library(shiny)
library(bslib)
library(mrgsolve)
library(dplyr)
library(ggplot2)

# ── mrgsolve model ─────────────────────────────────────────────────────────────
# 2-compartment IV continuous infusion model
# Source: Jonckheere et al. 2019, Antimicrob Agents Chemother 63:e01655-19
# All parameters allometrically scaled to 70 kg reference body weight
# CL_renal driven by estimated creatinine clearance (eCrCL, Cockcroft-Gault)

model_code <- "
$PARAM @annotated
THETA1  : 2.29   : CL_renal coefficient (L/h/70kg)
THETA2  : 0.943  : CL_renal eCrCL power exponent
CLOTHER : 0.795  : Non-renal clearance (L/h/70kg)
V1      : 10.7   : Central volume of distribution (L/70kg)
V2      : 12.2   : Peripheral volume of distribution (L/70kg)
Q       : 11.0   : Intercompartmental clearance (L/h/70kg)
WT      : 70     : Body weight (kg)
ECRCL   : 60     : Estimated creatinine clearance (mL/min)

$CMT @annotated
CENT   : Central compartment (mg)
PERIPH : Peripheral compartment (mg)

$MAIN
// Allometric weight scaling (reference = 70 kg, exponent = 1)
double WTSCALE = WT / 70.0;

// Renal clearance with Cockcroft-Gault eCrCL power covariate
double CLrenal_i = THETA1 * pow(ECRCL / 60.0, THETA2) * WTSCALE;
double CLi       = CLrenal_i + CLOTHER * WTSCALE;
double V1i       = V1 * WTSCALE;
double V2i       = V2 * WTSCALE;
double Qi        = Q  * WTSCALE;

$ODE
// mrgsolve adds zero-order infusion rate to CENT automatically
dxdt_CENT   = -(CLi / V1i) * CENT - (Qi / V1i) * CENT + (Qi / V2i) * PERIPH;
dxdt_PERIPH =  (Qi / V1i) * CENT  - (Qi / V2i) * PERIPH;

$TABLE
double CP = CENT / V1i;   // plasma concentration (mg/L)

$CAPTURE CP CLi V1i
"

mod <- mcode("cefepime_jonckheere2019", model_code)

# ── Theme ──────────────────────────────────────────────────────────────────────
app_theme <- bs_theme(
  version = 5, bootswatch = "flatly", primary = "#8b5cf6"
) |> bs_add_rules("
  .metric-card {
    background: #f8f9fa; border-radius: 8px; padding: 15px;
    margin: 5px; text-align: center; border: 1px solid #dee2e6;
  }
  .metric-value { font-size: 24px; font-weight: bold; color: #2c3e50; }
  .metric-label { font-size: 12px; color: #7f8c8d; margin-top: 4px; }
  .metric-success .metric-value { color: #10b981; }
  .metric-warning .metric-value { color: #f59e0b; }
  .metric-primary .metric-value { color: #8b5cf6; }
  .metric-info    .metric-value { color: #0dcaf0; }
  .ref-box {
    background: #f0f4ff; border-left: 4px solid #8b5cf6;
    padding: 12px 16px; border-radius: 4px; margin-top: 10px; font-size: 13px;
  }
  .ref-box a { color: #8b5cf6; }
")

# ── UI ─────────────────────────────────────────────────────────────────────────
ui <- page_sidebar(
  title = "Cefepime PK Simulator — ICU Continuous Infusion",
  theme = app_theme,

  sidebar = sidebar(
    title = "Simulation Settings", width = 340,

    h6("Dosing (Continuous IV Infusion)"),
    sliderInput("dose_g",   "Daily Dose (g/day)",  min = 0.5, max = 6,   value = 1.5, step = 0.5),
    numericInput("n_days",  "Duration (days)",     value = 3, min = 1, max = 7),
    hr(),

    h6("Patient Characteristics"),
    sliderInput("wt",       "Weight (kg)",          min = 40, max = 120, value = 70, step = 5),
    sliderInput("ecrcl",    "eCrCL (mL/min)",       min = 10, max = 150, value = 60, step = 5),
    hr(),

    h6("PK/PD Target"),
    selectInput("mic", "Pathogen MIC (mg/L)",
      choices  = c("0.25" = 0.25, "0.5" = 0.5, "1" = 1, "2" = 2, "4" = 4, "8" = 8),
      selected = 1
    ),
    hr(),
    checkboxInput("log_scale", "Log Scale (Y-axis)", value = FALSE)
  ),

  navset_card_underline(
    title = "Cefepime IV Continuous Infusion Simulator",
    full_screen = TRUE,

