R/calculate_isi_matsuda.R

In JMLuther/tabletools: Helpful Tools to Make html Tables

#' Matsuda-Defronzo Index Calculation (from standard 75g OGTT)
#'
#' This function calculates the Matsuda index using glucose and insulin sampled
#' during a standard 75g oral glucose tolerance test. Results can be checked
#' using the \href{http://mmatsuda.diabetes-smc.jp/MIndex.html}{Matsuda online
#' calculator}. Examples with reduced sampling times have been validated using the online calculator.
#'
#' Standard timepoints (`sample_times`) are 0, 30, 60, 90, and 120 min. This
#' option can be changed to use alternative sampling times, and if extra
#' timepoints are provided, they will be dropped for analysis with a warning.
#'
#' If time is provided out of order, it will be reordered and trigger a warning.
#'
#' Note: insulin unit conversion may differ differ depending on assay. Insulin
#' (pmol/l) = insulin (uU/ml)*6
#'
#' `calculate_isi_matsuda()` accepts 3 separate vectors for time, glucose,
#' insulin.
#'
#' The formula used for this calculation is described in
#' \href{https://pubmed.ncbi.nlm.nih.gov/10480510/}{Matsuda et al.}:
#' \deqn{10,000 \cdot \sqrt{Glucose_{0} \cdot Insulin_{0} \cdot Glucose_{mean}
#' \cdot Insulin_{mean} }}
#'
#' @param time  Vector of time values (in minutes)
#' @param glucose Vector of glucose values (in mg/dL)
#' @param insulin Vector of insulin values (in uU/mL)
#' @param time_units if units are not in "min", can indicate here for unit
#'   conversion (options "min" or "hr")
#' @param glucose_units if units are not in "mg/dl", can indicate here for unit
#'   conversion (options "mg/dl" or "mmol/l")
#' @param insulin_units if units are not in "uU/ml", can indicate here for unit
#'   conversion (options "uU/ml" or "pmol/l")
#' @param sample_times Timepoints to use for calculation (default =
#'   c(0,30,60,90,120)). Timepoints not listed will be dropped.
#'
#' @return Matsuda index as a single value ()
#' @export
#' @importFrom sfsmisc integrate.xy
#' @examples
#' # individual objects for each item
#' time=c(0, 30, 60, 90, 120)              # minutes
#' glucose=c(93, 129, 178, 164, 97)        # mg/dL
#' insulin=c(12.8, 30.7, 68.5, 74.1, 44.0) # uU/mL
#' calculate_isi_matsuda(time, glucose, insulin) # 3.43125
#'
#' # handling data stored in a dataframe
#' ogtt1 <- data.frame(time=c(0, 30, 60, 90, 120),              # minutes
#'                     glucose=c(93, 129, 178, 164, 97),        # mg/dL
#'                     insulin=c(12.8, 30.7, 68.5, 74.1, 44.0)) # uU/mL
#' calculate_isi_matsuda(ogtt1$time, ogtt1$glucose, ogtt1$insulin) # 3.43125
#'
#' # Convert units
#' ogtt5 <- data.frame(time = c(0,0.5,1,1.5,2), # time in hours
#'                     glucose = c(5.167, 7.167, 9.889, 9.111, 5.3889), # glucose in mmol/l
#'                     insulin = c(76.8,184.2,411,444.6,264)) # insulin in pmol/l
#'
#' calculate_isi_matsuda(time =  ogtt5$time,
#'                         glucose = ogtt5$glucose,
#'                         insulin = ogtt5$insulin,
#'                         time_units = "hr", insulin_units = "pmol/l", glucose_units = "mmol/l")
#'
#' # Handle different time points
#' calculate_isi_matsuda(time, glucose, insulin, sample_times = c(0,30,60,90)) # 3.43125
#' calculate_isi_matsuda(time, glucose, insulin, sample_times = c(0,30,60,120)) # 3.68
#' calculate_isi_matsuda(time, glucose, insulin, sample_times = c(0,30,120)) # 4.72
#' calculate_isi_matsuda(time, glucose, insulin, sample_times = c(0,60,120)) # 3.56
#' calculate_isi_matsuda(time, glucose, insulin, sample_times = c(0,120)) # 5.58

calculate_isi_matsuda <- function(time, glucose, insulin, 
                                  time_units = "min", 
                                  sample_times = c(0,30,60,90,120),
                                  glucose_units = "mg/dl", insulin_units = "uU/ml") {

  # convert hrs to minutes
  if (!tolower(time_units) %in% c("min", "minutes", "min.")) {
    time = convert_time_to_min(time, time_units)
  }
  
  # check time points and keep only sample times:
  indx_t_valid <- time %in% sample_times # valid time points T/F
  t_complete <- sum(indx_t_valid) == length(sample_times) # all time points present T/F (possibly with extra)
  t_extra <- length(time) > length(sample_times) # extra time points T/F
  t_n_extra <- sum(!indx_t_valid)         # number of extra time points
  
  # Data Incomplete, missing timepoints
  if (!t_complete){
    rlang::warn("Missing timepoint(s); NA returned") # Incomplete timepoints
    return(NA_real_)}
  
  # Data Complete, Extra data dropped
  if (t_complete & t_extra){ 
    time = time[indx_t_valid]
    glucose = glucose[indx_t_valid]
    insulin = insulin[indx_t_valid]
    rlang::warn(paste0(t_n_extra, " Nonstandard timepoint(s) omitted in analysis"))}
  
  # Data Complete, Reordered
  if (t_complete & !all(order(time) == seq_along(time))) {
    indx_reorder <- order(time)
    time <- time[indx_reorder]
    glucose <- glucose[indx_reorder]
    insulin <- insulin[indx_reorder]
    rlang::warn("Data reordered for analysis")}
  
  # Data Missing
  if (any(is.na(time)) | any(is.na(glucose)) | any(is.na(insulin))) {
    rlang::warn("Check for missing values in glucose and insulin")
    return(NA_real_)}
  
  # Data non-numeric
  if (any(!is.numeric(time)) | any(!is.numeric(glucose)) | any(!is.numeric(insulin))) {
    rlang::warn("non-numeric values in time, glucose, or insulin")
    return(NA_real_)}
  
  # convert units; 
  # time = convert_time_to_min(time, time_units)
  glucose = convert_glucose_to_mgdl(glucose, glucose_units)
  insulin = convert_insulin_to_uU_ml(insulin, insulin_units)
  
  ind0 = which(time==0)
  g0 = glucose[ind0]
  ins0 = insulin[ind0]
  time_max = max(time)
  g_bar <- sfsmisc::integrate.xy(time, glucose, use.spline = FALSE)/time_max 
  ins_bar <- sfsmisc::integrate.xy(time, insulin,use.spline = FALSE)/time_max
  return(10000/sqrt(g0*ins0*g_bar*ins_bar))
}