Analysis report (Supporting information)

Effect of oral nutritional supplements with or without nutritional counselling on mortality, treatment tolerance, and quality of life in head and neck cancer patients receiving (chemo)radiotherapy: a systematic review and meta-analysis

1 Abbreviations

CCRT: concurrent chemo-radiotherapy

CI: confidence interval

CT: chemotherapy

MD: mean difference

MID: minimal important difference

NC: nutritional counselling

ONS: oral nutritional supplements

RR: risk ratio

RT: radiotherapy

SD: standard deviation

seTE: standard error

SMD: standardized mean difference

TE: estimated treatment effect

2 Packages

library(meta)

## Loading 'meta' package (version 4.9-7).
## Type 'help(meta)' for a brief overview.

library(readr)
library(rmeta)
library(devtools)

## Loading required package: usethis

library(robvis)
library(patchwork)
library(ggplot2)
library(tidyr)

3 Standard configurations for the meta-analyses

settings.meta(hakn = TRUE) # Hartung-Knapp adjustment

Parsed with column specification:
cols(
  .default = col_double(),
  outclab = col_character(),
  D1 = col_character(),
  D2 = col_character(),
  D3 = col_character(),
  D4 = col_character(),
  D5 = col_character(),
  Overall = col_character(),
  bias = col_character(),
  site = col_character(),
  studlab = col_character(),
  X23 = col_logical()
)

See spec(...) for full column specifications.

4 Comparison 1

4.1 Mortality

4.1.1 Main analysis

Including only results at most at some concerns of bias.

4.1.1.1 Forest plot

mort_1_S <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = FALSE,
  HNC$studlab,
  subset = HNC$outclab == "Mortality" & HNC$C == 1 & HNC$studlab == "Cereda 2018",
  exclude = HNC$studlab == "Jiang 2019",
  hakn = TRUE,
  byvar = HNC$bias
)

mort_1_Sd <- data.frame(mort_1_S)
rob_mort_1_S <- subset(rob_mort_1, studlab == "Cereda 2018")
rob_mort_1_S <- merge(rob_mort_1_S, mort_1_Sd[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_mort_1_S <- rob_mort_1_S[1:3,]

forest(mort_1_S, 
       xlab="Mortality",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       comb.fixed = FALSE,
       xlim = c(0.1, 10),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       pooled.events = TRUE,
       resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.1.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mort_1_S, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.1.2 Sensitivity analysis

Including all available results.

4.1.2.1 Forest plot

mort_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = FALSE,
  HNC$studlab,
  subset = HNC$outclab == "Mortality" & HNC$C == 1,
  exclude = HNC$studlab == "Jiang 2019",
  hakn = TRUE,
  byvar = HNC$bias
)

## Warning in qt(1 - alpha/2, df = df): NaNs produzidos

## Warning in qt(1 - alpha/2, df = df): NaNs produzidos

mort_1d <- data.frame(mort_1)
rob_mort_1 <- merge(rob_mort_1, mort_1d[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_mort_1 <- rob_mort_1[1:3,]

forest(mort_1, 
       xlab="Mortality",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       comb.fixed = FALSE,
       xlim = c(0.01, 100),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       pooled.events = TRUE,
       resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.1.2.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mort_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.1.2.3 Risk of bias assessments by study

rob_traffic_light(rob_mort_1, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(mort_1,
#              pooled = "random"))

4.2 Treatment tolerance

4.2.1 RT Complete suspension

4.2.1.1 Forest plot

rt_susp <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  allstudies = FALSE,
  HNC$studlab,
  subset = HNC$outclab == "RT_Susp" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

## Warning: For a single study, inverse variance method used instead of Mantel-
## Haenszel method.

rt_susp_d <- data.frame(rt_susp)
rob_rt_susp_1 <- merge(rob_rt_susp_1, rt_susp_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_rt_susp_1_x <- subset(rob_rt_susp_1, studlab != "Jiang 2019")

forest(rt_susp, 
       xlab="RT Complete suspension",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.01, 100),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_rt_susp_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.1.3 Risk of bias assessments by study

rob_traffic_light(rob_rt_susp_1_x, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.2 CT Complete suspension

