scispace - formally typeset

Journal ArticleDOI

Predicting acute ovarian failure in female survivors of childhood cancer: a cohort study in the Childhood Cancer Survivor Study (CCSS) and the St Jude Lifetime Cohort (SJLIFE).

01 Mar 2020-Lancet Oncology (Elsevier)-Vol. 21, Iss: 3, pp 436-445

TL;DR: Both acute ovarian failure risk prediction models performed well and could help clinical discussions regarding the need for fertility preservation interventions in girls and young women newly diagnosed with cancer.
Abstract: Summary Background Cancer treatment can cause gonadal impairment. Acute ovarian failure is defined as the permanent loss of ovarian function within 5 years of cancer diagnosis. We aimed to develop and validate risk prediction tools to provide accurate clinical guidance for paediatric patients with cancer. Methods In this cohort study, prediction models of acute ovarian failure risk were developed using eligible female US and Canadian participants in the Childhood Cancer Survivor Study (CCSS) cohort and validated in the St Jude Lifetime Cohort (SJLIFE) Study. 5-year survivors from the CCSS cohort were included if they were at least 18 years old at their most recent follow-up and had complete treatment exposure and adequate menstrual history (including age at menarche, current menstrual status, age at last menstruation, and menopausal aetiology) information available. Participants in the SJLIFE cohort were at least 10-year survivors. Participants were excluded from the prediction analysis if they had an ovarian hormone deficiency, had missing exposure information, or had indeterminate ovarian status. The outcome of acute ovarian failure was defined as permanent loss of ovarian function within 5 years of cancer diagnosis or no menarche after cancer treatment by the age of 18 years. Logistic regression, random forest, and support vector machines were used as candidate methods to develop the risk prediction models in the CCSS cohort. Prediction performance was evaluated internally (in the CCSS cohort) and externally (in the SJLIFE cohort) using the areas under the receiver operating characteristic curve (AUC) and the precision-recall curve (average precision [AP; average positive predictive value]). Findings Data from the CCSS cohort were collected for participants followed up between Nov 3, 1992, and Nov 25, 2016, and from the SJLIFE cohort for participants followed up between Oct 17, 2007, and April 16, 2012. Of 11 336 female CCSS participants, 5886 (51·9%) met all inclusion criteria for analysis. 1644 participants were identified from the SJLIFE cohort, of whom 875 (53·2%) were eligible for analysis. 353 (6·0%) of analysed CCSS participants and 50 (5·7%) of analysed SJLIFE participants had acute ovarian failure. The overall median follow-up for the CCSS cohort was 23·9 years (IQR 20·4–27·9), and for SJLIFE it was 23·9 years (19·0–30·0). The three candidate methods (logistic regression, random forest, and support vector machines) yielded similar results, and a prescribed dose model with abdominal and pelvic radiation doses and an ovarian dose model with ovarian radiation dosimetry using logistic regression were selected. Common predictors in both models were history of haematopoietic stem-cell transplantation, cumulative alkylating drug dose, and an interaction between age at cancer diagnosis and haematopoietic stem-cell transplant. External validation of the model in the SJLIFE cohort produced an estimated AUC of 0·94 (95% CI 0·90–0·98) and AP of 0·68 (95% CI 0·53–0·81) for the ovarian dose model, and AUC of 0·96 (0·94–0·97) and AP of 0·46 (0·34–0·61) for the prescribed dose model. Based on these models, an online risk calculator has been developed for clinical use. Interpretation Both acute ovarian failure risk prediction models performed well. The ovarian dose model is preferred if ovarian radiation dosimetry is available. The models, along with the online risk calculator, could help clinical discussions regarding the need for fertility preservation interventions in girls and young women newly diagnosed with cancer. Funding Canadian Institutes of Health Research, Women and Children's Health Research Institute, National Cancer Institute, and American Lebanese Syrian Associated Charities.
Topics: Cohort (61%), Cohort study (60%), Fertility preservation (52%), Childhood Cancer Survivor Study (52%), Transplantation (50%)

