• aPLs are common in pediatric patients with acute provoked VTE and are mostly transitory and clinically insignificant.

  • Children with APS and provoked VTE appear to be at an increased risk of recurrent VTE compared with those with transitory or low-titer aPL.

Abstract

Few studies have prospectively evaluated the incidence and outcomes in children with provoked venous thromboembolism (VTE) and transient or persistent antiphospholipid antibodies (aPLs). We compared outcomes of patients aged <21 years with a first-episode acute provoked VTE and positive aPL at diagnosis, enrolled in the Multicenter Evaluation of the Duration of Therapy for Thrombosis in Children trial. aPLs were tested at enrollment and, when positive, repeated at 6 weeks after VTE diagnosis. Subsequent testing was performed at the discretion of the treating hematologist. Of 524 patients, 116 (22%) had positive aPLs at enrollment. At follow-up, 70 (60%) had transient (n = 66) or low-titer aPLs (n = 4), 11 (10%) had persistent aPLs meeting the criteria for antiphospholipid antibody syndrome (APS), and 35 (30%) had no repeat testing. Patients with APS were older (15.8 vs 9.9 years; P = .014) and had a statistically significant higher risk of symptomatic recurrent VTE (18% vs 1%; odds ratio [OR], 12.2; 95% confidence interval [CI], 1.4-108; P = .025) and a statistically nonsignificant but clinically meaningful difference in the risk of anticoagulant-related clinically relevant bleeding (9% vs 0%; OR, 20.1; 95% CI, 0.7-558; P = .077) compared with those in the transient or low-titer aPL group. In conclusion, aPLs are common in young patients with acute provoked VTE and are mostly transitory and clinically insignificant. Patients with APS and provoked VTE appear to have an increased risk of recurrent VTE compared with patients with transitory or low-titer aPLs. Future collaborative studies should investigate the optimal VTE management for children with provoked VTE who meet the criteria for APS. The trial was registered at www.ClinicalTrials.gov as #NCT00687882.

Low-titer and/or transient antiphospholipid antibodies (aPLs) can be seen in 2% to 28% of healthy children and are particularly common after infection or vaccination.1,2 These antibodies are thought to be of minimal clinical relevance in patients without venous thromboembolism (VTE).3 Prior work has shown that aPLs are also common in children presenting with first-time or recurrent vascular thrombosis.4,5 However, the prevalence and potential clinical importance of aPLs in children and young adults with VTE remains unclear and understudied; this is especially true for those with provoked VTE.

Accordingly, we aimed to prospectively determine the incidence of transient/low-titer aPL and persistent aPLs meeting the criteria for antiphospholipid antibody syndrome (APS) in children with acute provoked VTE; and to compare clinical outcomes (ie, development of symptomatic recurrent VTE and/or anticoagulant-associated clinically relevant bleeding [CRB]) between children with transient/low-titer aPLs and APS.

Study design

We performed a prespecified analysis of the National Heart, Lung, and Blood Institute–sponsored Multicenter Evaluation of the Duration of Therapy for Thrombosis in Children (Kids-DOTT) randomized clinical trial (RCT; ClinicalTrials.gov identifier: NCT00687882) on the duration of anticoagulant therapy in patients aged <21 years with a first-episode of acute provoked VTE.6 The Institutional Review Board at each participating site approved the study. Written informed consent was obtained in all cases.

The trial’s design and eligibility criteria have been previously reported in detail.6,7 In summary, patients were enrolled within 30 days after diagnostic confirmation of a provoked VTE; managed with anticoagulant agents in accordance with international guidelines8,9; and followed for 1 year for the development of symptomatic recurrent VTE (defined as radiologically confirmed VTE at a new site or progression at the index site occurring ≥10 days after incident VTE diagnosis) and CRB (defined as major and clinically relevant nonmajor bleeding).10 aPL testing was obtained in all participants at enrollment and at 6 weeks after VTE diagnosis in those with an initial positive aPL; subsequent testing in patients whose aPLs persisted at 6 weeks after VTE diagnosis was performed at the discretion of the primary hematologist. Patients with negative aPLs at enrollment (or if positive at enrollment, then negative at 6 weeks) were randomized to a 6-week vs 12-week course of anticoagulant therapy. Patients with positive aPLs at 6 weeks were included in a nonrandomized, parallel cohort arm in which a minimum anticoagulant duration of 3 months was prescribed.6,7 