4.2.2.1 Forest plot

ct_susp <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "CT_Susp" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

ct_susp_d <- data.frame(ct_susp)
rob_ct_susp_1 <- merge(rob_ct_susp_1, ct_susp_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(ct_susp, 
       xlab="CT Complete suspension",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.00001, 100000),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.2.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_ct_susp_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.2.3 Risk of bias assessments by study

rob_traffic_light(rob_ct_susp_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.3 RT interruption >= 5 days

4.2.3.1 Forest plot

rt_int5 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Interruption" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

rt_int5_d <- data.frame(rt_int5)
rob_rt_int5_1 <- merge(rob_rt_int5_1, rt_int5_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(rt_int5, 
       xlab="RT Interruption >= 5 days",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.5, 2),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.3.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_rt_int5_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.3.3 Risk of bias assessments by study

rob_traffic_light(rob_rt_int5_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.4 RT interruption

4.2.4.1 Forest plot

rt_int <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "RT_Interruption" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

rt_int_d <- data.frame(rt_int)
rob_rt_int_1 <- merge(rob_rt_int_1, rt_int_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(rt_int, 
       xlab="RT interruption",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       overall = FALSE,
       xlim = c(0.001, 1000),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.4.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_rt_int_1, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

4.2.4.3 Risk of bias assessments by study

rob_traffic_light(rob_rt_int_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.5 Incomplete CCRT

4.2.5.1 Forest plot

ccrt_inc <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "CCRT_Incomplete" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

ccrt_inc_d <- data.frame(ccrt_inc)
rob_ccrt_inc_1 <- merge(rob_ccrt_inc_1, ccrt_inc_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(ccrt_inc, 
       xlab="Incomplete CCRT",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.1, 10),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.5.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_ccrt_inc_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.5.3 Risk of bias assessments by study

rob_traffic_light(rob_ccrt_inc_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.6 RT Dose reduction

4.2.6.1 Forest plot

rt_dose <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "RT_Dose" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

rt_dose_d <- data.frame(rt_dose)
rob_rt_dose_1 <- merge(rob_rt_dose_1, rt_dose_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(rt_dose, 
       xlab="RT Dose reduction",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.01, 100),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.6.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_rt_dose_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.6.3 Risk of bias assessments by study

rob_traffic_light(rob_rt_dose_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.7 CT Dose reduction

4.2.7.1 Forest plot

ct_dose <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = TRUE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "CT_Dose" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

ct_dose_d <- data.frame(ct_dose)
rob_ct_dose_1 <- merge(rob_ct_dose_1, ct_dose_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(ct_dose, 
       xlab="CT Dose reduction",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       xlim = c(0.2, 5),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.2.7.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_ct_dose_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.7.3 Risk of bias assessments by study

rob_traffic_light(rob_ct_dose_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.8 Mucositis (severe)

4.2.8.1 Forest plot

muco_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Mucositis" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE
)

muco_1_d <- data.frame(muco_1)
rob_muco_1 <- merge(rob_muco_1, muco_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(muco_1, 
       xlab="Mucositis (grades 3-4)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.1, 10),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.8.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_muco_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.8.3 Risk of bias assessments by study

rob_traffic_light(rob_muco_1, tool = "ROB2", colour = "colourblind", psize = 10)

#orest(metainf(muco_1,
#               pooled = "random"))

4.2.9 Mucositis (overall)

4.2.9.1 Forest plot

muco_g_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Mucositis_g" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE
)

muco_g_1_d <- data.frame(muco_g_1)
rob_muco_g_1 <- merge(rob_muco_g_1, muco_g_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(muco_g_1, 
       xlab="Mucositis",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.1, 10),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.9.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_muco_g_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.9.3 Risk of bias assessments by study

rob_traffic_light(rob_muco_g_1, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(muco_g_1,
#               pooled = "random"))