Content maybe subject to copyright    Report

1
Predicting Acute Ovarian Failure in Female Childhood Cancer Survivors:
A Cohort Study in the Childhood Cancer Survivor Study (CCSS) and
the St. Jude Lifetime Cohort (SJLIFE)
Authorship List Line
Rebecca A. Clark, Sogol Mostoufi-Moab, Yutaka Yasui, Ngoc Khanh Vu, Charles A. Sklar
,
Tarek Motan, Russell J. Brooke, Todd M. Gibson, Kevin C. Oeffinger
, Rebecca M. Howell, Susan
A. Smith, Zhe Lu, Leslie L. Robison, Wassim Chemaitilly, Melissa M. Hudson
, Gregory T.
Armstrong, Paul C. Nathan*
, Yan Yuan*
.
Full professor
* Co-senior authors
Corresponding author
List of Affiliations
School of Public Health - University of Alberta, Edmonton, Alberta, Canada. (RA Clark MSc, NK
Vu PhD, Z Lu MEng, Y Yuan PhD)
University of Pennsylvania Perelman School of Medicine, The Children’s Hospital of
Philadelphia, Philadelphia, Pennsylvania, United States of America. (S Mostoufi-Moab MD)
St. Jude Children’s Research Hospital, Memphis, Tennessee, United States of America. (Y Yasui
PhD, RJ Brooke PhD, TM Gibson PhD, LL Robison PhD, W Chemaitilly MD, MM Hudson MD,
GT Armstrong MD)
Memorial Sloan Kettering Cancer Center, New York, New York, United States of America. (CA
Sklar MD)
Manuscript

2
Faculty of Medicine and Dentistry - University of Alberta, Edmonton, Alberta, Canada. (T Motan
MB ChB)
Duke University School of Medicine, Durham, North Carolina, United States of America. (KC
Oeffinger MD)
The University of Texas MD Anderson Cancer Center, Houston, Texas, United States of America.
(RM Howell PhD, SA Smith MPH)
The Hospital for Sick Children, Toronto, Ontario, Canada. (PC Nathan MD)
Corresponding Author Information
Yan Yuan, PhD
Associate Professor
yyuan@ualberta.ca
Tel: +1 (780) 248-5853
School of Public Health, University of Alberta
3-299 Edmonton Clinic Health Academy
11405 87 Avenue
Edmonton, Alberta
T6G 1C9
Canada
Word count: 4,123

3
Summary
Background
Cancer therapy can cause gonadal impairment. Acute ovarian failure (AOF) is defined as the
permanent loss of ovarian function within five years of cancer diagnosis. We aimed to develop
and validate risk prediction tools to provide accurate clinical guidance to paediatric cancer patients.
Methods
AOF risk prediction models were developed using eligible female participants in the Childhood
Cancer Survivor Study (CCSS) cohort and validated in the St. Jude Lifetime Cohort (SJLIFE).
Eligibility criteria were at least age 18, had complete treatment exposure and adequate menstrual
history information available. Logistic regression, random forest, and support vector machines
were used as candidate methods. Prediction performance was evaluated internally and externally
using the areas under the ROC curve (AUC) and the precision-recall curve (AP). An online risk
calculator was developed for clinical use.
Findings
Three-hundred and fifty-three (6%) of 5,886 CCSS participants and 50 (5.7%) of 875 SJLIFE
participants experienced AOF. The median follow-up for the CCSS and SJLIFE analysis samples
was 23·9 (IQR=20·4-27·9) and 23·9 (19·0-30·0) years, respectively. A prescribed dose model
with abdominal and pelvic radiation doses and an ovarian dose model with ovarian radiation
dosimetry using logistic regression were selected. Common predictors in both models were history
of hematopoietic stem cell transplantation (HSCT), cumulative alkylating agent dose, and an
interaction between age at cancer diagnosis and HSCT. External validation produced an estimated

4
AUC of 0·94 (95% CI=0·90-0·98) and AP of 0·68 (95% CI=0·53-0·81) for the ovarian dose
model, and AUC of 0·96 (0·94-0·97) and AP of 0·46 (0·34-0·61) for the prescribed dose model.
Interpretations
Both AOF risk prediction models perform very well. The ovarian model is preferred if ovarian
radiation dosimetry is available. The models, along with the online risk calculator, can aid clinical
discussions regarding the need for fertility preservation interventions in young females newly
diagnosed with cancer.
Funding
Canadian Institutes for Health Research, Women and Children’s Research Institute, National
Cancer Institute, American Lebanese Syrian Associated Charities.

