Figure 2. Memory Score Trajectories From Linear Mixed Models.
Estimated trajectories and 95% CIs (shaded area) of a hypothetical person with cancer diagnosis at 75 years of age (vertical line) in the reference categories of covariates (white race, female, 12 years of education, no history of alcohol use, no tobacco use, and no baseline comorbidities) compared with memory trajectories of a person with the same characteristics who remained cancer free. Memory score was standardized to its baseline mean (SD). Each unit in the memory score corresponds to approximately 1 SD in the baseline sample: yij = β0 + β1x1 + β2x2 + β3x3 + β4x4 + β5x5 + β6x6 + β7x7 + ∑βkxk, where y is the memory score for an individual i at j time, x1 is an indicator for having cancer diagnosis at any point in follow-up, x2 is an indicator for whether cancer has been diagnosed at j time (cancer now variable), x3 is j time minus time of diagnosis (number of years, always 0 for individuals who remained cancer free), x4 is years after diagnosis (number of years for people previously diagnosed with cancer, 0 if cancer had not yet been diagnosed, always 0 for individuals who remained cancer free), x5 is age at cancer diagnosis centered at 75 years (0 for individuals who remained cancer free), x6 is age at each interview centered at 75 years, and x7 is quadratic age at each interview centered at age 75 years, and xk is each of k baseline covariates. The estimated regression coefficients (β) are given in Table 2 and eTable 2 in the Supplement.