aPL testing

aPL testing included the following: (1) a lupus anticoagulant (LA) testing method meeting the International Society on Thrombosis and Haemostasis (ISTH) criteria11 (≥1 of LA-sensitive activated partial thromboplastin time, dilute Russell’s viper venom time, or hexagonal phase phospholipid assay [StaCLOT LA]); (2) anti-cardiolipin immunoglobulin M (IgM); and (3) anti–β-2-glycoprotein-1 (aβ2GPI) IgG and IgM. All aPL testing was performed at the local clinical laboratory at participating sites during the study. Testing was not blinded, and clinicians had access to the results to inform clinical decisions.

Statistical analyses

Descriptive statistics were used to summarize data on patient characteristics and outcomes of interest. Prespecified patient groups were classified as follows: (1) transient or low-titer aPL, which included patients with positive aPL at enrollment but negative on repeat testing at follow-up or patients with persistent (≥2 occasions at least 12 weeks apart) low-level positive aPL titers not meeting revised Sapporo criteria for APS12; and (2) APS, which included patients with persistent aPL on ≥2 occasions at least 12 weeks apart meeting revised Sapporo criteria for APS.12 Continuous variables were summarized as means with standard deviations, and their distributions were compared between groups using independent t test or Mann-Whitney U test, as appropriate. Categorical variables were summarized as counts with percentages and compared between groups via χ2 test or Fisher exact test, as appropriate. Associations between aPL status (transient/low titer vs APS) and outcomes were also determined by univariable logistic regression with Firth penalized likelihood approach to obtain odds ratios (ORs) and 95% confidence intervals (CIs). All statistical analyses were performed with SAS version 9.4. All statistical tests were 2-sided, and a P value < .05 was considered statistically significant.

Patient characteristic by aPL status

Of the 532 patients in the Kids-DOTT RCT, 524 (99%) underwent aPL testing at enrollment, and 116 (22%) tested positive (Figure 1). Patients with positive aPL at enrollment had a higher frequency of infection as their provoking VTE factor than those with negative aPL (28% vs 39%, respectively; P = .019). They were also older (7.6 vs 9.8 years, respectively; P = .087) and had a lower frequency of catheter-associated VTE (56% vs 47%, respectively; P = .068), although these differences were not statistically significant (Table 1).

Figure 1.

Distribution of patients on follow-up by aPL status.

Figure 1.

Distribution of patients on follow-up by aPL status.

Close modal
Table 1.

Patient characteristics by aPL status at enrollment

Positive aPL at baseline (n = 116)Negative aPL at baseline (n = 408)P value
Median age (IQR), y 9.8 (1.5-15.8) 7.6 (0.8-14.9) .087 
Female sex, n (%) 53 (46) 192 (47) .79 
White, n (%) 85 (73) 294 (72) .16 
Index VTE anatomical site, n (%)   .36 
Upper or lower extremity 86 (74) 312 (77)  
Cerebral sinus venous thrombosis 15 (13) 60 (15)  
Other (splanchnic vein, IVC, or IJ only) 15 (13) 35 (9)  
VTE-provoking factor, n (%)    
Catheter-associated VTE 54 (47) 229 (56) .068 
Infection 41 (39) 100 (28) .019 
Autoimmune/inflammatory disease 3 (3) 11 (3) .51 
Trauma or surgery 19 (18) 72 (20) .50 
Other (dehydration and genetic syndrome) 12 (11) 36 (10) .43 
Type of aPL, no. of positive/no. of tested (%)  
LAS-aPTT 14/29 (40) — — 
Dilute Russell’s viper venom time 40/43 (93) — — 
STACLOT LA 38/52 (73) — — 
Anti-cardiolipin IgM 17/112 (15) — — 
aβ2GP IgG 16/113 (14) — — 
aβ2GP IgM 11/113 (10) — — 
Positive aPL at baseline (n = 116)Negative aPL at baseline (n = 408)P value
Median age (IQR), y 9.8 (1.5-15.8) 7.6 (0.8-14.9) .087 
Female sex, n (%) 53 (46) 192 (47) .79 
White, n (%) 85 (73) 294 (72) .16 
Index VTE anatomical site, n (%)   .36 
Upper or lower extremity 86 (74) 312 (77)  
Cerebral sinus venous thrombosis 15 (13) 60 (15)  
Other (splanchnic vein, IVC, or IJ only) 15 (13) 35 (9)  
VTE-provoking factor, n (%)    
Catheter-associated VTE 54 (47) 229 (56) .068 
Infection 41 (39) 100 (28) .019 
Autoimmune/inflammatory disease 3 (3) 11 (3) .51 
Trauma or surgery 19 (18) 72 (20) .50 
Other (dehydration and genetic syndrome) 12 (11) 36 (10) .43 
Type of aPL, no. of positive/no. of tested (%)  
LAS-aPTT 14/29 (40) — — 
Dilute Russell’s viper venom time 40/43 (93) — — 
STACLOT LA 38/52 (73) — — 
Anti-cardiolipin IgM 17/112 (15) — — 
aβ2GP IgG 16/113 (14) — — 
aβ2GP IgM 11/113 (10) — — 