5
Research in Context
Evidence before this study
An increased risk of premature gonadal failure has been demonstrated in paediatric cancer
survivors treated with chemotherapy and radiation. Six percent of female childhood cancer
survivors lose ovarian function within five years of treatment (acute ovarian failure [AOF]), and
an additional nine percent experience premature, non-surgical menopause before age 40. The time
frame between primary cancer diagnosis and treatment resulting in AOF is limited to identify high-
risk patients that will benefit from interventions aimed at fertility preservation. We searched
PubMed with no date or language restrictions for all studies to evaluate the current knowledge of
AOF and the associated risk factors in childhood cancer survivors using the search terms “pediatric
cancer OR childhood cancer” AND “acute ovarian failure OR primary ovarian insufficiency”
AND “risk”. Five publications were considered for further review as they described AOF as an
independent condition without grouping patients in a broader premature menopause umbrella.
While high dose pelvic radiation, hematopoietic stem cell transplantation, and alkylating
chemotherapy have been identified as risk factors associated with AOF, clinicians lack a tool that
accurately estimates the risk of AOF for individual paediatric cancer patients at the time of cancer
diagnosis. We did not find any study that aimed to develop risk estimates of AOF for individual
paediatric cancer patients at the time of cancer diagnosis.
Added value of this study
To our knowledge, we have developed and validated the first models for predicting the risk of
AOF in female childhood cancer survivors. While physicians are aware of the gonadotoxic
treatment exposures with a high likelihood of causing AOF, there are no available prediction tools

Citations
More filters

Posted Content
L. Yu1, Zhe Lu1, P. C. Nathan2, S. Mostoufi-Moab3  +1 moreInstitutions (3)
TL;DR: This work contextualizes the computational methods with the working example: the modelling of acute ovarian failure risk in female childhood cancer survivors to quantify the risk of permanent ovarian failure due to exposure to lifesaving but nonetheless toxic cancer treatments.
Abstract: Statistical and computational methods are widely used in today's scientific studies. Using a female fertility potential in childhood cancer survivors as an example, we illustrate how these methods can be used to extract insight regarding biological processes from noisy observational data in order to inform decision making. We start by contextualizing the computational methods with the working example: the modelling of acute ovarian failure risk in female childhood cancer survivors to quantify the risk of permanent ovarian failure due to exposure to lifesaving but nonetheless toxic cancer treatments. This is followed by a description of the general framework of classification problems. We provide an overview of the modelling algorithms employed in our example, including one classic model (logistic regression) and two popular modern learning methods (random forest and support vector machines). Using the working example, we show the general steps of data preparation for modelling, variable selection steps for the classic model, and how model performance might be improved utilizing visualization tools. We end with a note on the importance of model evaluation.

2 citations


Journal ArticleDOI
TL;DR: Recognising reduced endocrine sequelae to lower intensity treatments and potential future toxicities of new treatments are both key for future long-term follow up strategies.
Abstract: Endocrine late effects are common sequelae of the treatment of childhood cancer, and surveillance and management of these are integral to the follow-up of childhood cancer survivors. Cancer treatments are evolving to improve survival and minimize toxicities, and novel treatment modalities such as proton radiotherapy and immunotherapies are becoming available. Recognising reduced endocrine sequelae to lower intensity treatments and potential future toxicities of new treatments are both key for future long-term follow-up strategies. Cancer therapies that impact on puberty are chiefly those that affect the hypothalamic–pituitary axis including surgery and radiotherapy and those that affect gonadal function including surgery, gonadotoxic chemotherapy and radiotherapy to the gonads. This short review looks at the impact on puberty of current and novel treatment modalities.

2 citations


Journal ArticleDOI
Abstract: Survivors of pediatric cancer are at increased risk for infertility and premature hormonal failure. Surgeons caring for children with cancer have an important role to play in understanding this risk, as well as advocating for and performing appropriate fertility preservation procedures. Fertility preservation options in males and females vary by pubertal status and include nonexperimental (oocyte harvest, ovarian tissue cryopreservation, sperm cryopreservation) and experimental (testicular tissue cryopreservation) options. This review summarizes the basics of risk assessment and fertility preservation options and explores unique considerations in pediatric fertility preservation.

