Comparing randomized trial designs to estimate treatment effect in rare diseases with longitudinal models: a simulation study showcased by Autosomal Recessive Cerebellar Ataxias using the SARA score - code

This repository contains the scripts for the paper submitted to the BMC Medical Research Methodology Journal : Comparing randomized trial designs to estimate treatment effect in rare diseases with longitudinal models: a simulation study showcased by Autosomal Recessive Cerebellar Ataxias using the SARA score

Authors : Hendrickx N, Mentré F, Hamdan A, Karlsson M O, Hooker A C, Traschütz A, Gagnon C, Schüle R, ARCA Study Group, EVIDENCE-RND consortium, Synofzik M, Comets E.

This repository contains the NONMEM code to reproduce the simulation study of the manuscript for the slow progression case. It contains three folders, one for each design : parallel, crossover and delayed start. 

Each folder contains one .csv file describing the design of the simulation, 8 NONMEM model files and 2 .sh files for the simulation. There are three models : Logistic, Linear and Linear Sparse models, twice each (*_full.mod = drug effect estimated, *_red.mod = drug effect fixed to 0). The Logistic model is also present twice : *H0_* = with a simulated drug effect of 0; *H1_* = with a simulated drug effect of 0.5. There are two .sh files : *_H0.sh = simulated drug effect of 0, for type 1 error ;  *_H1.sh = simulated drug effect of 0.5, for power.

To get the type 1 error and power calculation, you will need to unzip the "m1.zip" folders present in "ARSACS_type1" and "ARSACS_power" to "m1_*(t1/power)*" in each design directory and run the "script_type1_power.R" script in the parent directory to get the type 1 error and corrected power for all designs, as well as the code to get the figures 4 and 5 from the manuscript.

Please note that the computation time is long, especially if the commands are run locally. The commands in the .sh files are setup with 8 threads, and, depending on your CPU, the simulation study could take from 2 to 5 days per design and per simulation (Under H0 and H1) (with 500 replicates). Also, if you were to rerun the simulation, you should remove the "ARSACS_type_1" and "ARSACS_power" folders, or rename the target folder in the sse command and R script.

Softwares:

NONMEM 7.5.0PsN 5.3.1R 4.1.2

本仓库包含提交至《BMC Medical Research Methodology》期刊的论文配套脚本：比较基于纵向模型评估罕见病治疗效应的随机试验设计——以结合SARA评分的常染色体隐性小脑共济失调为例开展的模拟研究。

作者：Hendrickx N、Mentré F、Hamdan A、Karlsson M O、Hooker A C、Traschütz A、Gagnon C、Schüle R、ARCA研究组（ARCA Study Group）、EVIDENCE-RND联盟（EVIDENCE-RND consortium）、Synofzik M、Comets E。

本仓库包含用于复现论文慢进展病例模拟研究的NONMEM代码，内含三个文件夹，分别对应三类试验设计：平行组设计、交叉设计与延迟启动设计。

每个文件夹内包含1个描述模拟设计的.csv文件、8个NONMEM模型文件与2个用于执行模拟的.sh脚本文件。本研究涉及三类模型：Logistic模型、线性模型与稀疏线性模型，均设有两个子版本：*_full.mod 代表药物效应待估版本，*_red.mod 代表药物效应固定为0的版本。Logistic模型另有两组区分：*H0_* 对应模拟药物效应为0的场景，*H1_* 对应模拟药物效应为0.5的场景。脚本文件共两类：*_H0.sh 用于药物效应为0的场景，以计算一类错误率；*_H1.sh 用于药物效应为0.5的场景，以计算检验效能。

若需计算一类错误率与校正后的检验效能，请将各设计目录下“ARSACS_type1”与“ARSACS_power”文件夹中的“m1.zip”解压至对应设计目录下的“m1_*(t1/power)*”路径，随后运行父目录下的“script_type1_power.R”脚本，即可获取所有设计的一类错误率与校正后的检验效能，同时可复现论文中图4与图5的代码。

请注意，该计算耗时较长，若在本地运行则耗时更为可观。.sh脚本默认配置为8线程，根据您的CPU性能，每类设计的单次模拟（包含H0与H1两种场景，共500次重复）耗时约2至5天。若需重新运行模拟，请删除“ARSACS_type_1”与“ARSACS_power”文件夹，或在sse命令与R脚本中修改目标文件夹名称。

所需软件：NONMEM 7.5.0、PsN 5.3.1、R 4.1.2