DES vs BMS Therapeutic Decision Tree

Drug-eluting vs bare metal stents for coronary artery disease

R for HTA (Basics) — Workshop 2026

RRC-HTA, AIIMS Bhopal | HTAIn, DHR

The Clinical Question

Coronary artery disease (CAD) is a leading cause of mortality in India.

Percutaneous coronary intervention (PCI) with stent placement is standard.

The choice: Drug-eluting stents (DES) vs bare metal stents (BMS) has major cost and outcome implications.

The Indian Context

NPPA Price Ceilings (2025):

BMS: ~₹10,693
DES: ~₹38,933

The HTA Question

At current Indian price ceilings, is DES cost-effective compared to BMS over 5 years?

DES vs BMS: The Tradeoff

DES (Drug-Eluting Stent) - Higher initial cost - Lower restenosis rate (fewer repeat procedures) - Lower MACE rate (fewer MI/deaths)

BMS (Bare Metal Stent) - Lower initial cost - Higher restenosis rate (15–20% need repeat intervention) - Higher MACE rate

The question: Do DES’s downstream savings outweigh the upfront cost?

Key Parameters

Parameter	Value
DES stent cost	₹38,933
BMS stent cost	₹10,693
Procedure cost	₹120,000 (both)
DES restenosis	8%
BMS restenosis	20%
DES MACE (5yr)	8%
BMS MACE (5yr)	15%

Figure 1

All sourced from Indian data and NPPA 2025 guidelines.

The Decision Tree Structure

Code

library(DiagrammeR)

grViz("
digraph des_bms {
  graph [rankdir=LR, bgcolor='transparent', fontname='Helvetica', nodesep=0.3]
  node [fontname='Helvetica', fontsize=9]

  D [label='PCI Stent\nChoice', shape=square, style=filled, fillcolor='#4e79a7', fontcolor='white']

  DES [label='DES\n₹38,933', shape=box, style='filled,rounded', fillcolor='#4e79a7', fontcolor='white']
  BMS [label='BMS\n₹10,693', shape=box, style='filled,rounded', fillcolor='#e15759', fontcolor='white']

  DS [label='Success\n(97%)', shape=circle, style=filled, fillcolor='#d4e6f1', fontsize=8]
  DR [label='Restenosis\n(8%)', shape=circle, style=filled, fillcolor='#fdebd0', fontsize=8]
  DNR [label='No Restenosis\n(92%)', shape=circle, style=filled, fillcolor='#d5f5e3', fontsize=8]
  DM [label='MACE\n(8%)', shape=box, style='filled,rounded', fillcolor='#f5b7b1', fontsize=8]
  DW [label='Well', shape=box, style='filled,rounded', fillcolor='#d5f5e3', fontsize=8]

  BS [label='Success\n(96%)', shape=circle, style=filled, fillcolor='#d4e6f1', fontsize=8]
  BR [label='Restenosis\n(20%)', shape=circle, style=filled, fillcolor='#fdebd0', fontsize=8]
  BNR [label='No Restenosis\n(80%)', shape=circle, style=filled, fillcolor='#d5f5e3', fontsize=8]
  BM [label='MACE\n(15%)', shape=box, style='filled,rounded', fillcolor='#f5b7b1', fontsize=8]
  BW [label='Well', shape=box, style='filled,rounded', fillcolor='#d5f5e3', fontsize=8]

  D -> DES
  D -> BMS

  DES -> DS
  DS -> DR
  DS -> DNR
  DR -> DM
  DR -> DW
  DNR -> DM
  DNR -> DW

  BMS -> BS
  BS -> BR
  BS -> BNR
  BR -> BM
  BR -> BW
  BNR -> BM
  BNR -> BW
}
")

Figure 2: DES vs BMS decision tree for PCI

Initial PCI Cost

Code

# Total initial PCI cost per patient
initial_cost_des <- cost_procedure + cost_des_stent  # ₹120,000 + ₹38,933
initial_cost_bms <- cost_procedure + cost_bms_stent  # ₹120,000 + ₹10,693

# Cost difference
upfront_difference <- initial_cost_des - initial_cost_bms
# ₹28,240 per stent

Upfront cost: DES is ~₹28,240 more expensive per patient.

The question: Does this pay for itself downstream?

Tree Tracing: DES Arm

Code

# Per 1,000 patients
des_success <- 1000 * 0.97           # 970 patients
des_failure <- 1000 * 0.03           # 30 patients (CABG)

des_restenosis <- des_success * 0.08    # 77 restenosis
des_no_restenosis <- des_success * 0.92 # 893 no restenosis

# MACE among those with restenosis
des_mace_restenosis <- 77 * 0.15     # 12 MI/cardiac death

# MACE among those without restenosis
des_mace_no_restenosis <- 893 * 0.08 # 71 MI/cardiac death

des_total_mace <- 12 + 71            # 83 total MACE events

Tree Tracing: BMS Arm

Code

# Per 1,000 patients
bms_success <- 1000 * 0.96           # 960 patients
bms_failure <- 1000 * 0.04           # 40 patients

bms_restenosis <- bms_success * 0.20    # 192 restenosis (much higher!)
bms_no_restenosis <- bms_success * 0.80 # 768 no restenosis

# MACE outcomes
bms_mace_restenosis <- 192 * 0.25    # 48 MI/cardiac death
bms_mace_no_restenosis <- 768 * 0.15 # 115 MI/cardiac death

bms_total_mace <- 48 + 115           # 163 total MACE events

Patient Flow Comparison

Over 5 years per 1,000 patients:

DES: - Successful with no complications: ~893 - Restenosis requiring intervention: ~77 - MACE (MI or cardiac death): ~83

BMS: - Successful with no complications: ~768 - Restenosis requiring intervention: ~192 (2.5x higher!) - MACE: ~163 (2x higher!)