aPTT, activated partial thromboplastin time; IJ, internal jugular; IQR, interquartile range; IVC, inferior vena cava; LAS, lupus anticoagulant sensitive.

On follow-up testing, 70 patients (60%) had a transient (n = 66) or low-titer aPL (n = 4), 11 (10%) met the criteria for APS, and 35 (30%) had no repeat testing (Figure 1). Patients without repeat aPL testing (n = 35) were younger than patients with repeat aPL testing at follow-up (n = 81; 11.5 vs 5.7 years; P = .039) and had a higher proportion of positive aβ2GPI IgG antibodies at baseline (9% vs 26%; P = .014) and a lower proportion of positive dilute Russell’s viper venom time at baseline (88% vs 57%; P = .020). There were no other statistically significant differences between the groups (supplemental Table 1).

Patients meeting the criteria for APS were older than those in the transient or low-titer aPL group (15.8 vs 9.9 years; P = .014). They also had a higher proportion of females and lower rates of catheter-associated VTE, although these differences were not statistically significant (females, 64% vs 43% [P = .41]; catheter-associated VTE, 27% vs 43% [P = .51]; Table 2).

Table 2.

Patient characteristics and outcomes by aPL status at follow-up

APS (n = 11)Transient or low-titer aPL (n = 70)P value
Median age (IQR), y 15.8 (13.4-17.8) 9.9 (2-15.9) .014 
Sex female, n (%) 7 (64) 30 (43) .329 
Index VTE anatomical site, n (%)   1.00 
Upper or lower extremity 8 (73) 48 (69)  
Cerebral sinus venous thrombosis 2 (18) 11 (16)  
Other (splanchnic vein, IVC, or IJ only) 1 (9) 11 (16)  
Length of anticoagulation therapy, n (%)   
6-wk therapy 25 (36)  
3-mo therapy 11 (100) 45 (64)  
Complete veno-occlusion at 6 wk after VTE 1 (9) 7 (10) 1.00 
VTE-provoking factor, n (%)    
Catheter-associated VTE, n (%) 3 (27) 30 (43) .511 
Infection 4 (36) 26 (37) .764 
Autoimmune/inflammatory disease 3 (4) .72 
Trauma or surgery 3 (27) 9 (13) .164 
Other (dehydration and genetic syndrome) 1 (9) 8 (11) .686 
Type of aPL at enrollment, no. of positive (%)    
LAS-aPTT 3 (27) 7 (10) .325 
Dilute Russell’s viper venom time 5 (45) 31 (44) .567 
STACLOT LA 3 (27) 23 (33) .505 
Anti-cardiolipin IgM 1 (9) 12 (17) .681 
aβ2GP IgG/IgM 3 (27) 10 (14) .377 
High titer aPL at enrollment (>40 MPL, SGU, or SMU U/mL), n (%)    
Anti-cardiolipin IgM 
aβ2GP IgG 1 (9) 1 (1) .274 
aβ2GP IgM 1 (9) 1 (1) .274 
Outcomes (1 y after VTE)    
Symptomatic recurrent VTE 2 (18) 1 (1) .047 
Anticoagulant-associated CRB 1 (9) .135 
APS (n = 11)Transient or low-titer aPL (n = 70)P value
Median age (IQR), y 15.8 (13.4-17.8) 9.9 (2-15.9) .014 
Sex female, n (%) 7 (64) 30 (43) .329 
Index VTE anatomical site, n (%)   1.00 
Upper or lower extremity 8 (73) 48 (69)  
Cerebral sinus venous thrombosis 2 (18) 11 (16)  
Other (splanchnic vein, IVC, or IJ only) 1 (9) 11 (16)  
Length of anticoagulation therapy, n (%)   
6-wk therapy 25 (36)  
3-mo therapy 11 (100) 45 (64)  
Complete veno-occlusion at 6 wk after VTE 1 (9) 7 (10) 1.00 
VTE-provoking factor, n (%)    
Catheter-associated VTE, n (%) 3 (27) 30 (43) .511 
Infection 4 (36) 26 (37) .764 
Autoimmune/inflammatory disease 3 (4) .72 
Trauma or surgery 3 (27) 9 (13) .164 
Other (dehydration and genetic syndrome) 1 (9) 8 (11) .686 
Type of aPL at enrollment, no. of positive (%)    
LAS-aPTT 3 (27) 7 (10) .325 
Dilute Russell’s viper venom time 5 (45) 31 (44) .567 
STACLOT LA 3 (27) 23 (33) .505 
Anti-cardiolipin IgM 1 (9) 12 (17) .681 
aβ2GP IgG/IgM 3 (27) 10 (14) .377 
High titer aPL at enrollment (>40 MPL, SGU, or SMU U/mL), n (%)    
Anti-cardiolipin IgM 
aβ2GP IgG 1 (9) 1 (1) .274 
aβ2GP IgM 1 (9) 1 (1) .274 
Outcomes (1 y after VTE)    
Symptomatic recurrent VTE 2 (18) 1 (1) .047 
Anticoagulant-associated CRB 1 (9) .135 

