Neuropathic pain

Comparison and analysis of the treatment of fibromyalgia with antidepressants

Current draft for discussion with Tamar, Gav, and Hollie. Will cut down for presentation to whole group.

packages

Show code 
library(meta)
library(gt)
library(tidyverse)
library(skimr)
library(metafor)
library(targets)
library(multinma)

source("R/hpp_themes.R")

replication of amitriptyline and fibromyalgia

In this section I replicate the substantial pain reduction findings of Moore et al. (Moore et al. 2015).

notes

Objective: measure efficacy of analgesic efficacy and adverse events of amitriptyline on pain in fibromyalgia patients.

Three tiers of evidence

Possibly should incorporate these as variables into our analysis.

First tier

Second tier

Third tier

Only third-tier studies were found in the review

Outcomes

Analgesic efficacy: NNT (number needed to treat)

Harm: NNH number heeded to treat to harm

This is only important for transformation in the summary of findings, that is, after the model’s log risk ratios have been calculated. Not important for model. Importantly, to calculate NNT and NNH, we require a something somehting.

Model notes

Risk ratio of 50% pain relief using a fixed-effect meta-analysis

Results

import data from revman format

Show code 
fib_ami_dat_raw <- read.rm5("data/fibromyalgia-amitriptyline.rm5") %>% 
    janitor::clean_names() 

# tidy it up a bit
fib_ami_dat <-
  fib_ami_dat_raw %>% 
  rename(study = studlab, outcome =  outclab) %>% 
  select(study, event_c, event_e, n_c, n_e, outcome)

third-tier efficacy

Our objective is to replicate the results shown in this analysis.

Show code 
knitr::include_graphics("external-figs/moore-third-tier.png")
Show code 
third_tier_dat <- 
  fib_ami_dat %>% 
  filter(str_detect(outcome, "Third-tier efficacy"))
  
third_tier_dat %>% 
  gt(groupname_col = "outcome") %>% 
  hpp_tab(vertical_divider = "study") %>% 
  tab_header("Studies included in third-tier efficacy analysis")
Show code 
third_tier_escalc <-
  third_tier_dat %>%
  # does adding the complement help?
  mutate(comp_c = n_c - event_c, comp_e = n_e - event_e) %>%
  escalc(
    measure = "RR",
    ai = event_e,
    bi = comp_e,
    n1i = n_e,
    ci = event_c,
    di = comp_c,
    n2i = n_c,
    dat = .
  ) 
  
third_tier_rma <-   
  rma(
    yi = yi,
    vi = vi,
    slab = study,
    data = third_tier_escalc
  )



third_tier_mh <- 
  rma.mh(
    slab = study,
        measure = "RR",
    ai = event_e,
    bi = comp_e,
    n1i = n_e,
    ci = event_c,
    di = comp_c,
    n2i = n_c,
    data = third_tier_dat %>%   
      mutate(comp_c = n_c - event_c, comp_e = n_e - event_e),
    add = 1/2
  )  


forest(third_tier_mh, transf = exp)

Of course, Gav and I would likely do a random-effects model for this analysis, the results of which will likely adhere better to the NMA.

Show code 
forest(third_tier_rma, transf = exp)

reproduction of amitriptyline and fibromyalgia

Discuss with Hollie why Carette 1995 is omitted from the review.

Show code 
tar_read(w_covidence) %>% 
  filter(str_detect(study, "care|gins")) %>% 
  count(study)
Show code 
tar_load(w_obs)

reproduction of third-tier analysis

Show code 
# number of studies per outcome
tar_read(w_obs_long) %>% 
  group_by(outcome) %>% 
  summarise(studies = n_distinct(study))


# outcomes that have carette and ginsberg 
tar_read(w_obs_long) %>% 
  filter(str_detect(study, "caret|gins")) %>%
  count(outcome, study)

Discuss with Hollie scale labelling of Carette 1986; is analog scale 1-10 the same as vas?