    # ── Simulation tab ────────────────────────────────────────────────────────
    nav_panel("Simulation",
      layout_column_wrap(
        width = 1/4, fill = FALSE,
        div(class = "metric-card metric-success",
          div(class = "metric-value", textOutput("css")),
          div(class = "metric-label", "Predicted Css (mg/L)")
        ),
        div(class = "metric-card metric-warning",
          div(class = "metric-value", textOutput("tmic")),
          div(class = "metric-label", "T > MIC (%)")
        ),
        div(class = "metric-card metric-primary",
          div(class = "metric-value", textOutput("auc24")),
          div(class = "metric-label", "AUC (mg·h/L per day)")
        ),
        div(class = "metric-card metric-info",
          div(class = "metric-value", textOutput("thalf")),
          div(class = "metric-label", "Terminal t½ (h)")
        )
      ),
      card(
        full_screen = TRUE,
        plotOutput("pkPlot", height = "500px")
      )
    ),

    # ── Model Information tab ─────────────────────────────────────────────────
    nav_panel("Model Information",
      card(
        card_body(
          markdown("
## Cefepime — 2-Compartment IV Population PK Model

**Source:** Jonckheere et al. (2019). *Antimicrobial Agents and Chemotherapy* 63(9):e01655-19.  
**Fitting method:** NONMEM FOCE-I  
**Population:** 21 critically ill ICU patients; combined dataset (n = 2 studies)

---

### Population PK Parameter Estimates (Reference: 70 kg, eCrCL = 60 mL/min)

| Parameter | Estimate | Units | RSE (%) | Interpretation |
|-----------|----------|-------|---------|----------------|
| θ₁ (CL_renal coeff.) | 2.29 | L/h/70kg | 5.4 | Renal CL at reference eCrCL |
| θ₂ (eCrCL exponent) | 0.943 | — | 9.6 | Power relationship with eCrCL |
| CL_other | 0.795 | L/h/70kg | 9.0 | Non-renal clearance |
| V₁ (central) | 10.7 | L/70kg | 8.1 | Central volume |
| V₂ (peripheral) | 12.2 | L/70kg | 7.2 | Peripheral volume |
| Q₂ | 11.0 | L/h/70kg | 14 | Intercompartmental clearance |

**Total CL** (70 kg, eCrCL = 60 mL/min) ≈ 3.09 L/h

---

### Covariate Model

```
CL_renal = 2.29 × (eCrCL / 60)^0.943 × (WT / 70)
CL_total = CL_renal + CL_other × (WT / 70)
```

All PK parameters allometrically scaled to body weight (exponent = 1 for L parameters, 1 for L/h parameters per allometric theory).

---

### Between-Subject Variability (BSV, CV%)

| Parameter | BSV (CV%) |
|-----------|-----------|
| CL_renal  | 24.6% |
| V₁        | 45.7% |
| CL_other  | 69.4% |

---

### Therapeutic Drug Monitoring
- **TCI target:** 16 mg/L (continuous infusion steady state)
- **Therapeutic range:** 11.2–20.8 mg/L (±30% around target)
- **Toxicity threshold:** >35 mg/L (neurotoxicity risk)
- **PK-PD index:** T > MIC (time-dependent killing)
- **fT > MIC thresholds:** ≥50–70% for bacteriostatic/bactericidal effect
          ")
        )
      )
    ),