4.2.10 Radiation dermatitis

4.2.10.1 Forest plot

derma_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Derma" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

derma_1_d <- data.frame(derma_1)
rob_derma_1 <- merge(rob_derma_1, derma_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(derma_1, 
       xlab="Radiation dermatitis",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.1, 10),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.10.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_derma_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.10.3 Risk of bias assessments by study

rob_traffic_light(rob_derma_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.11 Nausea

4.2.11.1 Forest plot

nausea_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Nausea" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

nausea_1_d <- data.frame(nausea_1)
rob_nausea_1 <- merge(rob_nausea_1, nausea_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(nausea_1, 
       xlab="Nausea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.2, 5),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.11.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_nausea_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.11.3 Risk of bias assessments by study

rob_traffic_light(rob_nausea_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.12 Dry mouth

4.2.12.1 Forest plot

mouth_dry_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Mouth_dry" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE
)

mouth_dry_1_d <- data.frame(mouth_dry_1)
rob_mouth_dry_1 <- merge(rob_mouth_dry_1, mouth_dry_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(mouth_dry_1, 
       xlab="Dry mouth",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.2, 5),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.12.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mouth_dry_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.12.3 Risk of bias assessments by study

rob_traffic_light(rob_mouth_dry_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.13 Swallowing difficulty

4.2.13.1 Forest plot

swallow_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Swallow" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

swallow_1_d <- data.frame(swallow_1)
rob_swallow_1 <- merge(rob_swallow_1, swallow_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(swallow_1, 
       xlab="Swallowing difficulty",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.5, 2),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.13.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_swallow_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.13.3 Risk of bias assessments by study

rob_traffic_light(rob_swallow_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.14 Taste and appetite changes

4.2.14.1 Forest plot

taste_app_1 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Taste_App" & HNC$C == 1,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = TRUE,
)

taste_app_1_d <- data.frame(taste_app_1)
rob_taste_app_1 <- merge(rob_taste_app_1, taste_app_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(taste_app_1, 
       xlab="Taste and appetite changes",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       subgroup = FALSE,
       xlim = c(0.5, 2),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = FALSE,
       overall.hetstat = TRUE
)

4.2.14.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_taste_app_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.2.14.3 Risk of bias assessments by study

rob_traffic_light(rob_taste_app_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.2.15 Summary of treatment tolerance outcomes

4.2.15.1 Results of meta-analysis

tol_1_b <- metabind(#rt_susp,
                    ct_susp,
                    #rt_int5,
                    #ccrt_inc,
                    #rt_dose,
                    #ct_dose,
                    muco_1,
                    muco_g_1,
                    derma_1,
                    #nausea_1,
                    mouth_dry_1,
                    #swallow_1,
                    #taste_app_1,
                    name = c(#"RT suspension", 
                             "CT suspension", 
                             #"RT interruption >= 5 days", 
                             #"Incomplete CCRT", 
                             #"RT dose reduction",
                             #"CT dose reduction",
                             "Mucositis (grades 3-4)",
                             "Mucositis",
                             "Radiation dermatitis",
                             #"Nausea",
                             "Dry mouth"
                             #"Swallowing difficulties",
                             #"Taste and appetite changes"
                             ), 
                    pooled = "random")

forest(tol_1_b, 
       #layout = "meta",
       #leftcols = c("Subgroup"),
       digits = 2,
       digits.se = 1,
       subgroup = FALSE,
       xlim = c(0.01, 100),
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = TRUE,
       overall.hetstat = FALSE,
       digits.addcols.left = 0,
       test.effect.subgroup.random = FALSE,
       test.subgroup = FALSE,
       test.overall = FALSE
)