1 citations


Journal ArticleDOI
David H Noyd1, Nigel B. Neely2, Kristin Schroeder1, Paul M. Lantos1  +4 moreInstitutions (2)
Abstract: BACKGROUND This retrospective study harnessed an institutional cancer registry to construct a childhood cancer survivorship cohort, integrate electronic health record (EHR) and geospatial data to stratify survivors based on late-effect risk, analyze follow-up care patterns, and determine factors associated with suboptimal follow-up care. PROCEDURE The survivorship cohort included patients ≤18 years of age reported to the institutional cancer registry between January 1, 1994 and November 30, 2012. International Classification of Diseases for Oncology, third revision (ICD-O-3) coding and treatment exposures facilitated risk stratification of survivors. The EHR was linked to the cancer registry based on medical record number (MRN) to extract clinic visits. RESULTS Five hundred and ninety pediatric hematology-oncology (PHO) and 275 pediatric neuro-oncology (PNO) survivors were included in the final analytic cohort. Two hundred and eight-two survivors (32.6%) were not seen in any oncology-related subspecialty clinic at Duke 5-7 years after initial diagnosis. Factors associated with follow-up included age (p = .008), diagnosis (p < .001), race/ethnicity (p = .010), late-effect risk strata (p = .001), distance to treatment center (p < .0001), and area deprivation index (ADI) (p = .011). Multivariable logistic modeling attenuated the association for high-risk (OR 1.72; 95% CI 0.805, 3.66) and intermediate-risk (OR 1.23, 95% CI 0.644, 2.36) survivors compared to survivors at low risk of late effects among the PHO cohort. PNO survivors at high risk for late effects were more likely to follow up (adjusted OR 3.66; 95% CI 1.76, 7.61). CONCLUSIONS Nearly a third of survivors received suboptimal follow-up care. This study provides a reproducible model to integrate cancer registry and EHR data to construct risk-stratified survivorship cohorts to assess follow-up care.

1 citations


Posted Content
TL;DR: This article examines the analytical connections and differences between two IncV metrics: Incv in AUC (IncV-AUC) and IncV in AP ( IncV-AP), and compares them with a strictly proper scoring rule: the IncV of the scaled Brier score (incV-sBrS), via a numerical study.
Abstract: Incremental value (IncV) evaluates the performance change from an existing risk model to a new model. It is one of the key considerations in deciding whether a new risk model performs better than the existing one. Problems arise when different IncV metrics contradict each other. For example, compared with a prescribed-dose model, an ovarian-dose model for predicting acute ovarian failure has a slightly lower area under the receiver operating characteristic curve (AUC) but increases the area under the precision-recall curve (AP) by 48%. This phenomenon of conflicting conclusions is not uncommon, and it creates a dilemma in medical decision making. In this article, we examine the analytical connections and differences between two IncV metrics: IncV in AUC (IncV-AUC) and IncV in AP (IncV-AP). Additionally, since they are both semi-proper scoring rules, we compare them with a strictly proper scoring rule: the IncV of the scaled Brier score (IncV-sBrS), via a numerical study. We demonstrate that both IncV-AUC and IncV-AP are weighted averages of the changes (from the existing model to the new one) in separating the risk score distributions between events and non-events. However, IncV-AP assigns heavier weights to the changes in the high-risk group, whereas IncV-AUC weights the changes equally. In the numerical study, we find that IncV-AP has a wide range, from negative to positive, but the size of IncV-AUC is much smaller. In addition, IncV-AP and IncV-sBr Sare highly consistent, but IncV-AUC is negatively correlated with IncV-sBrS and IncV-AP at a low event rate. IncV-AUC and IncV-AP are the least consistent among the three pairs, and their differences are more pronounced as the event rate decreases.

Cites background or methods from "Predicting acute ovarian failure in..."

  • ...For example, a prescribed-dose model and an ovarian-dose model were compared in Clark et al. (2020) for predicting acute ovarian failure among female childhood cancer survivors....

    [...]

  • ...Clark et al. (2020) created four risk groups: low (< 5%), medium-low (5% to <20%), medium (20% to <50%), and high risk (≥ 50%)....