    # ── References tab ────────────────────────────────────────────────────────
    nav_panel("References",
      card(
        card_body(
          div(class = "ref-box",
            tags$h5("📚 Key References"),
            tags$ol(
              tags$li(
                "Jonckheere S et al. (2019). Target-Controlled Infusion of Cefepime in Critically Ill Patients. ",
                tags$em("Antimicrobial Agents and Chemotherapy"), " 63(9):e01655-19. ",
                tags$a(href = "https://doi.org/10.1128/AAC.01655-19", target = "_blank", "DOI")
              ),
              tags$li(
                "Jonckheere S et al. (2018). Development and external evaluation of a population pharmacokinetic model for cefepime dosing optimization. ",
                tags$em("J Antimicrob Chemother")
              ),
              tags$li(
                "Roberts JA et al. (2014). DALI study — Defining antibiotic levels in intensive care unit patients: are current beta-lactam antibiotic doses sufficient for critically ill patients? ",
                tags$em("Clin Infect Dis"), " 58(8):1072-1083."
              ),
              tags$li(
                "Tam VH et al. (2003). Pharmacodynamics of cefepime in patients with Gram-negative infections. ",
                tags$em("J Antimicrob Chemother")
              )
            ),
            tags$h5("💊 Drug Context"),
            tags$ul(
              tags$li(tags$strong("Drug class:"), " 4th-generation cephalosporin (beta-lactam)"),
              tags$li(tags$strong("Spectrum:"), " Broad-spectrum Gram-negative (Enterobacterales, Pseudomonas aeruginosa), limited Gram-positive"),
              tags$li(tags$strong("Indications:"), " Febrile neutropenia, pneumonia, UTI, sepsis in ICU patients"),
              tags$li(tags$strong("Renal elimination:"), " ~70–80% renally excreted unchanged; dose adjustment required for renal impairment"),
              tags$li(tags$strong("EUCAST breakpoints:"), " Enterobacterales S ≤1 mg/L; Pseudomonas S ≤8 mg/L"),
              tags$li(tags$strong("Toxicity concern:"), " Cefepime-induced neurotoxicity (CEFN) at unbound trough >22 mg/L")
            )
          )
        )
      )
    )
  ),

  # ── Footer ─────────────────────────────────────────────────────────────────
  div(
    style = "text-align: center; padding: 20px; margin-top: 30px; border-top: 1px solid #e9ecef; color: #6c757d; font-size: 12px;",
    "Powered by ",
    tags$a(
      href = "https://www.pkpdbuilder.com", target = "_blank",
      style = "color: #8b5cf6; font-weight: 500;", "PKPDBuilder.com"
    ),
    " • Cefepime ICU PopPK Simulator (Jonckheere 2019)", br(),
    tags$span(
      style = "font-size: 10px;",
      "For research and educational purposes only. Not for clinical decision-making."
    )
  )
)

# ── Server ──────────────────────────────────────────────────────────────────────
server <- function(input, output, session) {
  # thematic removed — not available on shinyapps.io

  # ── Reactive simulation ────────────────────────────────────────────────────
  sim_data <- reactive({
    shiny::req(input$dose_g, input$n_days, input$wt, input$ecrcl)

    inf_rate_mg_h <- input$dose_g * 1000 / 24           # mg/h
    total_mg      <- input$dose_g * 1000 * input$n_days # total mg over full duration

    ev1 <- ev(amt = total_mg, rate = inf_rate_mg_h, cmt = 1, time = 0)

    mod |>
      param(WT = input$wt, ECRCL = input$ecrcl) |>
      ev(ev1) |>
      mrgsim(end = input$n_days * 24, delta = 0.25) |>
      as.data.frame() |>
      mutate(time_h = time)
  })

