Table 3.
Risk of ASD in offspring of a mother with RA, by RA serostatus and the timing of exposure
| Analysis group | ASD rate (per 100 000 person year) | ASD cases (person years) | Model 1 HR (95% CI) | Model 3 HR (95% CI) |
|---|---|---|---|---|
| By serostatusa | ||||
| Seronegative | 314.8 | 35 (11 119) | 1.65 (1.16–2.37) | 1.61 (1.12–2.30) |
| Seropositive | 240.8 | 35 (14 534) | 1.28 (0.91–1.82) | 1.29 (0.91–1.82) |
| Non-RA | 187.8 | 28 892 (15 383 092) | Reference | Reference |
| Seronegative v. seropositive | – | – | 1.29 (0.78–2.12) | 1.25 (0.76–2.05) |
| By exposure timingb | ||||
| RA before delivery | 278.5 | 70 (25 133) | 1.48 (1.17–1.87) | 1.70 (1.35–2.15) |
| RA after delivery | 290.4 | 125 (43 047) | 1.17 (0.98–1.40) | 1.09 (0.92–1.30) |
| Non-RA | 187.7 | 28 767 (15 328 041) | Reference | Reference |
| RA before v. RA after delivery | – | – | 1.26 (0.94–1.69) | 1.56 (1.16–2.09) |
Abbreviations: RA, Rheumatoid Arthritis; ASD, Autism Spectrum Disorders; HR, Hazard Ratio; CI, Confidence interval.
Note: aTo decide serostatus status, we used the mother's all available RA diagnoses up to delivery, and defined seropositive or seronegative status as the most common type among these. If equally frequent, we used the last diagnosis to define the RA subtype.
Note: Relative risks were quantified by hazard ratios with 95% confidence intervals fitting Cox proportional hazard regression models with robust standard errors. Model 1: Adjusted for birth year by natural cubic splines with 5 knots; Model 3: Additionally adjusted for maternal and paternal age at delivery (by natural cubic splines), maternal and paternal education at delivery (categorized; < 9 years of primary education, 9 years of primary education, 1–2 years of secondary school education, 3 years of secondary school education, 1–2 years of postgraduate education, ⩾ 3 years of postgraduate education, PhD education), maternal and paternal income at delivery (by natural cubic splines), maternal and paternal psychiatric diagnoses at delivery (yes/ no).
bRelative risks with 95% confidence intervals were calculated using Poisson regression models, including RA diagnosis as a time-varying exposure. The cases and person-years are summarized at the end of follow-up. Model 1: Adjusted for birth year by natural cubic splines with 5 knots; Model 3: Additionally adjusted for maternal and paternal age at delivery (by natural cubic splines), maternal and paternal education at delivery (categorized), maternal and paternal income at delivery (by natural cubic splines), maternal and paternal psychiatric diagnoses at delivery (yes/ no).