4.2.15.2 Results of structured reporting (no meta-analysis)

tol_1_c <- metabind(rt_susp,
                    rt_int5,
                    ccrt_inc,
                    rt_dose,
                    ct_dose,
                    nausea_1,
                    swallow_1,
                    taste_app_1,
                    name = c("RT suspension", 
                             "RT interruption >= 5 days",
                             "Incomplete CCRT", 
                             "RT dose reduction",
                             "CT dose reduction",
                             "Nausea",
                             "Swallowing difficulties",
                             "Taste and appetite changes"
                             ), 
                    pooled = "random")

forest(tol_1_c, 
       #layout = "meta",
       #leftcols = c("Subgroup"),
       digits = 2,
       digits.se = 1,
       subgroup = FALSE,
       xlim = c(0.01, 100),
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       hetstat = TRUE,
       overall.hetstat = FALSE,
       digits.addcols.left = 0,
       test.effect.subgroup.random = FALSE,
       test.subgroup = FALSE,
       test.overall = FALSE
)

forest(rt_int, 
       xlab="RT interruption",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       pooled.events = TRUE,
       #overall = FALSE,
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       overall = FALSE,
       xlim = c(0.01, 100),
       #xlim    = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.3 Quality of life (end of treatment)

4.3.1 Global quality of life

4.3.1.1 Forest plot

QoL <- metacont(
  HNC$n.e,
  HNC$mean.e,
  HNC$sd.e,
  HNC$n.c,
  HNC$mean.c,
  HNC$sd.c,
  HNC$studlab,
  data = HNC,
  subset = HNC$outclab == "QoL" & HNC$C == 1, 
  byvar = HNC$bias,
) 

QoL_d <- data.frame(QoL)
rob_QoL_1_1 <- merge(rob_QoL_1_1, QoL_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_QoL_1_1_S <- subset(rob_QoL_1_1,studlab != "Jiang 2019")

forest(QoL, 
       xlab="Quality of life",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-15,15),
       plotwidth = "5cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       lower.equi = -6.5, # Equivalence limits
       upper.equi = 5.4,
       lty.equi = 1,
       fill.equi = "#f5f5f5",
       col.equi = "white",
       lab.e = "Intervention",
       lab.c = "Comparator"
)

MIDs for the global health status (QL) scale for improvement (deterioration) were QL: 5.4 (- 6.5) and SF: 4.9 (- 7.7) in head and neck cancer patients (shaded area)

Musoro J, Coens C, Fiteni F, et al. Evidence-based approach to determine meaningful change in scores of the EORTC QLQ-C30 in breast and head and neck cancer: on behalf of the EORTC Breast, Head and Neck and Quality of Life Groups. 25th annual conference of the international society for quality of life research, Dublin, Ireland. Qual Life Res 2018;27 (Suppl 1): ab101.4, 18. https://doi.org/10.1007/s11136-018-1946-9

4.3.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_QoL_1_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.3.1.3 Risk of bias assessments by study

rob_traffic_light(rob_QoL_1_1, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(QoL,
#               pooled = "random"))

4.3.2 Sensitivity analysis

Excluding Jiang 2019, because mean and standard deviation for this study were inputted (as described in the methods section of the primary report).

4.3.2.1 Forest plot

QoL_S <- update.meta(QoL, subset = HNC$outclab == "QoL" & HNC$C == 1 & HNC$studlab != "Jiang 2019")

forest(QoL_S, 
       xlab="Quality of life",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-15,15),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       lower.equi = -6.5, # Equivalence limits
       upper.equi = 5.4,
       lty.equi = 1,
       fill.equi = "#f5f5f5",
       col.equi = "white",
)

4.3.2.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_QoL_1_1_S, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.3.2.3 Risk of bias assessments by study

rob_traffic_light(rob_QoL_1_1_S, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(bw_1_1,
#               pooled = "random"))

4.3.3 Quality of life subscales

4.3.3.1 Appetite loss

QoL_Appetite <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Appetite" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Appetite, 
       xlab="QoL_Appetite",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-2, 2),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE,
)

4.3.3.2 Cognitive

QoL_Cognitive <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Cognitive" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Cognitive, 
       xlab="QoL_Cognitive",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-3, 3),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.3 Constipation

QoL_Constipation <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Constipation" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Constipation, 
       xlab="QoL_Constipation",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm",
       xlim = c(-1, 1),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.3.3.4 Diarrhoea

QoL_Diarrhoea <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Diarrhoea" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Diarrhoea, 
       xlab="QoL_Diarrhoea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm",
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.5 Dyspnoea

