Loss of a co-twin at birth and subsequent risk of psychiatric disorders
Figures

Study design.
*GA, Gestational age, categorized as <28 week, 28–31 weeks, 32–36 weeks, >36 weeks. †Birth weight for gestational age was generated by calculating birth weight z-score for each gestational age and sex-specific population, and was categorized as <10th, 10–30th, 31–50th, 51–70th, 71–90th, >90th, or unknown. We did this separately for twins and singletons. £Birth order was categorized as 1, 2, 3, ≥4.

The association between loss of a co-twin at birth and subsequent risk of psychiatric disorders by attained age, analyses of population-based matched cohort.
*Time-varying hazard ratios were derived from flexible parametric survival models, allowing relative risk of psychiatric disorders to vary over attained age. A spline with five df (four intermediate knots and two knots at each boundary, placed at quintiles of distribution of events) was used for the baseline rate, while three df was used for the time-varying effect. All models were adjusted for birth year, sex, gestational age, birth weight for gestational age, birth order, maternal age at childbirth, low Apgar score (≤7) at 5/10 min, maternal educational level at childbirth, maternal cohabitation status during pregnancy, and family history of psychiatric disorders.
-
Figure 2—source data 1
Summary data for Figure 2.
- https://cdn.elifesciences.org/articles/63514/elife-63514-fig2-data1-v1.xls
Tables
Characteristics of the study cohorts.
Population-based matched cohort: twins exposed to loss of co-twin at birth vs. unexposed twins or singletons | Twin-sibling family cohort: twins exposed to loss of a co-twin at birth vs. their full siblings | ||||
---|---|---|---|---|---|
Exposed twins | Matched unexposed twins | Matched singletons | Exposed twins | Exposed full siblings | |
Number of individuals | 787 | 3935 | 3935 | 569 | 880 |
Age at end of follow-up, median (interquartile range), year | 18.0 (9.5–28.0) | 19.8 (10.7–29.2) | 19.5 (10.8–28.9) | 18.3 (9.7–26.7) | 20.5 (11.8–28.0) |
Follow-up time, median (IQR), year | 17.8 (9.3–27.8) | 19.3 (10.1–28.7) | 19.0 (10.3–28.5) | 18.1 (9.5–26.5) | 18.6 (9.9–25.1) |
% of male | 53.9 | 53.8 | 53.9 | 55.7 | 52.7 |
Gestational age, n (%) | |||||
<28 weeks | 181 (23.0) | 814 (20.7) | 833 (21.2) | 142 (25.0) | 2 (0.23) |
28–31 weeks | 184 (23.4) | 1001 (25.4) | 1052 (26.7) | 139 (24.4) | 6 (0.68) |
32–36 weeks | 216 (27.5) | 1090 (27.7) | 1030 (26.2) | 151 (26.5) | 62 (7.05) |
≥37 weeks | 162 (20.6) | 810 (20.6) | 800 (20.3) | 132 (23.2) | 809 (91.9) |
Unknown | 44 (5.59) | 220 (5.59) | 220 (5.59) | 5 (0.88) | 1 (0.11) |
Birth weight for gestational age*, n (%) | |||||
<10th | 82 (10.4) | 287 (7.29) | 417 (10.6) | 59 (10.4) | 76 (8.64) |
10–30th | 156 (19.8) | 738 (18.8) | 774 (19.7) | 113 (19.9) | 151 (17.2) |
31–50th | 207 (26.3) | 908 (23.1) | 1031 (26.2) | 151 (26.5) | 173 (19.7) |
51–70th | 136 (17.3) | 832 (21.1) | 681 (17.3) | 101 (17.8) | 175 (19.9) |
71–90th | 108 (13.7) | 589 (15.0) | 536 (13.6) | 78 (13.7) | 194 (22.1) |
>90th | 54 (6.86) | 361 (9.17) | 276 (7.01) | 39 (6.9) | 108 (12.3) |
Unknown | 44 (5.59) | 220 (5.59) | 220 (5.59) | 28 (4.9) | 3 (0.34) |
Apgar score ≤ 7 at 5/10 min, n (%) | |||||
