
1
):
223
. .
21.
Enrico
D
,
Paci
A
,
Chaput
N
,
Karamouza
E
,
Besse
B
.
Antidrug antibodies against immune checkpoint blockers: impairment of drug efficacy or indication of immune activation?
Clin Cancer Res
.
2020
;
26
(
4
):
787
92
. .
22.
Gilbert
PB
,
Hudgens
MG
,
Wolfson
J
.
Commentary on “principal stratification: a goal or a tool?” by Judea Pearl
.
Int J Biostat
.
2011
;
7
(
1
):
Article 36
. .
23.
Yang
J
,
Zhao
H
,
Garnett
C
,
Rahman
A
,
Gobburu
JV
,
Pierce
W
,
The combination of exposure-response and case-control analyses in regulatory decision making
.
J Clin Pharmacol
.
2013
;
53
(
2
):
160
6
. .
24.
Thompson
IM
,
Goodman
PJ
,
Tangen
CM
,
Lucia
MS
,
Miller
GJ
,
Ford
LG
,
The influence of finasteride on the development of prostate cancer
.
N Engl J Med
.
2003
;
349
(
3
):
215
24
. .
25.
Goodman
PJ
,
Tangen
CM
,
Darke
AK
,
Lucia
MS
,
Ford
LG
,
Minasian
LM
,
Long-term effects of finasteride on prostate cancer mortality
.
N Engl J Med
.
2019
;
380
(
4
):
393
4
. .
26.
Shepherd
BE
,
Redman
MW
,
Ankerst
DP
.
Does finasteride affect the severity of prostate cancer? A causal sensitivity analysis
.
J Am Stat Assoc
.
2008
;
103
(
484
):
1392
404
. .

Additional information

Clinical trial registration: ClinicalTrials.gov: NCT01994889.

Open Access License / Drug Dosage / Disclaimer
This article is licensed under the Creative Commons Attribution 4.0 International License (CC BY). Usage, derivative works and distribution are permitted provided that proper credit is given to the author and the original publisher.Drug Dosage: The authors and the publisher have exerted every effort to ensure that drug selection and dosage set forth in this text are in accord with current recommendations and practice at the time of publication. However, in view of ongoing research, changes in government regulations, and the constant flow of information relating to drug therapy and drug reactions, the reader is urged to check the package insert for each drug for any changes in indications and dosage and for added warnings and precautions. This is particularly important when the recommended agent is a new and/or infrequently employed drug. Disclaimer: The statements, opinions and data contained in this publication are solely those of the individual authors and contributors and not of the publishers and the editor(s). The appearance of advertisements or/and product references in the publication is not a warranty, endorsement, or approval of the products or services advertised or of their effectiveness, quality or safety. The publisher and the editor(s) disclaim responsibility for any injury to persons or property resulting from any ideas, methods, instructions or products referred to in the content or advertisements.