Table 3.

Radiation dose response for breast cancer and evaluation of effect modification of the established breast cancer risk factors and characteristics of scoliosis diagnosis and treatment

Model*Potential effect modifierSubgroupRelative risk (main effect)ERR/Gy (95% CI)*Deviance differenceP§
ReferenceNoneAll2.86 (−0.07 to 8.62)Reference
1Age at first birth (y)<251.01.25 (<0-9.44)1.1080.57
25-341.236.21 (−0.55 to 29.61)
≥352.302.30 (<0-13.82)
2Menopausal status at questionnaire completionPremenopausal Postmenopausal1.0 3.822.94 (<0-9.89) 2.84 (−0.21 to 9.03)<0.0010.99
3Household income, dollars<30,0001.03.69 (<0-22.29)0.8680.83
30,000-59,0001.421.54 (<0-11.34)
≥60,0002.745.14 (<0-24.13)
4Any family history of breast cancerNo1.0−0.16 (<0-4.41)4.6140.03
Yes1.208.37 (1.50-28.16)
5Family history of breast cancer by degree of family relationshipNone Second degree only1.0 1.56−0.09 (<0-4.60) 6.90 (0.23-30.50)4.6100.10
Any first degree0.8311.83 (<0-122)
6Family history of early-onset breast cancer (before age 50 y)No1.02.76 (−0.12 to 8.50)1.320.25
Yes2.8417.65 (<0-91.4)
  • NOTE: An alternative reference model was used for model 4, which includes the “any family history” variable as confounder in the background risk term rather than the variable “family history by degree of relationship” that is used in all other models in this table.

  • * All models adjusted for attained age, menopausal status at questionnaire completion, age at first birth, household income, and an indicator of family history of breast cancer in the baseline; all analyses exclude 325 person-years (8 women, 0 cases) for women with unknown age at first birth.

  • The main effect relative risks are the coefficients from the log-linear element of the statistical model and reflect the effect of the potential effect modifier on the background risk of breast cancer, whereas the ERR/Gy reflects their effect as modifier of the radiation dose-response relationship for radiation-related breast cancer risk.

  • Relative to the main reference model (deviance 855.544) for models 1, 2, 3, and 5 and to an alternative reference model (deviance 856.848) for model 4.

  • § Likelihood ratio test of homogeneity of ERR/Gy across categories of effect modifier.

  • Lower confidence bound could not be obtained using likelihood-based methods.

  • Includes nulliparous women.