In windyzn/urinaryDBP:

knitr::opts_chunk$set(collapse = TRUE, echo = FALSE, message = FALSE, warning = FALSE)
library(dplyr)
devtools::load_all()
load_data(update = TRUE)
set_options()
extrafont::loadfonts(device="win")

ds <- project_data
# 
# dsBase <- ds %>% 
#   dplyr::filter(fVN == "Baseline")
# 
dsComplete <- ds %>%
  dplyr::group_by(SID) %>%
  dplyr::filter(n() == 3) %>%
  dplyr::ungroup()

Subject Characterization

Subject characteristics across visit numbers

dsTableOne <- ds %>% 
  dplyr::filter(Sex == "Female")

subCharTable <- tableone::CreateTableOne(
  vars = c("Age",
           "Sex",
           "eGFR",
           "ACR",
           "UDBP",
           "VitaminD",
           "dmStatus",
           "BirthControl",
           "PeriodsStopped"),
  strata = c("fVN"),
  data = dsTableOne,
  factorVars = c("SmokeCigs",
                 "BirthControl",
                 "PeriodsStopped")
) %>% 
  print(nonnormal = c("UDBP",
                      "ACR")) %>% 
  knitr::kable()

Part 1 - Cross-Sectional

ACR

# Box plot of uVDBP in different albuminuria categories
ds %>% 
  filter(fVN == "Baseline") %>% 
  select(acrStatus, udbpCrRatio) %>%
  na.omit() %>%
  box_plot_slides("acrStatus", "log(udbpCrRatio)", 
            "Albuminuria",
            "log uVDBP:Creatinine (ug/mmol)")

# ANOVA
anova <- aov(formula = log(udbpCrRatio)~acr_status, data = ds1)
summary(anova)
TukeyHSD(anova)
rm(anova)

# Scatterplot of ACR and uVDBP ----------------------------------

ds %>% 
  filter(fVN == "Baseline") %>% 
  select(ACR, udbpCrRatio) %>% 
  na.omit() %>% 
  scatter_plot("log(ACR)", "log(udbpCrRatio)",
               "log Albumin:Creatinine (mg/mmol)",
               "log uVDBP:Creatinine (ug/mmol)",
               line = TRUE)

# Spearman Correlation ------------------------------------------

ds %>% 
  filter(fVN == "Baseline") %>% 
  cor.test(formula = ~ ACR + udbpCrRatio, data = ., method = "spearman")

eGFR

# Boxplot of uVDBP concentrations across eGFR categories --------------

ds %>% 
  filter(fVN == "Baseline") %>% 
  select(eGFRStatus, udbpCrRatio) %>% 
  na.omit() %>% 
  box_plot_slides("eGFRStatus", "log(udbpCrRatio)", 
            "Estimated GFR (ml/min/1.73m^2)",
            "log uVDBP:Creatinine (ug/mmol)")

# ANOVA

anova <- aov(formula = log(udbpCrRatio)~eGFR_status, data = ds1)
summary(anova)
TukeyHSD(anova)
rm(anova)

# Scatterplot of eGFR and uVDBP ----------------------------------

ds %>% 
  # dplyr::filter(eGFR < 150) %>% 
  dplyr::filter(fVN == "Baseline") %>% 
  dplyr::select(eGFR, udbpCrRatio) %>% 
  na.omit() %>% 
  scatter_plot("eGFR", "log(udbpCrRatio)",
               "Estimated Glomerular Filtration Rate (ml/min/1.73m^2)",
               "log uVDBP:Creatinine (ug/mmol)")

# Spearman Correlation ------------------------------------------

ds %>% 
  # filter(!(acrStatus == "Macroalbuminuria")) %>% 
  dplyr::filter(fVN == "Baseline") %>% 
  cor.test(formula = ~ eGFR + log(udbpCrRatio), data = ., method = "spearman")

# Multiple Regression -------------------------------------------

ds %>%
  dplyr::filter(fVN == "Baseline") %>% 
  # dplyr::mutate(UDBP = UDBP/1000) %>%
  dplyr::arrange(SID, VN) %>%

  mason::design("glm") %>% 
  mason::add_settings(family = stats::gaussian()) %>% 
  mason::add_variables("yvars", "UDBP") %>% 
  mason::add_variables("xvars", c("ACR", "eGFR")) %>% 
  mason::construct() %>% 
  # mason::add_variables("covariates", c("DM", "MedsBloodPressure")) %>%
  # mason::construct() %>%
  mason::add_variables("yvars", "VitaminD") %>% 
  mason::add_variables("xvars", "UDBP") %>% 
  mason::construct() %>% 
  mason::scrub()

Inflammation

ds %>% 
  dplyr::filter(fVN == "Baseline") %>% 
  scatter_plot("log(CRP)", "VitaminD",
               "CRP", "25(OH)D")


ds %>%
  dplyr::filter(fVN == "Baseline") %>% 
  # dplyr::mutate(UDBP = UDBP/1000) %>%
  dplyr::arrange(SID, VN) %>%

  mason::design("glm") %>% 
  mason::add_settings(family = stats::gaussian()) %>% 
  mason::add_variables("yvars", "VitaminD") %>% 
  mason::add_variables("xvars", "CRP") %>% 
  mason::construct() %>% 
  mason::scrub()