    [...]

  • ...We evaluate and compare two recently published risk models (Clark et al., 2020) that predict AOF on an external validation dataset, the St....

    [...]


References
More filters

Journal ArticleDOI
Leo Breiman1Institutions (1)
01 Oct 2001
TL;DR: Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the forest, and are also applicable to regression.
Abstract: Random forests are a combination of tree predictors such that each tree depends on the values of a random vector sampled independently and with the same distribution for all trees in the forest. The generalization error for forests converges a.s. to a limit as the number of trees in the forest becomes large. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation between them. Using a random selection of features to split each node yields error rates that compare favorably to Adaboost (Y. Freund & R. Schapire, Machine Learning: Proceedings of the Thirteenth International conference, aaa, 148–156), but are more robust with respect to noise. Internal estimates monitor error, strength, and correlation and these are used to show the response to increasing the number of features used in the splitting. Internal estimates are also used to measure variable importance. These ideas are also applicable to regression.

58,232 citations


Journal ArticleDOI
TL;DR: It is suggested that reporting discrimination and calibration will always be important for a prediction model and decision-analytic measures should be reported if the predictive model is to be used for clinical decisions.
Abstract: The performance of prediction models can be assessed using a variety of methods and metrics. Traditional measures for binary and survival outcomes include the Brier score to indicate overall model performance, the concordance (or c) statistic for discriminative ability (or area under the receiver operating characteristic [ROC] curve), and goodness-of-fit statistics for calibration.Several new measures have recently been proposed that can be seen as refinements of discrimination measures, including variants of the c statistic for survival, reclassification tables, net reclassification improvement (NRI), and integrated discrimination improvement (IDI). Moreover, decision-analytic measures have been proposed, including decision curves to plot the net benefit achieved by making decisions based on model predictions.We aimed to define the role of these relatively novel approaches in the evaluation of the performance of prediction models. For illustration, we present a case study of predicting the presence of residual tumor versus benign tissue in patients with testicular cancer (n = 544 for model development, n = 273 for external validation).We suggest that reporting discrimination and calibration will always be important for a prediction model. Decision-analytic measures should be reported if the predictive model is to be used for clinical decisions. Other measures of performance may be warranted in specific applications, such as reclassification metrics to gain insight into the value of adding a novel predictor to an established model.

2,704 citations


Journal ArticleDOI
12 Jun 2013-JAMA
TL;DR: Among adult survivors of childhood cancer, the prevalence of adverse health outcomes was high, and a systematic risk-based medical assessment identified a substantial number of previously undiagnosed problems that are more prevalent in an older population.
Abstract: Importance Adult survivors of childhood cancer are known to be at risk for treatment-related adverse health outcomes. A large population of survivors has not been evaluated using a comprehensive systematic clinical assessment to determine the prevalence of chronic health conditions. Objective To determine the prevalence of adverse health outcomes and the proportion associated with treatment-related exposures in a large cohort of adult survivors of childhood cancer. Design, Setting, and Participants Presence of health outcomes was ascertained using systematic exposure–based medical assessments among 1713 adult (median age, 32 [range, 18-60] years) survivors of childhood cancer (median time from diagnosis, 25 [range, 10-47] years) enrolled in the St Jude Lifetime Cohort Study since October 1, 2007, and undergoing follow-up through October 31, 2012. Main Outcomes and Measures Age-specific cumulative prevalence of adverse outcomes by organ system. Results Using clinical criteria, the crude prevalence of adverse health outcomes was highest for pulmonary (abnormal pulmonary function, 65.2% [95% CI, 60.4%-69.8%]), auditory (hearing loss, 62.1% [95% CI, 55.8%-68.2%]), endocrine or reproductive (any endocrine condition, such as hypothalamic-pituitary axis disorders and male germ cell dysfunction, 62.0% [95% CI, 59.5%-64.6%]), cardiac (any cardiac condition, such as heart valve disorders, 56.4% [95% CI, 53.5%-59.2%]), and neurocognitive (neurocognitive impairment, 48.0% [95% CI, 44.9%-51.0%]) function, whereas abnormalities involving hepatic (liver dysfunction, 13.0% [95% CI, 10.8%-15.3%]), skeletal (osteoporosis, 9.6% [95% CI, 8.0%-11.5%]), renal (kidney dysfunction, 5.0% [95% CI, 4.0%-6.3%]), and hematopoietic (abnormal blood cell counts, 3.0% [95% CI, 2.1%-3.9%]) function were less common. Among survivors at risk for adverse outcomes following specific cancer treatment modalities, the estimated cumulative prevalence at age 50 years was 21.6% (95% CI, 19.3%-23.9%) for cardiomyopathy, 83.5% (95% CI, 80.2%-86.8%) for heart valve disorder, 81.3% (95% CI, 77.6%-85.0%) for pulmonary dysfunction, 76.8% (95% CI, 73.6%-80.0%) for pituitary dysfunction, 86.5% (95% CI, 82.3%-90.7%) for hearing loss, 31.9% (95% CI, 28.0%-35.8%) for primary ovarian failure, 31.1% (95% CI, 27.3%-34.9%) for Leydig cell failure, and 40.9% (95% CI, 32.0%-49.8%) for breast cancer. At age 45 years, the estimated cumulative prevalence of any chronic health condition was 95.5% (95% CI, 94.8%-98.6%) and 80.5% (95% CI, 73.0%-86.6%) for a serious/disabling or life-threatening chronic condition. Conclusions and Relevance Among adult survivors of childhood cancer, the prevalence of adverse health outcomes was high, and a systematic risk-based medical assessment identified a substantial number of previously undiagnosed problems that are more prevalent in an older population. These findings underscore the importance of ongoing health monitoring for adults who survive childhood cancer.