No | 567 (72.1) | 3121 (79.3) | 3222 (81.9) | 411 (72.2) | 838 (95.2) |
Yes | 118 (15.0) | 332 (8.44) | 392 (9.96) | 84 (14.8) | 9 (1.02) |
Unknown | 102 (13.0) | 482 (12.3) | 321 (8.16) | 74 (13.0) | 33 (3.75) |
Maternal age at birth, n (%) | |||||
≤28 | 345 (43.8) | 1788 (45.4) | 2179 (55.4) | 267 (46.9) | 397 (45.1) |
29–32 | 209 (26.6) | 1098 (27.9) | 912 (23.2) | 158 (27.8) | 245 (27.8) |
≥33 | 233 (29.6) | 1049 (26.7) | 844 (21.5) | 144 (25.3) | 238 (27.1) |
Maternal educational level, n (%) | |||||
<9 years | 36 (4.57) | 190 (4.83) | 178 (4.52) | 19 (3.34) | 31 (3.52) |
9–12 years | 498 (63.3) | 2418 (61.5) | 2618 (66.5) | 359 (63.1) | 569 (64.7) |
>12 years | 241 (30.6) | 1272 (32.3) | 1089 (27.7) | 185 (32.5) | 273 (31.0) |
Unknown | 12 (1.52) | 55 (1.40) | 50 (1.27) | 6 (1.05) | 7 (0.80) |
Maternal cohabitation status, n (%) | |||||
Yes | 671 (85.3) | 3165 (80.4) | 3067 (77.9) | 494 (86.8) | 783 (89.0) |
No | 112 (14.2) | 748 (19.0) | 847 (21.5) | 72 (12.7) | 90 (10.2) |
Unknown | 4 (0.51) | 22 (0.56) | 21 (0.53) | 3 (0.5) | 7 (0.80) |
Family history of psychiatric disorders including suicide, n (%) | |||||
Yes | 78 (9.91) | 334 (8.49) | 395 (10.0) | 49 (8.61) | 93 (10.6) |
No | 709 (90.1) | 3601 (91.5) | 3540 (90.0) | 520 (91.4) | 787 (89.4) |
-
* Birth weight was standardized by singletons/twins, sex, and gestational age.
Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among twins after loss of a co-twin at birth, derived from different Cox models and by subtypes of psychiatric disorders.
Population-based matched cohort | Twin-sibling family cohort | |||||
---|---|---|---|---|---|---|
Number of cases (crude incidence rate, per 1000 person years), exposed twins/unexposed twins | HR (95% CI)* | Number of cases (crude incidence rate, per 1000 person years), exposed twins/matched singletons | HR (95% CI)* | Number of cases (crude incidence rate, per 1000 person years), exposed twins/full siblings | HR (95% CI)* | |
Model information: Model 1 Controlled for attained age, (as underlying time scale), sex, and birth characteristics (i.e., GA, birth weight for GA, maternal age at birth) | 178 (12.08)/600 (7.76) | 1.59 (1.33–1.90) | 178 (12.08)/723 (9.33) | 1.42 (1.19–1.68) | 130 (12.32)/130 (8.17) | 1.44 (0.83–2.51) |
Model 2 above + neonatal factors (Apgar score) | 1.57 (1.31–1.87) | 1.37 (1.15–1.63) | 1.43 (0.82–2.49) | |||
Model 3 above + family history of psychiatric disorders, maternal educational level, maternal cohabitation status | 1.56 (1.30–1.87) | 1.41 (1.19–1.69) | - | |||
Full adjusted HRs† for subtypes of psychiatric disorders | ||||||
Neurodevelopmental disorders (ADHD, ASD, and intellectual disabilities) | 71 (4.53)/224 (2.80) | 1.56 (1.16–2.08) | 71 (4.53)/270 (3.34) | 1.44 (1.09–1.92) | 52 (4.58)/35 (2.10) | 0.24 (0.05–1.30) |
Emotional disorders (depression, anxiety, stress-related disorders) | 105 (6.75)/293 (3.67) | 1.90 (1.49–2.42) | 105 (6.75)/386 (4.82) | 1.57 (1.25–1.98) | 79 (7.04)/85 (5.21) | 1.75 (0.89–3.44) |
Other psychiatric disorders | 84 (5.43)/310 (3.92) | 1.32 (1.02–1.70) | 84 (5.43)/361 (4.54) | 1.28 (1.00–1.64) | 60 (5.39)/61 (3.72) | 1.37 (0.61–3.08) |
-
ADHD, attention deficit hyperactivity disorder; ASD, autism spectrum disorder; GA,gestational age.