MPL, microgram of IgM antibody; SGU, standard IgG anti-beta 2 glycoprotein unit; SMU, standard IgM anti-beta 2 glycoprotein unit.

aPL characteristics and outcomes

Table 1 shows the aPL characteristics at the time of enrollment. LA was the most common aPL detected. Among patients with positive aβ2GPI IgG and/or IgM (n = 27), 4 (15%) had titers >40 SGU or SMU, whereas none of the patients with positive anti-cardiolipin IgM (n = 17) had titers >40 MPL. Of the 11 children in the APS group, 9 had single-aPL positivity (LA, n = 5; aβ2GPI IgG/IgM, n = 4), 1 was dual-positive, and 1 patient had triple-positive aPL.

Patients in the APS group had a statistically significant higher risk of symptomatic recurrent VTE than those in the transient or low-titer aPL group (18% vs 1%; OR, 12.2; 95% CI, 1.4-108; P = .025) and a statistically nonsignificant but clinically meaningful difference in the risk of anticoagulant-related CRB (9% vs 0%; OR, 20.1; 95% CI, 0.7-558; P = .077).

This observational of the Kids-DOTT trial demonstrates positive aPL acutely in nearly one-quarter of children and young adults presenting with a first episode of provoked VTE, with ∼10% of these patients eventually meeting the criteria for APS. Patients fulfilling ISTH APS criteria appear to have a statistically significant and clinically meaningful increased risk of symptomatic recurrent VTE compared with patients with transient or persistent low-titer aPL. They also exhibited a nonstatistically significant albeit clinically meaningful increased risk of anticoagulant-associated CRB; however, the small sample size likely limited the power to detect a difference. Given that the overall number of events for both outcomes of interest was small, these findings should be interpreted with caution and warrant further investigation. Nevertheless, they have potentially important implications for the field of pediatric provoked VTE treatment, in that they suggest that current approaches to anticoagulation duration and/or intensity in children with acute provoked VTE and persistent aPL positivity may be suboptimal, with a potentially increased risk of symptomatic recurrent VTE in those who eventually meet the criteria for APS.