772 citations


Journal ArticleDOI
Tara M. Cousineau1, Alice D. Domar1Institutions (1)
TL;DR: Evidence is emerging of an association between stress of fertility treatment and patient drop-out and pregnancy rates, and further research is needed to understand the association between distress and fertility outcome.
Abstract: The inability to conceive children is experienced as a stressful situation by individuals and couples all around the world. The consequences of infertility are manifold and can include societal repercussions and personal suffering. Advances in assisted reproductive technologies, such as IVF, can offer hope to many couples where treatment is available, although barriers exist in terms of medical coverage and affordability. The medicalization of infertility has unwittingly led to a disregard for the emotional responses that couples experience, which include distress, loss of control, stigmatization, and a disruption in the developmental trajectory of adulthood. Evidence is emerging of an association between stress of fertility treatment and patient drop-out and pregnancy rates. Fortunately, psychological interventions, especially those emphasizing stress management and coping-skills training, have been shown to have beneficial effects for infertility patients. Further research is needed to understand the association between distress and fertility outcome, as well as effective psychosocial interventions.

701 citations


Journal ArticleDOI
Abstract: Background Increased attention has been directed toward the long-term health outcomes of survivors of childhood cancer. To facilitate such research, a multi-institutional consortium established the Childhood Cancer Survivor Study (CCSS), a large, diverse, and well-characterized cohort of 5-year survivors of childhood and adolescent cancer. Procedure Eligibility for the CCSS cohort included a selected group of cancer diagnoses prior to age 21 years between 1970–1986 and survival for at least 5 years. Results A total of 20,276 eligible subjects were identified from the 25 contributing institutions, of whom 15% are considered lost to follow-up. Currently, 14,054 subjects (69.3% of the eligible cohort) have participated by completing a 24-page baseline questionnaire. The distribution of first diagnoses includes leukemia (33%), lymphoma (21%), neuroblastoma (7%), CNS tumor (13%), bone tumor (8%), kidney tumor (9%), and soft-tissue sarcoma (9%). Abstraction of medical records for chemotherapy, radiation therapy, and surgical procedures has been successfully completed for 98% of study participants. Overall, 78% received radiotherapy and 73% chemotherapy. Conclusion The CCSS represents the largest and most extensively characterized cohort of childhood and adolescent cancer survivors in North America. It serves as a resource for addressing important issues such as risk of second malignancies, endocrine and reproductive outcome, cardiopulmonary complications, and psychosocial implications, among this unique and ever-growing population. Med. Pediatr. Oncol. 2002;38:229–239. © 2002 Wiley-Liss, Inc.

577 citations


Performance
Metrics
No. of citations received by the Paper in previous years
YearCitations
20219
20206