*Cox regression models were stratified by matching identifiers or family identifier, and adjusted for covariates mentioned in the ‘model information’ column. Attained age was applied as the underlying time scale.
-
†HRs were derived from fully adjusted Cox regression models, that is, model 3.
Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among the surviving twins after co-twin loss at birth, by characteristics of the twin pairs.
Population-based matched cohort | Twin-sibling family cohort | |||||
---|---|---|---|---|---|---|
Number of cases (crude incidence rate, per 1000 person years), exposed twins/unexposed twins | HR (95% CI)* | Number of cases (crude incidence rate, per 1000 person years), exposed twins /matched singletons | HR (95% CI) † | Number of cases (crude incidence rate, per 1000 person years), exposed twins/full siblings | HR (95% CI) £ | |
By gender of the surviving twins | ||||||
Male | 103 (13.08)/334 (7.87) | 1.74 (1.37–2.22) | 103 (13.08)/366 (8.61) | 1.63 (1.29–2.07) | 76 (12.96)/64 (7.51) | 2.46 (0.77–7.91) |
Female | 75 (10.94)/266 (7.61) | 1.37 (1.04–1.81) | 75 (10.94)/357 (10.23) | 1.21 (0.93–1.58) | 54 (11.53)/66 (8.93) | 1.22 (0.25–5.85) |
By family history of psychiatric disorders | ||||||
Yes | 20 (17.98)/80 (15.63) | 3.61 (0.43–30.1) | 20 (17.98)/106 (19.17) | 0.79 (0.25–2.56) | 12 (18.00)/16 (12.32) | - |
No | 158 (11.60)/520 (7.20) | 1.62 (1.33–1.97) | 158 (11.60)/617 (8.58) | 1.44 (1.19–1.74) | 118 (11.94)/114 (7.80) | 1.55 (0.80–3.00) |
By gender difference of the twin pair | ||||||
Same-sex twin pair | 130 (12.25)/423 (7.56) | 1.69 (1.34–2.12) | 130 (12.25)/78 (7.97) | 1.78 (1.21–2.63) | 98 (12.70)/93 (7.90) | 1.50 (0.73–3.11) |
Opposite-sex twin pair | 48 (11.65)/177 (8.27) | 1.30 (0.81–2.10) | 48 (11.65)/38 (9.64) | 1.18 (0.61–2.28) | 32 (11.29)/37 (8.94) | 0.96 (0.29–3.20) |
By survival days of the deceased twin | ||||||
0–6 days | 124 (11.72)/415 (7.58) | 1.57 (1.26–1.94) | 124 (11.72)/516 (9.42) | 1.34 (1.08–1.65) | 91 (12.09)/82 (7.54) | 1.67 (0.70–3.99) |
7–27 days | 35 (13.09)/134 (9.21) | 1.35 (0.89–2.05) | 35 (13.09)/142 (9.69) | 1.59 (1.07–2.36) | 26 (13.07)/24 (7.65) | 7.04 (0.79–62.4) |
28–59 days | 19 (12.87)/51 (6.33) | 2.56 (1.36–4.81) | 19 (12.87)/65 (8.13) | 1.66 (0.92–2.99) | 13 (12.58)/24 (12.71) | 0.23 (0.03–1.80) |
-
* Cox regression models were stratified by matching identifiers (sex, birth year, and gestational age), and adjusted for birth weight for gestational age, maternal age at childbirth, low Apgar score (≤7) at 5/10 min, maternal educational level at childbirth, maternal cohabitation status during pregnancy, and family history of psychiatric disorders.