Cost Calculation: DES

Code

# Initial PCI
cost_initial_des <- des_success * initial_cost_des +
                    des_failure * cost_cabg

# Medications (5 years)
cost_medication_des <- n_cohort * cost_dapt_des_annual +  # Year 1
                      n_cohort * cost_followup_annual * 4  # Years 2-5

# Restenosis management (repeat PCI or CABG)
cost_restenosis_des <- des_restenosis * (
  0.80 * cost_repeat_pci +     # 80% repeat PCI
  0.20 * cost_cabg              # 20% CABG
)

# MACE management
cost_mace_des <- des_mi * cost_mi_management +
                 des_cardiac_death * cost_cardiac_death

total_cost_des <- cost_initial_des + cost_medication_des +
                  cost_restenosis_des + cost_mace_des

Cost Calculation: BMS

Code

# Initial PCI (lower stent cost)
cost_initial_bms <- bms_success * initial_cost_bms +
                    bms_failure * cost_cabg

# Medications (shorter DAPT)
cost_medication_bms <- n_cohort * cost_dapt_bms_annual +
                       n_cohort * cost_followup_annual * 4

# Restenosis management (higher rate!)
cost_restenosis_bms <- bms_restenosis * (
  0.80 * cost_repeat_pci +
  0.20 * cost_cabg
)

# MACE management (higher rate!)
cost_mace_bms <- bms_mi * cost_mi_management +
                 bms_cardiac_death * cost_cardiac_death

total_cost_bms <- cost_initial_bms + cost_medication_bms +
                  cost_restenosis_bms + cost_mace_bms

QALY Calculation

Code

# DES: QALYs over 5 years
qaly_des_well <- 893 * 0.85 * 5         # No complications
qaly_des_restenosis <- 77 * (0.78 * 4)  # After intervention
qaly_des_mace <- 83 * (0.65 * 3)        # Post-MI survival
qaly_des_death <- (83 * 0.3) * 0         # Death contributes 0 QALYs

total_qaly_des <- qaly_des_well + qaly_des_restenosis + qaly_des_mace

# BMS: QALYs (similar structure, with higher MACE)
# ...similar calculations but with higher adverse event rates

Higher MACE in BMS arm → Lower QALYs in BMS.

ICER Result

Code

inc_cost <- total_cost_des - total_cost_bms
inc_qaly <- total_qaly_des - total_qaly_bms
icer <- inc_cost / inc_qaly

# 4-quadrant interpretation (always check signs!)
if (inc_cost < 0 & inc_qaly > 0) {
  cat("DES is DOMINANT (cheaper AND better)")
} else if (inc_cost > 0 & inc_qaly < 0) {
  cat("DES is DOMINATED")
} else if (inc_cost > 0 & inc_qaly > 0) {
  if (icer < 170000) cat("Cost-effective") else cat("Not CE")
} else {
  cat("Trade-off")
}

Net Monetary Benefit (NMB)

Recall from Session 3: NMB = WTP × QALYs − Cost

Code

nmb_des <- wtp * total_qaly_des - total_cost_des
nmb_bms <- wtp * total_qaly_bms - total_cost_bms
inc_nmb <- nmb_des - nmb_bms
# Positive ΔNMB → adopt DES

If DES is dominant (negative ΔCost, positive ΔQALYs), ΔNMB is always positive — no ambiguity.

Number Needed to Treat (NNT)

How many patients need DES to prevent one adverse event?

Code

# Prevent one restenosis event
nnt_restenosis <- 1 / (0.20 - 0.08)  # NNT = 8.3

# Prevent one MACE event
nnt_mace <- 1 / ((163 - 83) / 1000)  # NNT = 12.5

cat("NNT to prevent 1 restenosis:", round(nnt_restenosis, 1), "\n")
cat("NNT to prevent 1 MACE:", round(nnt_mace, 1), "\n")

For every 8-13 DES stents placed instead of BMS, you prevent one adverse event.

The Key Insight

DES costs more upfront (₹28,240 per stent).

BUT: - Fewer repeat revascularisations - Fewer MI events - Fewer cardiac deaths

Downstream savings may outweigh upfront cost — depending on: - Restenosis rate differential - Cost of managing complications - Strength of evidence from Indian data

Therapeutic Decision Trees

This model structure applies to any treatment choice:

Drug A vs Drug B (cost vs efficacy)
Surgery vs Medical (invasive vs conservative)
Screening intensity (more tests vs fewer tests)

Each branches on: 1. Procedural success/failure 2. Complication rates 3. Long-term outcomes (MACE, relapse, etc.)

Why R Excels Here

In Excel: If NPPA revises DES price from ₹38,933 to ₹35,000, you manually update cells and pray you didn’t break formulas.

In R: Change one number at the top → entire model re-calculates in milliseconds.

This is essential for rapid scenario analysis in HTA policy work.

Key Takeaway

A therapeutic decision tree is another reusable template:

Treatment choice (two or more strategies)
Probability branches (success, complication, outcome rates)
Cost & QALY tracking through each pathway
ICER with 4-quadrant interpretation + NMB

Just like diagnostics, but for treatment and intervention decisions.

Bonus: rdecision Package

The same DES vs BMS model rebuilt with rdecision:

Object-oriented tree with DecisionNode, ChanceNode, LeafNode
Built-in tornado DSA and PSA (1,000 iterations)
CE plane + CEAC for decision uncertainty

→ See the Session 4 Bonus page on the workshop website.