Desktop version

Home arrow Economics

  • Increase font
  • Decrease font


<<   CONTENTS   >>

Advanced Gastric Cancer: A Meta-Analysis of 20 Trials

Two meta-analyses of randomized clinical trials were conducted by the GASTRIC (Global Advanced/Adjuvant Stomach Tumor Research International Collaboration) Group. The first meta-analysis included individual data on 3,838 patients with curatively resected gastric cancer randomized in 17 trials. This meta-analysis confirmed the benefit of adjuvant chemotherapy as compared with no adjuvant treatment in terms of both disease-free survival (DFS) and overall survival (OS) (GASTRIC Group, 2010). It was also used to show that DFS is a valid surrogate for OS in the adjuvant setting (Oba et al., 2013). The second meta-analysis included individual data on 4,245 patients with advanced or recurrent gastric cancer randomized in 22 trials. This meta-analysis confirmed the benefit of adding experimental agents to standard chemotherapy regimens in terms of both progression-free survival (PFS) and overall survival (OS) (GASTRIC Group, 2013). It was used to show that PFS cannot be used reliably as a surrogate in advanced disease (Paoletti et al., 2013). These two meta-analyses illustrate how a quantitative evaluation can inform the use of potential surrogate endpoints, since two situations that appear relatively similar at face value lead to opposite conclusions as to whether a putative surrogate can be used in practice (Buyse et al., 2016).

In this book we will focus on the situation of advanced gastric cancer. Data are available on 4,069 patients randomized in 20 trials with documented overall survival and progression-free survival (Paoletti et al., 2013). The variables available for analysis include patient identifier (patJd), trial identifier (icLstudy), treatment assigned by randomization (arm), time to progression or death with the first allocated treatment (pfs), progression or death status (pfs_status), time to death or last visit (os), and survival status (last_status). Details on the coding of treatments (1=Control vs. 2=Experimental) are proTABLE 2.7

Advanced Gastric Cancer Trial. Observed treatment effects [95% confidence intervals] on PFS and OS, and predicted treatment effect on OS [95% prediction intervals] in twelve validation trials in advanced gastric cancer (Paoletti et al., 2013). Asterisks mark intervals that exclude no effect.

Trial

Observed HRpps

Observed HRqs

Predicted HRqs

Jeung

0.63

[0.38,

1.05]

0.56

[0.35,

0.88]*

0.73

[0.46,

1.04]

AIO

0.67

[0.43,

1.04]

0.82

[0.47,

1.45]

0.76

[0.53,

1.07]

ToGA

0.71

[0.59,

0.85]*

0.74

[0.60,

0.91]*

0.80

[0.58,

1.09]

AVAGAST

0.80

[0.68,

0.93]*

0.87

[0.73,

1.03]

0.88

[0.76,

1.14]

Kang

0.80

[0.63,

1.03]

0.85

[0.64,

1.13]

0.88

[0.76,

1.14]

Park

0.86

[0.54,

1.37]

0.96

[0.60,

1.52]

0.93

[0.71,

1.18]

REAL(a)

0.92

[0.80,

1.04]

0.92

[0.80,

1.10]

0.98

[0.77,

1.22]

REAL(b)

0.92

[0.81,

1.05]

0.86

[0.80,

0.99]*

0.98

[0.77,

1.22]

Ross

0.95

[0.80,

1.08]

0.91

[0.76,

1.04]

1.00

[0.79,

1.29]

FLAGS

0.99

[0.86,

1.14]

0.92

[0.80,

1.05]

1.03

[0.81,

1.31]

Rao

1.13

[0.63,

2.01]

1.02

[0.61,

1.70]

1.14

[0.89,

1.46]

Moehler

1.14

[0.59,

2.21]

0.77

[0.51,

1.17]

1.15

[0.90,

1.48]

vided in the original meta-analysis (GASTRIC Group, 2013) and surrogate evaluation paper (Paoletti et al., 2013).

One interesting feature of this meta-analysis was that the results of the surrogate evaluation could be externally validated using twelve trials not included in the meta-analysis, using treatment effects extracted from reports published in the literature after the meta-analysis was completed (Table 2.7). The observed treatment effects on survival (HROS), along with their 95% confidence intervals, could be compared with the treatment effects on survival predicted from the treatment effects on the surrogate (HRPFS) in each of these 12 trials. For reasons that are fully detailed in Section 5.3, the prediction intervals were quite wide and included one (no treatment effect on OS) in all 12 trials, which means that the observed effects on PFS would not have allowed predicting an effect on OS in any of these twelve trials. Yet, three of the twelve trials showed a statistically significant effect of treatment on survival (Paoletti et al., 2013), which provided further evidence that treatment effects on PFS were unreliable predictors for treatment effects (or lack thereof) on OS.

In the remainder of this book, this dataset will be referred to as the Gastric dataset. The dataset is included in the R library Surrogate (where it can be accessed using the command data(Gastric)) and it can be downloaded from http://ibiostat.be/online-resources (file gastric.txt). Table 2.8 provides an overview of the variables that are included in the Gastric dataset. The data were provided by the GASTRIC (Global Advanced/Adjuvant StomTABLE 2.8

Advanced Gastric Cancer Trial. Overview of the variables in the dataset.

Variable name

Description

icLstudy

The trial in which the patient was treated

id_pat

The identification number of a patient

arm

The treatment indicator (1=Control, 2=Experimental)

pfs

Progression-free survival in days (the candidate surrogate)

pfs_status

Censoring indicator for progression-free survival (0=alive and progression-free, 1=with progression or dead)

os

Overall survival time in days (the true endpoint)

last .status

Censoring indicator for survival time (0=alive, l=dead)

ach Tumor Research International Collaboration) Group (see Paoletti et al., 2013).

FIGURE 2.7

Advanced Prostate Cancer Trials. Survival curves by treatment arm.

FIGURE 2.8

Advanced Prostate Cancer Trials. Mean PSA profiles by treatment arm.

 
<<   CONTENTS   >>

Related topics