QoL_Dyspnoea <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Dyspnoea" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Dyspnoea, 
       xlab="QoL_Dyspnoea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.6 Emotional

QoL_Emotional <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Emotional" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Emotional, 
       xlab="QoL_Emotional",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-5, 5),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

4.3.3.7 Fatigue

QoL_Fatigue <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Fatigue" & HNC$C == 1, 

  byvar = HNC$bias
) 

forest(QoL_Fatigue, 
       xlab="QoL_Fatigue",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-15, 15),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       lower.equi = 0, # Equivalence limits
       upper.equi = 12,
       lty.equi = 1,
       fill.equi = "#f5f5f5",
       col.equi = "white",
       overall = FALSE
)

MID: 12.0 (deteriorate)

Musoro J, Coens C, Fiteni F, et al. Minimally important differences for interpreting EORTC QLQ-C30 scores in melanoma, breast cancer and head and neck cancer patients on behalf of the EORTC breast, Head and Neck, Melanoma and Quality of life groups. ISPOR Europe 2018 Barcelona, November, 2018.

4.3.3.8 Financial

QoL_Financial <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Financial" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Financial, 
       xlab="QoL_Financial",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.3.3.9 Insomnia

QoL_Insomnia <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Insomnia" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Insomnia, 
       xlab="QoL_Insomnia",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-3, 3),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.10 Nausea

QoL_Nausea <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Nausea" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Nausea, 
       xlab="QoL_Nausea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.11 Pain

QoL_Pain <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Pain" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Pain, 
       xlab="QoL_Pain",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.12 Physical

QoL_Physical <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Physical" & HNC$C == 1, 

  byvar = HNC$bias,
) 

forest(QoL_Physical, 
       xlab="QoL_Physical",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-15, 15),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       lower.equi = -7.3, # Equivalence limits
       upper.equi = 0,
       lty.equi = 1,
       fill.equi = "#f5f5f5",
       col.equi = "white",
       overall = FALSE
)

MID: -7.3 (deteriorate)

4.3.3.13 Role

QoL_Role <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Role" & HNC$C == 1, 

  byvar = HNC$bias,
  sm = "SMD",
) 

forest(QoL_Role, 
       xlab="QoL_Role",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2, 2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = FALSE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       overall = FALSE
)

4.3.3.14 Social

QoL_Social <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Social" & HNC$C == 1, 

  byvar = HNC$bias
) 

forest(QoL_Social, 
       xlab="QoL_Social",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-30, 30),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       lower.equi = -7.3, # Equivalence limits
       upper.equi = 6.1,
       lty.equi = 1,
       fill.equi = "#f5f5f5",
       col.equi = "white",
)

MIDs for the social functioning (SF) scale for improvement (deterioration) were SF: 6.1 (- 7.3) in HNC

Musoro J, Coens C, Fiteni F, et al. Evidence-based approach to determine meaningful change in scores of the EORTC QLQ-C30 in breast and head and neck cancer: on behalf of the EORTC Breast, Head and Neck and Quality of Life Groups. 25th annual conference of the international society for quality of life research, Dublin, Ireland. Qual Life Res 2018;27 (Suppl 1): ab101.4, 18. https://doi.org/10.1007/s11136-018-1946-9

4.4 Functional status (end of treatment)

4.4.1 Forest plot

functional <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "Functional status" & HNC$C == 1, 

  byvar = HNC$bias
) 

functional_d <- data.frame(functional)
rob_functional_1_1 <- merge(rob_functional_1_1, functional_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(functional, 
       xlab="Functional status",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-10,10),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

4.4.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_functional_1_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.4.3 Risk of bias assessments by study

rob_traffic_light(rob_functional_1_1, tool = "ROB2", colour = "colourblind", psize = 10)

4.5 Body weight (end of treatment)

4.5.1 Main analysis

Including only results at most at some concerns of bias.