  # ── Metrics ────────────────────────────────────────────────────────────────
  output$css <- renderText({
    d <- sim_data()
    last_h <- max(d$time_h)
    css_val <- d |> dplyr::filter(time_h >= last_h - 1) |> dplyr::pull(CP) |> mean(na.rm = TRUE)
    sprintf("%.1f", css_val)
  })

  output$tmic <- renderText({
    d     <- sim_data()
    mic_v <- as.numeric(input$mic)
    last24 <- dplyr::filter(d, time_h >= (input$n_days - 1) * 24)
    pct    <- mean(last24$CP > mic_v, na.rm = TRUE) * 100
    sprintf("%.0f%%", pct)
  })

  output$auc24 <- renderText({
    d          <- sim_data()
    last_start <- (input$n_days - 1) * 24
    seg        <- dplyr::filter(d, time_h >= last_start, time_h <= last_start + 24)
    auc        <- sum(diff(seg$time_h) * (head(seg$CP, -1) + tail(seg$CP, -1)) / 2)
    sprintf("%.1f", auc)
  })

  output$thalf <- renderText({
    shiny::req(input$wt, input$ecrcl)
    wt_s     <- input$wt / 70.0
    cl_renal <- 2.29  * (input$ecrcl / 60.0)^0.943 * wt_s
    cl_tot   <- cl_renal + 0.795 * wt_s
    v1i      <- 10.7 * wt_s
    v2i      <- 12.2 * wt_s
    qi       <- 11.0 * wt_s
    # 2-compartment terminal beta half-life
    k10  <- cl_tot / v1i
    k12  <- qi / v1i
    k21  <- qi / v2i
    s    <- k10 + k12 + k21
    beta <- 0.5 * (s - sqrt(max(s^2 - 4 * k10 * k21, 0)))
    if (beta <= 0) return("N/A")
    sprintf("%.1f", log(2) / beta)
  })

  # ── Plot ──────────────────────────────────────────────────────────────────
  output$pkPlot <- renderPlot({
    d       <- sim_data()
    mic_v   <- as.numeric(input$mic)
    d_plot  <- if (input$log_scale) dplyr::filter(d, CP > 0) else d
    max_t   <- max(d_plot$time_h)
    max_cp  <- max(d_plot$CP, na.rm = TRUE)
    y_tox   <- 35
    y_tci_lo <- 11.2
    y_tci_hi <- 20.8

    p <- ggplot(d_plot, aes(x = time_h, y = CP)) +
      # Therapeutic window band (±30% around 16 mg/L TCI target)
      annotate("rect", xmin = -Inf, xmax = Inf,
               ymin = y_tci_lo, ymax = y_tci_hi,
               fill = "#10b981", alpha = 0.12) +
      geom_hline(yintercept = c(y_tci_lo, y_tci_hi),
                 linetype = "dashed", color = "#10b981", alpha = 0.6) +
      annotate("text", x = max_t * 0.02, y = y_tci_hi * 1.06,
               label = "TCI target range (\u00b130%, 11.2\u201320.8 mg/L)",
               hjust = 0, size = 3.5, color = "#10b981") +
      # Toxicity threshold
      geom_hline(yintercept = y_tox,
                 linetype = "dotted", color = "#ef4444", alpha = 0.7) +
      annotate("text", x = max_t * 0.02, y = y_tox * 1.04,
               label = "Neurotoxicity threshold (35 mg/L)",
               hjust = 0, size = 3.5, color = "#ef4444") +
      # MIC line
      geom_hline(yintercept = mic_v,
                 linetype = "dashed", color = "#f59e0b", alpha = 0.8) +
      annotate("text", x = max_t * 0.02, y = mic_v * 1.12 + 0.2,
               label = paste0("MIC = ", mic_v, " mg/L"),
               hjust = 0, size = 3.5, color = "#f59e0b") +
      # PK concentration-time curve
      geom_line(color = "#8b5cf6", linewidth = 0.9) +
      labs(
        x     = "Time (hours)",
        y     = "Cefepime Plasma Concentration (mg/L)",
        title = paste0(
          "Cefepime ", input$dose_g, " g/day continuous IV infusion",
          " | ", input$wt, " kg, eCrCL = ", input$ecrcl, " mL/min"
        )
      ) +
      theme_minimal(base_size = 14) +
      theme(plot.title = element_text(size = 13, face = "bold"))

    if (input$log_scale) p <- p + scale_y_log10()
    p
  })
}

shinyApp(ui = ui, server = server)