Hyperfiltration

ds %>% 
  dplyr::filter(eGFR < 45) %>% 
  dplyr::select(SID, VN, Age, eGFR, dmStatus, acrStatus)

ds %>% 
  scatter_plot("Age", "eGFR",
               "Age", "eGFR")

ds %>% 
  dplyr::select(dmStatus, eGFR, fVN) %>% 
  na.omit() %>% 
  box_plot("dmStatus", "eGFR",
               "Diabetic Status", "eGFR",
           facet = TRUE)

Part 2 - GEE

ds %>% 
  line_plot("")

dsGEE <- ds %>% 
  dplyr::mutate(udbpBase = ifelse(fVN == "Baseline", UDBP, NA),
                ageBase = ifelse(fVN == "Baseline", Age, NA),
                bmiBase = ifelse(fVN == "Baseline", BMI, NA),
                waistBase = ifelse(fVN == "Baseline", Waist, NA), 
                Sex = ifelse(Sex == "Male", 1, 0),
                Ethnicity = ifelse(Ethnicity == "European", Ethnicity, "Other"),
                Ethnicity = ifelse(Ethnicity == "European", 0, 1),
                fMedsBP = ifelse(fMedsBP == "Yes", 1, 0)
                # vitdBase = ifelse(fVN == "6Year", NA, VitaminD)
                ) %>% 
  dplyr::arrange(SID, fVN) %>% 
  dplyr::group_by(SID) %>% 
  tidyr::fill(udbpBase, ageBase, bmiBase, waistBase) %>% 
  dplyr::ungroup() %>%
  dplyr::arrange(SID, VN) %>% 
  dplyr::select(udbpBase, ACR, eGFR, VN, ageBase, Sex, Ethnicity, BMI, 
                Waist, DM, fMedsBP)

round(cor(dsGEE[sapply(dsGEE, is.numeric)], use = "complete.obs"), 2)

# Heatmap

corrplot::corrplot(cor(dsGEE[sapply(dsGEE, is.numeric)], use = "complete.obs"),
                   method = "color")

gee <- ds %>%
  dplyr::mutate(
  udbpBase = ifelse(fVN == "Baseline", UDBP, NA),
  ageBase = ifelse(fVN == "Baseline", Age, NA),
  fDM = relevel(as.factor(DM), "0")
  # Ethnicity = ifelse(Ethnicity == "European", Ethnicity, "Other"),
  ) %>%
  dplyr::filter(!(fVN == "Baseline" &
                    acrStatus == "Macroalbuminuria")) %>%
  dplyr::filter(!(fVN == "Baseline" & eGFRStatus == "Moderate")) %>%
  dplyr::arrange(SID, fVN) %>%
  dplyr::group_by(SID) %>%
  tidyr::fill(udbpBase, ageBase) %>%
  dplyr::ungroup() %>%
  dplyr::mutate(UDBP = UDBP/1000) %>%
  dplyr::arrange(SID, VN) %>%

  mason::design("gee") %>% 
  mason::add_settings(
    family = stats::gaussian(),
    corstr = "ar1",
    cluster.id = "SID"
  ) %>% 
  mason::add_variables("yvars", c("ACR", "eGFR")) %>% 
  mason::add_variables("xvars", "UDBP") %>% 
  mason::add_variables("covariates", c("VN", "ageBase", "fDM")) %>% 
  mason::construct() %>% 
  mason::scrub()

gee3 <- readRDS(file = "../data/ds.Rds") %>%
  dplyr::mutate(
  udbpBase = ifelse(fVN == "Baseline", UDBP, NA),
  ageBase = ifelse(fVN == "Baseline", Age, NA),
  VitaminD3 = ifelse(fVN == "6Year", NA, VitaminD)
  # Ethnicity = ifelse(Ethnicity == "European", Ethnicity, "Other"),
  ) %>%
  dplyr::arrange(SID, fVN) %>%
  dplyr::group_by(SID) %>%
  tidyr::fill(udbpBase, ageBase) %>%
  dplyr::ungroup() %>%
  # dplyr::group_by(VN) %>%
  # dplyr::mutate(udbpBase = as.numeric(scale(udbpBase))) %>%
  dplyr::mutate(UDBP = UDBP/1000) %>%
  # dplyr::ungroup() %>% 
  dplyr::arrange(SID, VN) %>%
  mason_gee(
  yvars = c("VitaminD3"),
  xvars = "UDBP",
  covars = c("VN", "ageBase", "DM")
  )

Graph

gee %>%
  dplyr::mutate(p.value = ifelse(p.value > 0.05, 1, 0.04)) %>%
  seer::view_main_effect(
  graph.options = 'dot.size',
  groups = '~Yterms',
  legend.title = 'p-values',
  xlab = 'Percent difference with 95% CI in the outcomes\nfor each unit increase in baseline uVDBP',
  ylab = 'Kidney Measures'
  ) +
  graph_theme(ticks = FALSE, legend.pos = 'right') +
  ggplot2::theme(legend.margin = grid::unit(0, 'cm'))