4.5.1.1 Forest plot

bw_1_1 <- metagen(
  HNC$TE,
  HNC$seTE,
  sm = "MD",
  HNC$studlab,
  data = HNC,
  subset = HNC$outclab == "Body weight" & HNC$C == 1,
  byvar = HNC$bias,
)

bw_1_1_d <- data.frame(bw_1_1)
rob_bw_1_1 <- merge(rob_bw_1_1, bw_1_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_bw_1_1_S <- subset(rob_bw_1_1, Overall != "High")
rob_bw_1_1_S_2 <- subset(rob_bw_1_1,studlab != "Chitapanarux 2016")

bw_1_1_LB <- update.meta(bw_1_1, subset = HNC$outclab == "Body weight" & HNC$C == 1 & HNC$bias != "High risk of bias")

forest(bw_1_1_LB, 
       xlab="Body weight (end of treatment)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-5, 5),
       #leftcols = c("studlab", "n.e", "n.c", "TE", "seTE", "random.w"),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

4.5.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_bw_1_1_S, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.5.2 Sensitivity analysis 1

Including all available results.

4.5.2.1 Forest plot

forest(bw_1_1, 
       xlab="Body weight (end of treatment)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-10, 10),
       leftcols = c("studlab", "n.e", "n.c", "TE", "seTE"),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

4.5.2.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_bw_1_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.5.2.3 Risk of bias assessments by study

rob_traffic_light(rob_bw_1_1, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(bw_1_1,
#               pooled = "random"))

4.5.3 Sensitivity analysis 2

Excluding Chitapanarux 2016, because mean and standard deviation for this study were inputted (as described in the methods section of the primary report).

4.5.3.1 Forest plot

bw_1_1_S_2 <- update.meta(bw_1_1, subset = HNC$outclab == "Body weight" & HNC$C == 1 & HNC$studlab != "Chitapanarux 2016")

forest(bw_1_1_S_2, 
       xlab="Body weight (end of treatment)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-10, 10),
       leftcols = c("studlab", "n.e", "n.c", "TE", "seTE"),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

4.5.3.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_bw_1_1_S_2, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.5.3.3 Risk of bias assessments by study

rob_traffic_light(rob_bw_1_1_S_2, tool = "ROB2", colour = "colourblind", psize = 10)

#forest(metainf(bw_1_1,
#               pooled = "random"))

4.6 Adverse effects

4.6.1 Forest plot

ae <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = FALSE,
  HNC$studlab,
  subset = HNC$outclab == "Adverse events" & HNC$C == 1,
  byvar = HNC$bias,
)

ae_d <- data.frame(ae)
rob_ae_1 <- merge(rob_ae_1, ae_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(ae, 
       xlab="Adverse effects (nausea, vomiting, feeling of fullness)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(0.001, 1000),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

4.6.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_ae_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

4.6.3 Risk of bias assessments by study

rob_traffic_light(rob_ae_1, tool = "ROB2", colour = "colourblind", psize = 10)

5 Comparison 2

5.1 Mortality

5.1.1 Forest plot

mort_2 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = FALSE,
  HNC$studlab,
  subset = HNC$outclab == "Mortality" & HNC$C == 2,
  byvar = HNC$bias
)

mort_2d <- data.frame(mort_2)
rob_mort_2 <- merge(rob_mort_2, mort_2d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(mort_2, 
       xlab="Mortality",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(0.5, 2),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mort_2, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

5.1.3 Risk of bias assessments by study

rob_traffic_light(rob_mort_2, tool = "ROB2", colour = "colourblind", psize = 10)

5.2 Quality of life (end of treatment)

5.2.1 Global quality of life

5.2.1.1 Forest plot

QoL_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

QoL_2_d <- data.frame(QoL_2)
rob_QoL_2_1 <- merge(rob_QoL_2_1, QoL_2_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(QoL_2, 
       xlab="Quality of life",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       plotwidth = "5cm",
       xlim = c(-1, 1),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10,
       lab.e = "Intervention",
       lab.c = "Comparator",
)

5.2.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_QoL_2_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

5.2.1.3 Risk of bias assessments by study

rob_traffic_light(rob_QoL_2_1, tool = "ROB2", colour = "colourblind", psize = 10)

