Das sind jetzt diese zwei Faktoren, die als Ursache ausgeschlossen wurden? Oder auch noch andere?
Aber es gibt bestimmt noch mehr variablen, die zur Einnahme von kontrazeptiva führen und später zu depressiven Symptomen führen als die aktuelle Einnahme oder der Beginn der Einnahme?
Die Autoren sagen dazu:
Covariate selection
Covariates for the propensity score model were selected in a two step procedure. First, we identified central demographic and sample characteristics (NHANES cohort, age and educational level [high school graduate (reference); less than high school education; education beyond high school], marital status [married or living with partner (reference); never married; widowed, divorced or separated], poverty income ratio (a ratio of family income to poverty threshold),and ethnicity [non-Hispanic white (reference); non-Hispanic black; Hispanic; other]), and other variables that are theoretically or empirically linked to OC use and/or depression and were available in the dataset: bodymass index (BMI; weight in kilograms divided by height in metres squared; Luppino et al., 2010), age at menarche (KaltialaHeino, Kosunen, & Rimpel€a, 2003), age at sexual debut (Kaltiala-Heino et al., 2003), smoking history [no (reference); yes] (Kendleret al.,1993), sex of sexual partners [other-sex only (reference); same-sex only or both sexes] (Charlton et al., 2013; Marshal et al., 2011), diagnosed endometriosis [no (reference); yes] (Chen et al., 2016), ever use of Depo-Provera or injectables to prevent pregnancy (no [reference]; yes) (Skovlund et al., 2016), ever use of other medication containing female hormones such as oestrogen and progesterone [no(reference);yes] (Soares, Almeida, Joffe, & Cohen, 2001), current use of OCs for ever users [no (reference); yes] (Skovlund et al., 2016), total duration of lifetime OC use (ever users only) and pregnancy in the past year [no (reference); yes] (O’Hara, Schlechte, Lewis, & Varner, 1991). Second, following recommendations in the literature (DuGoff, Schuler, & Stuart, 2014), we selected those variables for our propensity score model that were related to the outcome (MDD), independently of whether they were also related to the exposure (history of OC use). Simulation studies show that inclusion of variables that are related to the outcome increases the precision of the estimated exposure effect without increasing bias; in contrast,including variables that are related to the exposure but not the outcome decrease the precision of the estimated exposure effect without decreasing bias (Brookhart et al., 2006). We used p < .20 as a conservative threshold for variables to be selected as covariates (Cheslack-Postava et al., 2015). This resulted in 11 covariates being included in the propensity score model: age at menarche (p = .15), age at sexual debut (p = .012), education (p = .10), marital status (p < .001), BMI (p = .003), diagnosis of endometriosis (p = .036), ever use of other medication containing female hormones such as oestrogen and progesterone (p = .012), pregnancy in the past year (p = .071), smoking history (p = .007), sex of sexual partners (p = .064) and current OC use (p = .17) (ever users of OCs only). NHANES cohort, age, poverty income ratio, ethnicity, ever use of Depo-Provera or injectables to prevent pregnancy and total duration of lifetime OC use were unrelated to MDD (all ps > .38) and were therefore not entered into the propensity score model.
[...]
Results
Of the 1,236 eligible women, 561 (45%) had first used OCs in adolescence, 353 (29%) had first used OCs as adults and 322 (26%) had never used OCs. Overall, 131 (11%) met the criteria for MMD according to DSM-5. Sample characteristics by history of OCs use are displayed in Table 1. Addressing the proposed relationship between history of OC use and depression, adult women who had used OCs during adolescence showed a higher 1-year prevalence of MDD compared with both adult women who had never used OCs [odds ratio (OR) = 0.31, 95% CI = 0.16–0.60] as well as women who had only started using OCs in adulthood (OR = 0.54, 95% CI = 0.30–0.95) in the unadjusted model. These effects remained stable when controlling for a large number of potential confounders using propensity score weighting (Table 2). Balancing of potential confounders was successful: Except for diagnosis of endometriosis (p = .14), ever use of other medication containing female hormones (p = .001) and pregnancy in the past year (p = .11), all variables were well balanced after applying propensity weighting (all ps > .26). Excluding women who had been pregnant in the past year, women with a diagnosis of endometriosis and ever users of other medication containing female hormones from analyses did not change the core results. Our sensitivity analysis among women who had their sexual debut in adolescence confirmed the results of our main analysis: Women who had used OCs during adolescence showed a significant increase in 1-year prevalence of MDD compared with both women who had only started using OCs as adults as well as women who had never used OC, both in the unadjusted and the propensity score weighted models (Table 3). Moreover, the effect was robust to using age at first OC use (in years) as a continuous predictor for ever users of OCs, both across the whole sample (OR = 0.90, 95% CI = 0.83– 0.98) and women who had their sexual debut in adolescence (OR = 0.89, 95% CI = 0.81–0.98).
Ich hab mich jetzt nicht in die Tabelle vertieft, aber ich würde daraus ableiten dass einige Faktoren die das Ergenis beeinflussen könnten mitanalysiert wurden und am Ende eben doch der Faktor Alter bei Beginn der Pilleneinnahme als bester Prediktor für spätere Depression herausgekommen ist. (Und dass es noch ganz andere Faktoren gibt die eine Depression auslösen/verstärken können ist auch klar.)