. 2019 Jan 31;5(6):864–871. doi: 10.1001/jamaoncol.2018.6634

Table 4. Dose-Response Models Examined and Corresponding Akaike Information Criterion and Log-Likelihood Deviance.

Description	Model^a	Akaike Information Criterion	Log-Likelihood Deviance
Daunorubicin vs doxorubicin
Log-linear model^b	Risk = exp (ΣαX) exp (0.56 doxoi1 + 1.51 doxoi2 + 2.52 doxoi3 + 0.34 dauni1 + 0.99 dauni2 + 1.76 dauni3)	6123.1	6089.1
Linear model^c	Risk = exp (ΣαX) [1 + 0.02963 doxo_dose + 0.01571 daun_dose]	6098.8	6072.8
Linear exponential model^c	Risk = exp (ΣαX) [1 + 0.01674 doxo_dose × exp (0.00157 doxo_dose) + 0.0129 daun_dose × exp (0.000597 daun_dose)]	6092.4	6062.4
Epirubicin vs doxorubicin
Log-linear model^b	Risk = exp (ΣαX) exp (0.37 doxoi1 + 1.41 doxoi2 + 2.41 doxoi3 + 0.73 epii1 + 0.71 epii2 + 1.83 epii3)	6247.5	6213.5
Linear model^c	Risk = exp (ΣαX) [1 + 0.02203 doxo_dose +0.01685 epi_dose]	6235.8	6209.8
Linear exponential model^c	Not estimable
Mitoxantrone vs doxorubicin^d
Log-linear model^b	Risk = exp (ΣαX) exp (0.43 doxoi1 + 1.46 doxoi2 + 2.44 doxoi3 + 1.51 mitoxi1 + 1.34 mitoxi2 + 3.89 mitoxi3)	6197.7	6163.7
Linear model^c	Risk = exp (ΣαX) [1 + 0.02313 doxo_dose + 0.07966 mitox_dose]	6192.0	6166.0
Linear exponential model^c	Risk = exp (ΣαX) [1 + 0.01265 doxo_dose × exp (0.0017 doxo_dose) + 0.02498 mitox_dose × exp (0.0036 mitox_dose)]	6178.7	6148.7

Abbreviations: daun, daunorubicin; doxo, doxorubicin; epi, epirubicin; mitox, mitoxantrone.

^{^a}

exp (ΣαX) contains these covariates: sex, age at cancer diagnosis, current age, chest radiotherapy exposure, exposure to another anthracycline or anthraquinone, and cohort.

^{^b}

Based on dose categories with no dose as referent, where listed covariates are indicators (i1 = 1-149; i2 = 150-299; and i3 = ≥300 mg/m²).

^{^c}

Based on continuous dose (in units of milligrams per square meter).

^{^d}

Mitoxantrone dose (in units of milligrams per square meter) first multiplied by 4 to facilitate comparability to doxorubicin on the same mg/m² scale.