5.2.2 Quality of life subscales

5.2.2.1 Appetite loss

QoL_Appetite_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Appetite" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

forest(QoL_Appetite_2, 
       xlab="QoL_Appetite",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-1,1),
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.2.2.2 Constipation

QoL_Constipation_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Constipation" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

forest(QoL_Constipation_2, 
       xlab="QoL_Constipation",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-1,1),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.2.2.3 Diarrhoea

QoL_Diarrhoea_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Diarrhoea" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

forest(QoL_Diarrhoea_2, 
       xlab="QoL_Diarrhoea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-1,1),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.2.2.4 Nausea

QoL_Nausea_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Nausea" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

forest(QoL_Nausea_2, 
       xlab="QoL_Nausea",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2,2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.2.2.5 Pain

QoL_Pain_2 <- metacont(
  HNC$n.e,
  
  HNC$mean.e,
  
  HNC$sd.e,

  HNC$n.c,
 
  HNC$mean.c,

  HNC$sd.c,
 
  HNC$studlab,

  data = HNC,

  subset = HNC$outclab == "QoL_Pain" & HNC$C == 2, 

  sm = "SMD",
  byvar = HNC$bias
) 

forest(QoL_Pain_2, 
       xlab="QoL_Pain",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 1,
       digits.se = 1,
       comb.fixed = FALSE,
       #plotwidth = "2cm"
       xlim = c(-2,2),
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

5.3 Body weight (end of treatment)

5.3.1 Forest plot

bw_2_1 <- metacont(
  HNC$n.e,
  HNC$mean.e,
  HNC$sd.e,
  HNC$n.c,
  HNC$mean.c,
  HNC$sd.c,
  HNC$studlab,
  data = HNC,
  subset = HNC$outclab == "Body weight" & HNC$C == 2,
  byvar = HNC$bias,
) 

bw_2_1_d <- data.frame(bw_2_1)
rob_bw_2_1 <- merge(rob_bw_2_1, bw_2_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)
rob_bw_2_1_x <- subset(rob_bw_2_1, studlab != "Moriarty 1989")

forest(bw_2_1, 
       xlab="Body weight (end of treatment)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-15, 15),
       #leftcols = c("studlab", "n.e", "n.c", "TE", "seTE", "random.w"),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

Moriarty 1981 measured body weight but only reported the result as statistically non-significant, so it could not be included in a meta-analysis.

5.3.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_bw_2_1, tool = "ROB2", weighted = TRUE, overall = TRUE, colour = "colourblind")

5.3.3 Risk of bias assessments by study

rob_traffic_light(rob_bw_2_1_x, tool = "ROB2", colour = "colourblind", psize = 10)

6 Comparison 3

6.1 Mortality

6.1.1 Forest plot

mort_3 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Mortality" & HNC$C == 3,
  byvar = HNC$bias,
  comb.fixed = FALSE,
)

mort_3_d <- data.frame(mort_3)
rob_mort_3 <- merge(rob_mort_3, mort_3_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(mort_3, 
       xlab="Treatment interruption",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       #subgroup = FALSE,
       xlim = c(0.2, 5),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

6.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mort_3, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

6.1.3 Risk of bias assessments by study

rob_traffic_light(rob_mort_3, tool = "ROB2", colour = "colourblind", psize = 10)

6.2 Interruption of treatment

6.2.1 Forest plot

int_3 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Interruption of anti-cancer treatment" & HNC$C == 3,
  byvar = HNC$bias,
  comb.fixed = FALSE,
)

int_3_d <- data.frame(int_3)
rob_int_3 <- merge(rob_int_3, int_3_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(int_3, 
       xlab="Treatment interruption",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       #subgroup = FALSE,
       xlim = c(0.001, 1000),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

6.2.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_int_3, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

6.2.3 Risk of bias assessments by study

rob_traffic_light(rob_int_3, tool = "ROB2", colour = "colourblind", psize = 10)

6.3 Summary of non-hematological toxicity outcomes

6.3.1 Forest plot

tol_3 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Tolerance" & HNC$C == 3,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = FALSE,
)

forest(tol_3, 
       xlab="Non-hematological toxicity",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       #subgroup = FALSE,
       xlim = c(0.1, 10),
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

