Table 5.

Association between indoor tanning and risk of melanoma by possible recall and selection bias among cases and controls (Skin Health Study)

ObservedCasesControlsCrude OR (95% CI)Adjusted OR (95% CI)*
All participants
    n1,1671,101
    % ever tanned indoors62.951.11.62 (1.37-1.92)1.74 (1.42-2.14)
Evaluation of recall bias
    Participants who talked with their physician
        n213
        % ever tanned indoors71.466.71.25 (0.10-16.50)
    Participants who did not talk with their physician
        n130188
        % ever tanned indoors57.752.71.23 (0.78-1.92)1.72 (0.92-3.22)
Evaluation of selection bias
    Nonparticipants who answered brief questionnaire
        n107180
        % ever tanned indoors60.848.31.62 (1.00-3.61)§
  • *Adjusted for age, gender, eye color, natural hair color, skin color, freckles, moles, income, education, family history of melanoma, routine sun exposure, outdoor activity sun exposure, outdoor job exposure, mean sunscreen use, and number of lifetime painful sunburns; analysis among all participants excludes an additional 16 cases and 12 controls because the number of missing values was too small to be included as its own category. Analysis of recall bias excludes only two additional cases and three controls for the same reason.

  • Excludes nine cases and three controls who responded “don't know” or whose response was missing.

  • Not possible to estimate due to small numbers.

  • §Confounders not collected on nonparticipants.

  • *Adjusted for age, gender, eye color, natural hair color, skin color, freckles, moles, income, education, family history of melanoma, routine sun exposure, outdoor activity sun exposure, outdoor job exposure, mean sunscreen use, and number of lifetime painful sunburns; analysis among all participants excludes an additional 16 cases and 12 controls because the number of missing values was too small to be included as its own category. Analysis of recall bias excludes only two additional cases and three controls for the same reason.

  • Excludes nine cases and three controls who responded “don't know” or whose response was missing.

  • Not possible to estimate due to small numbers.

  • §Confounders not collected on nonparticipants.