Table 3. Probabilistic sensitivity analyses correcting for differential non-reporting of cases with NMSC with sensitivity and specificity drawn from trapezoidal and uniform distributions.
| Analysis |
Basal cell carcinomaa
|
Squamous cell carcinoma
|
||||
|---|---|---|---|---|---|---|
| IRR 2.5 percentile | IRR median estimate | IRR 97.5 percentile | IRR 2.5 percentile | IRR median estimate | IRR 97.5 percentile | |
| Conventional analysis | 1.07 | 1.15 | 1.25 | 0.94 | 1.15 | 1.40 |
| Differential sensitivity analysis–systematic error | 1.17 | 1.45 | 1.92 | 1.15 | 1.47 | 2.00 |
| Differential sensitivity analysis–systematic and random error | 1.14 | 1.45 | 1.94 | 1.06 | 1.48 | 2.16 |
BCC=basal cell carcinoma; IRR=incidence rate ratios; NMSC=non-melanoma skin cancer.
Sensitivity among cases using glucocorticoids: Trapezodial distribution (minimum=0.55, mode 1=0.6, mode 2=0.9, maximum=0.95). Sensitivity among cases not using glucocorticoids: Uniform distribution (minimum=0.8, maximum=1.0). Specificity among non-cases: Uniform distribution (minimum=0.95, maximum=1.0).
a
The sensitivity analysis should be cautiously interpreted for risk of BCC among oral glucocorticoid users, because the SAS-macro did not converge after approximately 25 000 iterations. Therefore, the macro was reran and stopped after 5000 iterations.