The incidence of aPL positivity (22%) at the time of an acute provoked VTE episode observed in our trial is similar to that reported in prior smaller retrospective studies.7 Most aPLs identified in our cohort were transient, low titer, and resolved within the first 6 to 12 weeks after VTE diagnosis. Despite being present at the time of an acute VTE episode, we suspect that these aPLs are immunologically different from aPLs present in APS, may be expressed as a result of the primary VTE-provoking condition, and are likely nonpathogenic. Our findings suggest specific characteristics that may help in the early identification of those children who will eventually meet APS criteria including older age, female sex, and non–catheter-associated VTE. These findings are consistent with those recently reported in a single-institution retrospective cohort study investigating risk factors for recurrent VTE in children with APS.13 

Although previous studies have shown that pediatric patients with VTE with APS have risks of recurrent and/or progressive thrombotic events as high as 19% to 29%, most of these studies have not distinguished between those with provoked and unprovoked index VTEs; yet, this is an important distinction, because children with unprovoked VTE in general have a higher risk of recurrence than those with provoked VTE.14,15 Adult data have indicated that the risk of recurrent VTE after a provoked VTE is similar between patients with aPL positivity and those with negative aPL, suggesting no significant clinical impact of aPL positivity.16,17 We report both a statistically significant and clinically meaningful increase in the risk of symptomatic recurrent VTE among pediatric patients with APS after an episode of provoked VTE, suggesting that (unlike in adults) aPL positivity may contribute to recurrence risk.

The strengths of this study include the prospective follow-up of a well-defined, multinational population of children and young adults with objectively diagnosed VTE and the systematic approach to the diagnosis of VTE outcomes using ISTH guidelines.10 A limitation of the study is the fact that decision on performing subsequent aPL testing after the 6-week post-VTE time point was left to the discretion of the treating hematologist, such that approximately one-third of patients with persistent aPL positivity at 6 weeks after VTE diagnosis did not undergo further aPL testing. Although this attrition could have potentially affected our findings, this risk is mitigated by our demonstration that the clinical characteristics of these patients were similar to those of the retained cohort. Notwithstanding this potential limitation, this observational analysis demonstrates that aPL positivity is frequent in patients aged <21 years with acute, first-episode, provoked VTE, and those who meet the criteria for APS appear to have an increased risk of symptomatic recurrent VTE with current conventional approaches to anticoagulation. Future collaborative studies should investigate the optimal management of young patients with provoked VTE who have persistent aPL.

The authors thank the patients and families for their generous participation in the Kids-DOTT multicenter trial. The authors thank the members of the Kids-DOTT Executive Steering Committee and the Kids-DOTT Trial Investigators.

The Kids-DOTT trial was supported by a U01 grant (1U01HL130048; Clinical Coordinating Center and Data Coordinating Center) from the National Institutes of Health (NIH), National Heart, Lung, and Blood Institute (NHLBI); and received support from an American Society of Hematology Bridge Grant, an Institutional Research Grant from the Johns Hopkins All Children's Foundation, an NIH NHLBI K23 award (1K23HL084055), and a Hemophilia and Thrombosis Research Society Thrombosis Studies Award.

Contribution: M.B. and N.A.G. designed the study; M.B., C.T., A.V., R.B., T.N., N.E.K., C.N., G.W., L.R.B., and N.A.G. performed patient enrollment and data collection; M.M. and E.A. performed data analyses; M.B. and N.A.G. performed the interpretation of findings and drafted the manuscript; and all authors provided critical revisions and approved the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

A complete list of the Kids-DOTT Trial investigators and Kids-DOTT Executive Steering Committee members appears in “supplemental Appendix.”

Correspondence: Marisol Betensky, Johns Hopkins All Children's Hospital, 501 6th Ave S, St. Petersburg, FL 33701; email: [email protected].