† Cox regression models were stratified by matching identifiers (sex, birth year, gestational age, birth weight for gestational age, birth order), and adjusted for maternal age at childbirth, low Apgar score (≤7) at 5/10 min, maternal educational level at childbirth, maternal cohabitation status during pregnancy, and family history of psychiatric disorders.
-
£ Cox regression models were stratified by family identifiers, and adjusted for sex, birth year, gestational age, birth weight for gestational age, low Apgar score (≤7) at 5/10 min, maternal educational level at childbirth, and maternal cohabitation status during pregnancy.
Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among the surviving twins after co-twin loss at birth, by survival days of the deceased twin or using “28 days” to define the exposed twins.
Population-based matched cohort | Twin-sibling family cohort | |||||
---|---|---|---|---|---|---|
Number of cases (Crude incidence rate, per 1000 person years), exposed twins/unexposed twins | HR(95% CI) * | Number of cases (Crude incidence rate, per 1000 person years), exposed twins /matched singletons | HR(95% CI) † | Number of cases (Crude incidence rate, per 1000 person years), exposed twins/full siblings | HR(95% CI) £ | |
Main analysis | 178(12.08)/600(7.76) | 1.56 (1.30-1.87) | 178(12.08)/723(9.33) | 1.41 (1.19-1.69) | 130(12.32)/130(8.17) | 1.43 (0.82-2.49) |
Subgroup analysis: by survival days of the deceased twin | ||||||
0-6 days | 124(11.72)/415(7.58) | 1.57 (1.26-1.94) | 124(11.72)/516(9.42) | 1.34 (1.08-1.65) | 91(12.09)/82(7.54) | 1.67 (0.70-3.99) |
7-27 days | 35(13.09)/134(9.21) | 1.35 (0.89-2.05) | 35(13.09)/142(9.69) | 1.59 (1.07-2.36) | 26(13.07)/24(7.65) | 7.04 (0.79-62.4) |
28-59 days | 19(12.87)/51(6.33) | 2.56 (1.36-4.81) | 19(12.87)/65(8.13) | 1.66 (0.92-2.99) | 13(12.58)/24(12.71) | 0.23 (0.03-1.80) |
Sensitivity analysis: twins who lost a co-twin within 28 days after birth | 159(12.00)/549(7.92) | 1.52 (1.25-1.84) | 159(12.00)/658(9.47) | 1.38 (1.15-1.67) | 117(12.29)/106(7.56) | 1.74 (0.88-3.45) |
Additional files
-
Source code 1
SAS script for the primary analyses.
- https://cdn.elifesciences.org/articles/63514/elife-63514-code1-v1.zip
-
Supplementary file 1
Supplementary tables.
Supplementary Table 1. International Classification of Diseases (ICD), eighth (ICD-8; 1969–1986), ninth (ICD-9; 1987–1996), and tenth (ICD-10; 1997–2013) revisions codes for diagnoses used in this study Supplementary Table 2. Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among twins who lose a co-twin at birth, calculated separately for older and younger siblings of the surviving twins, subgroups by psychiatric disorders among parents during follow-up Supplementary Table 3. Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among twins who lose a co-twin at birth, subgrouped by or additionally adjusted for diagnosis of congenital abnormalities and severe somatic diseases during follow-up Supplementary Table 4. Hazard ratios (HRs) with 95% confidence intervals (CIs) for any psychiatric disorder among twins who lost a co-twin within 28 days after birth, derived from different Cox models and by subtypes of psychiatric disorders
- https://cdn.elifesciences.org/articles/63514/elife-63514-supp1-v1.docx
-
Transparent reporting form
- https://cdn.elifesciences.org/articles/63514/elife-63514-transrepform-v1.pdf