6.3.2 Proportion of the summary at each level of risk of bias

rob_summary(rob_tol_3, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

6.3.3 Risk of bias assessments by study

rob_traffic_light(rob_tol_3, tool = "ROB2", colour = "colourblind", psize = 10)

#labbe.metabin(x = tol_1,
#              bg = "grey",
#              studlab = TRUE,
#              comb.random = FALSE,
#              cex.studlab = 0.4)

6.4 Body weight (end of treatment)

6.4.1 Forest plot

bw_3_1 <- metacont(
  HNC$n.e,
  HNC$mean.e,
  HNC$sd.e,
  HNC$n.c,
  HNC$mean.c,
  HNC$sd.c,
  HNC$studlab,
  data = HNC,
  subset = HNC$outclab == "Body weight" & HNC$C == 3,
  byvar = HNC$bias,
) 

bw_3_1_d <- data.frame(bw_3_1)
rob_bw_3_1 <- merge(rob_bw_3_1, bw_3_1_d[,c("w.random", "studlab")], by = "studlab", all = TRUE)

forest(bw_3_1, 
       xlab="Body weight (end of treatment)",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       xlim = c(-10, 10),
       #leftcols = c("studlab", "n.e", "n.c", "TE", "seTE", "random.w"),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = I1,
       label.left = C1,
       fs.axis = 10,
       fs.lr = 10
)

6.4.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_bw_3_1, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

6.4.3 Risk of bias assessments by study

rob_traffic_light(rob_bw_3_1, tool = "ROB2", colour = "colourblind", psize = 10)

7 Comparison 4

7.1 Mortality

7.1.1 Forest plot

mort_4 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Mortality" & HNC$C == 4,
  byvar = HNC$bias,
  comb.fixed = FALSE,
)

forest(mort_4, 
       xlab="Mortality",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       #subgroup = FALSE,
       xlim = c(0.1, 10),
       #xlim    = "s",
       #plotwidth = "2cm",
       subgroup = FALSE,
       hetstat = FALSE,
       overall.hetstat = TRUE,
       #pooled.events = TRUE,
       #resid.hetstat = FALSE,
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

7.1.2 Proportion of information at each level of risk of bias

Risk of bias assessments are weighted by the relative contribution of each study to the meta-analysis result (when applicable)

rob_summary(rob_mort_4, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

7.1.3 Risk of bias assessments by study

rob_traffic_light(rob_mort_4, tool = "ROB2", colour = "colourblind", psize = 8)

7.2 Summary of non-hematological toxicity outcomes

7.2.1 Forest plot

tol_4 <- metabin(
  HNC$event.e,
  HNC$n.e,
  HNC$event.c,
  HNC$n.c,
  sm = "RR",
  method = "MH",
  RR.cochrane = TRUE,
  MH.exact = FALSE,
  #allstudies = TRUE,
  HNC$studlab,
  subset = HNC$outclab == "Tolerance" & HNC$C == 4,
  byvar = HNC$bias,
  comb.fixed = FALSE,
  comb.random = FALSE,
)

forest(tol_4, 
       xlab="Non-hematological toxicity",
       #ref = 10,
       #layout = "Revman5",
       layout = "meta",
       digits = 2,
       digits.se = 1,
       comb.fixed = FALSE,
       #overall = FALSE,
       #subgroup = FALSE,
       xlim = c(0.1, 10),
       #lim = "s",
       #plotwidth = "2cm",
       colgap.forest = "0.5 cm",
       just = "right",
       label.right = C1,
       label.left = I1,
       fs.axis = 10,
       fs.lr = 10
)

7.2.2 Proportion of the summary at each level of risk of bias

rob_summary(rob_tol_4, tool = "ROB2", weighted = FALSE, overall = TRUE, colour = "colourblind")

7.2.3 Risk of bias assessments by study

rob_traffic_light(rob_tol_4, tool = "ROB2", colour = "colourblind", psize = 10)