1.
Avcin
T
,
Ambrozic
A
,
Kuhar
M
,
Kveder
T
,
Rozman
B
.
Anticardiolipin and anti-beta(2) glycoprotein I antibodies in sera of 61 apparently healthy children at regular preventive visits
.
Rheumatology
.
2001
;
40
(
5
):
565
-
573
.
2.
Burk
CD
,
Miller
L
,
Handler
SD
,
Cohen
AR
.
Preoperative history and coagulation screening in children undergoing tonsillectomy
.
Pediatrics
.
1992
;
89
(
4 Pt 2
):
691
-
695
.
3.
Avcin
T
,
Toplak
N
.
Antiphospholipid antibodies in response to infection
.
Curr Rheumatol Rep
.
2007
;
9
(
3
):
212
-
218
.
4.
Angelini
L
,
Ravelli
A
,
Caporali
R
,
Rumi
V
,
Nardocci
N
,
Martini
A
.
High prevalence of antiphospholipid antibodies in children with idiopathic cerebral ischemia
.
Pediatrics
.
1994
;
94
(
4 Pt 1
):
500
-
503
.
5.
Manco-Johnson
MJ
,
Nuss
R
.
Lupus anticoagulant in children with thrombosis
.
Am J Hematol
.
1995
;
48
(
4
):
240
-
243
.
6.
Goldenberg
NA
,
Brandão
LR
,
Male
C
;
Kids-DOTT Trial Investigators
.
Anticoagulant therapy for 6 weeks vs 3 months and recurrence and bleeding in patients younger than 21 years with provoked venous thromboembolism-reply
.
JAMA
.
2022
;
327
(
17
):
1709
-
1710
.
7.
Goldenberg
NA
,
Tripputi
M
,
Crowther
M
, et al
.
The "parallel-cohort RCT": Novel design aspects and application in the Kids-DOTT trial of pediatric venous thromboembolism
.
Contemp Clin Trials
.
2010
;
31
(
1
):
131
-
133
.
8.
Monagle
P
,
Cuello
CA
,
Augustine
C
, et al
.
American Society of Hematology 2018 guidelines for management of venous thromboembolism: treatment of pediatric venous thromboembolism
.
Blood Adv
.
2018
;
2
(
22
):
3292
-
3316
.
9.
Monagle
P
,
Chan
AKC
,
Goldenberg
NA
, et al
.
Antithrombotic therapy in neonates and children: antithrombotic therapy and prevention of thrombosis, 9th ed: American College of Chest Physicians evidence-based clinical practice guidelines
.
Chest
.
2012
;
141
(
suppl 2
):
e737S
-
e801S
.
10.
Mitchell
LG
,
Goldenberg
NA
,
Male
C
,
Kenet
G
,
Monagle
P
,
Nowak-Göttl
U
;
Perinatal and Paediatric Haemostasis Subcommittee of the SSC of the ISTH
.
Perinatal and Paediatric Haemostasis Subcommittee of the SSC of the ISTH. Definition of clinical efficacy and safety outcomes for clinical trials in deep venous thrombosis and pulmonary embolism in children
.
J Thromb Haemost
.
2011
;
9
(
9
):
1856
-
1858
.
11.
Tripodi
A
,
Cohen
H
,
Devreese
KMJ
.
Lupus anticoagulant detection in anticoagulated patients. Guidance from the Scientific and Standardization Committee for lupus anticoagulant/antiphospholipid antibodies of the International Society on Thrombosis and Haemostasis
.
J Thromb Haemost
.
2020
;
18
(
7
):
1569
-
1571575
.
12.
Miyakis
S
,
Lockshin
MD
,
Atsumi
T
, et al
.
International consensus statement on an update of the classification criteria for definite antiphospholipid syndrome (APS)
.
J Thromb Haemost
.
2006
;
4
(
2
):
295
-
306
.
13.
Avila
ML
,
Amiri
N
,
Zavareh
ZT
, et al
.
Characterization of recurrent thrombosis in pediatric antiphospholipid syndrome
.
Am J Hematol
.
2022
;
97
(
7
):
E268
-
E270
.
14.
Berkun
Y
,
Padeh
S
,
Barash
J
, et al
.
Antiphsophoipid syndrome and recurrent thrombosis in children
.
Arthr Rheum
.
2006
;
55
(
6
):
850
-
855
.
15.
Avcin
T
,
Cimaz
R
,
Silverman
ED
, et al
.
Pediatric antiphospholipid syndrome: clinical and immunologic features of 121 patients in an international registry
.
Pediatric
.
2008
;
122
(
5
):
e1100
-
e1107
.
16.
Ortel
TL
,
Meleth
S
,
Catellier
D
, et al
.
Recurrent thrombosis in patients with antiphospholipid antibodies and an initial venous or arterial thromboembolic event: a systematic review and meta-analysis
.
J Thromb Haemost
.
2020
;
18
(
9
):
2274
-
2286
.
17.
Garcia
D
,
Erkan
D
.
Diagnosis and management of the antiphospholipid syndrome
.
N Engl J Med
.
2018
;
378
(
21
):
2010
-
2021
.

Author notes

Data are available on the request from the corresponding author, Marisol Betensky ([email protected]).

The full-text version of this article contains a data supplement.

Supplemental data