Depressive symptoms and the onset of menarche

1. Depressive symptoms and the onset of menarche Jon Heron [1,2] Carol Joinson, Ricardo Araya, Glyn Lewis [1] Tim Croudace [2] [1] University of Bristol [2] University of Cambridge

2. ALSPAC • “Avon Longitudinal Study of Parents and Children” • AKA Children of the Nineties • Birth Cohort study of ~14,000 children and their families • Based in South-West of England • Started 1991 • Self completion questionnaires • Hands on assessments • Data from external sources • Biological samples

3. Where is Avon?

4. Data for current models • Depressive symptoms • Short form of Mood and Feelings Questionnaire Angold, A., Costello, E. J., Messer, S. C., Pickles, A., Winder, F., & Silver, D. (1995). Development of a short questionnaire for use in epidemiological studies of depression in children and adolescents. International Journal of Methods in Psychiatric Research, 5, 237–249. Costello, E. J., & Angold, A. (1988). Scales to assess child and adolescent depression: Checklists, screens and nets. Journal of the American Academy of Child and Adolescent Psychiatry, 27, 726–737. • Assessed in clinic (self report) • 3 time points: 10y7m, 12y10m, 13y10m

5. Depressive symptoms – Short MFQ • I felt miserable or unhappy (True/Sometimes/Not at all) • I didn't enjoy anything at all • I felt so tired I just sat around and did nothing • I was very restless • I felt I was no good any more • I cried a lot • I found it hard to think properly or concentrate • I hated myself • I was a bad person • I felt lonely • I thought nobody really loved me • I thought I could never been as good as other kids • I did everything wrong

6. Subscales: odd / even Measured using short MFQ (13 items) • I felt miserable or unhappy • I didn't enjoy anything at all • I felt so tired I just sat around and did nothing • I was very restless • I felt I was no good any more • I cried a lot • I found it hard to think properly or concentrate • I hated myself • I was a bad person • I felt lonely • I thought nobody really loved me • I thought I could never been as good as other kids • I did everything wrong

7. Data for current models • Age at onset of menarche • Assessed repeatedly by questionnaire • Respondent = mother and/or daughter • 3 groups: • Before 11½ (14.7%) • 11½ - 13½ (65.5%) • After 13½ (19.8%)

8. [1] Simple C/S models (n = 2063)

9. [1] Standardized estimates

10. However… • Scales are skewed • Data is longitudinal • Heterogeneity in patterns of change • Scale properties may not be invariant across time

11. [2] Measurement Invariance model

12. Item-1 Item-12 Item-1 Item-12 Item-1 Item-12 …… …… …… F_t1 F_t2 F_t3 Menarche onset [2] Measurement Invariance model • Constrained: • factor loadings • Item thresholds • correlated residual variances

13. Measurement Invariance model ~ change Item-1 Item-12 Item-1 Item-12 Item-1 Item-12 …… …… …… F_t1 F_t2 F_t3 Menarche onset

14. Measurement Invariance model CFI 0.952 TLI 0.975 Number of Free Parameters 57 RMSEA (Root Mean Square Error Of Approximation) 0.043 WRMR (Weighted Root Mean Square Residual) 1.892 Decent fit for fully invariant model

15. Compare models so far

16. Compare models so far

17. [1] C/S [2] Meas Invariant

18. [3] 2nd Order Growth Model

19. F_s2 [3] 2nd Order Growth Model – piecewise t1_odd t1_even t2_odd t2_even t3_odd t3_even F_t1 F_t2 F_t3 Zero residual variance F_i F_s1

20. [3] 2nd Order Growth Model INTCPT ON EARLYMEN 0.061 (0.071), p = 0.386 LATEMEN 0.037 (0.063), p = 0.559 INTCPT* ON EARLYMEN 0.237 (0.068), p = 0.001 LATEMEN -0.003 (0.061), p = 0.959 INTCPT** ON EARLYMEN 0.210 (0.066), p = 0.002 LATEMEN -0.185 (0.059), p = 0.002 SLOPE1 ON EARLYMEN 0.189 (0.073), p = 0.010 LATEMEN -0.033 (0.065), p = 0.606 SLOPE2 ON EARLYMEN 0.029 (0.071), p = 0.683 LATEMEN -0.214 (0.063), p = 0.001

21. Compare models so far

22. Compare models so far

23. [4] Growth model with mixture

24. [4] Two class results Movers (50.1%) Stayers (49.9%) C#1 ON EARLYMEN 1.363 [1.016, 1.828], p = 0.039 LATEMEN 0.863 [0.619, 1.112], p = 0.254

25. Compare models so far t1 t2 t3 s1 s2

26. Has that really helped with the skewness? 1st level factors within class:

27. What next? • Semi-continuous models • Zero-inflated models • Split measures in binary (zero/non-zero) and cts (e.g. log of non-zero scores) • However • ideal for instance where a score of zero is substantively different to score of 1+ • Prevalence of non-zero is 60-70% • Limited scope for modelling of 3 binaries (low df)

28. Summary / conclusions • We have 3 skewed measures of MFQ and wish to assess the extend to which early menarche leads to an earlier and/or greater onset in depressive symptoms • The models considered thus far are in good agreement and have not altered the conclusions made with the initial ANOVAs • Only factor model based on items dealt with scale skewness

29. Acknowledgements • Core Funding • MRC • Welcome • University of Bristol • Specific funding • ESRC • Thanks to all the study families

32. Restless: non-response more common

33. Restless: non-monotonic change

34. Measurement Invariance model ~ change Item-1 Item-12 Item-1 Item-12 Item-1 Item-12 …… …… …… F_t1 F_t2 F_t3 Early onset Late onset