Show code 
neuropathic_dat <- 
w_obs %>% 
  filter(condition_general == "fibromyalgia", 
         str_detect(arm, "placebo|amitrip")) %>%
  group_by(outcome, study) %>% 
  mutate(arms_per_study = 1:n()) %>% 
  filter(max(arms_per_study) > 1) %>% 
  # select(-arms_per_study) %>% 
  ungroup()
Show code 
# look for studies
neuropathic_dat %>% 
  filter(str_detect(study, "care|gins"), str_detect(outcome, "pain")) %>% 
  select(outcome, study) %>% 
  distinct()
Show code 
three_studies_dat <- 
neuropathic_dat %>% 
  filter(str_detect(study, "care|gins"), str_detect(outcome, "pain")) %>% 
  select(study, arm, covidence, scale, mean, n, sd, intervention, type)

Look into multilevel variance-covariance matrix structure

Show code 
three_studies_dat %>% 
  filter(type != "placebo") %>% 
  left_join(
    three_studies_dat %>% 
      filter(type == "placebo") %>% 
      select(study, covidence, mean_p = mean, n_p = n, sd_p = sd),
    by = c("study", "covidence")
  ) %>% 
  escalc(
    m1i = mean,
    sd1i = sd,
    n1i = n,
    m2i = mean_p,
    sd2i = sd_p,
    n2i = n_p,
    measure = "SMD",
    data = .,
    slab = study
  ) %>% 
  rma.mv(
    yi,
    vi,
    random = ~ 1 | study/covidence,
    data = .
  ) %>% forest()

reproduction with all studies

Show code 
all_dat <- neuropathic_dat %>% 
  filter(str_detect(outcome, "pain")) %>% 
  select(study, arm, covidence, scale, mean, n, sd, intervention, type) 

all_dat_rma <- 
all_dat %>% 
  filter(type != "placebo") %>% 
  left_join(
    all_dat %>% 
      filter(type == "placebo") %>% 
      select(study, covidence, mean_p = mean, n_p = n, sd_p = sd),
    by = c("study", "covidence")
  ) %>% 
  escalc(
    m1i = mean,
    sd1i = sd,
    n1i = n,
    m2i = mean_p,
    sd2i = sd_p,
    n2i = n_p,
    measure = "SMD",
    data = .,
    slab = study
  ) %>% 
  rma.mv(
    yi = yi,
    V = vi,
    random = ~ 1 |study/ covidence,
    data = .
  ) 


summary(all_dat_rma)

forest(all_dat_rma)

extension

Now we’ll try network meta-analysis. Again, we have to solve the multi-level thing.

nma of fibromyalgia

Show code 
# get fibromyalgia data
fibro_dat <- 
w_obs %>% 
  filter(condition_general == "fibromyalgia") %>%
  group_by(outcome, study) %>% 
  mutate(arms_per_study = 1:n()) %>% 
  filter(max(arms_per_study) > 1) %>% 
  # select(-arms_per_study) %>% 
  ungroup()
Show code 
# outcomes
fibro_dat %>% 
  count(outcome) %>% 
  gt()

pain nma

Show code 
pain_sub_net <- 
  set_agd_arm(
  data = fibro_dat %>% filter(outcome == "pain_sub"),
  trt = intervention,
  r = r,
  n = n,
  sample_size = n,
  trt_ref = "placebo",
  study = study
) 

plot(pain_sub_net)
Show code 
pain_int_net <- 
  set_agd_arm(
  data = fibro_dat %>% filter(outcome == "pain_int"),
  trt = intervention,
  y = mean,
  se = se,
  sample_size = n,
  trt_ref = "placebo",
  study = study
) 

pain_int_net

plot(pain_int_net)
Show code 
print(pain_int_rma)

relative_effects(pain_int_rma) %>% 
  plot()
Show code 
# gussy up the plot a bit
summary(pain_int_rma) %>% 
  as_tibble()
Moore, R. Andrew, Sheena Derry, Dominic Aldington, Peter Cole, and Philip J. Wiffen. 2015. “Amitriptyline for Fibromyalgia in Adults.” Cochrane Database of Systematic Reviews, no. 7. https://doi.org/10.1002/14651858.CD011824.

References