Skip to main content

Table 2 Computational times for nested case-control and survival analyses of time-dependent data for cohorts of increasing sizes: Models* of the risk of pacemaker insertion in elderly Quebec residents with atrial fibrillation

From: Comparison of nested case-control and survival analysis methodologies for analysis of time-dependent exposure

Cohort Size†: #Subjects (#Cases) 1340 (53) 2680 (106) 5360 (212) 10720 (424) 21440 (848) 42880 (1696)
Cohort Size: multiple of original 1 2 4 8 16 32
CPU‡ Time (seconds):       
   Nested: 4 Controls per Case 0.03 0.05 0.07 0.11 0.18 0.4
   Nested: 32 Controls per Case 0.08 0.1 0.2 0.41 0.7 1.49
   Survival Analysis 1.26 2.51 5.06 9.54 19.53 49.91
CPU Time (multiple of Nested 4):       
   Nested: 4 Controls per Case 1 1 1 1 1 1
   Nested: 32 Controls per Case 3 2 3 4 4 4
   Survival Analysis 42 50 72 87 109 125
  1. * All models included four fixed variables (age, sex, calendar year of cohort entry, and baseline sinus node or conduction disorder) and seven time-dependent variables (amiodarone dose, ventricular arrhythmia, and exposure to sotalol, class I antiarrhythmic agents, beta-blockers, calcium channel blockers, and digoxin).
  2. † The original cohort of 1340 subjects was increased by progressive duplication to 2, 4, 8, 16, and 32 times its original size.
  3. ‡ CPU, central processing unit.