Risk factors for overweight and obesity after childhood acute lymphoblastic leukemia in North America and Switzerland: A comparison of two cohort studies

Abstract Background After childhood acute lymphoblastic leukemia (ALL), sequelae include overweight and obesity, yet with conflicting evidence. We compared the prevalence of overweight and obesity between ≥5‐year ALL survivors from the North American Childhood Cancer Survivor Study (CCSS) and the Swiss Childhood Cancer Survivor Study (SCCSS) and described risk factors. Methods We included adult childhood ALL survivors diagnosed between 1976 and 1999. We matched CCSS participants (3:1) to SCCSS participants by sex and attained age. We calculated body mass index (BMI) from self‐reported height and weight for 1287 CCSS and 429 SCCSS participants; we then compared those with siblings (2034) in North America and Switzerland (678) siblings. We assessed risk factors for overweight (BMI 25–29.9 kg/m2) and obesity (≥30 kg/m2) using multinomial regression. Results We found overweight and obesity significantly more common among survivors in North America when compared with survivors in Switzerland [overweight: 30%, 95% confidence interval (CI): 27–32 vs. 24%, 21–29; obesity: 29%, 27–32 vs. 7%, 5–10] and siblings (overweight: 30%, 27–32 vs. 25%, 22–29; obesity: 24%, 22–26 vs. 6%, 4–8). Survivors in North America [odds ratio (OR) = 1.24, 1.01–1.53] and Switzerland (1.27, 0.74–2.21) were slightly more often obese than siblings. Among survivors, risk factors for obesity included residency in North America (5.8, 3.7–9.0); male (1.7, 1.3–2.3); attained age (≥45 years: 5.1, 2.4–10.8); Non‐Hispanic Black (3.4, 1.6–7.0); low household income (2.3, 1.4–3.5); young age at diagnosis (1.6, 1.1–2.2). Cranial radiotherapy ≥18 Gray was only a risk factor for overweight (1.4, 1.0–1.8); steroids were not associated with overweight or obesity. Interaction tests found no evidence of difference in risk factors between cohorts. Conclusions Although treatment‐related risk for overweight and obesity were similar between regions, higher prevalence among survivors in North America identifies important sociodemographic drivers for informing health policy and targeted intervention trials.


| INTRODUCTION
Acute lymphoblastic leukemia (ALL) is the most common childhood cancer, comprising 25% of all childhood cancer diagnoses. 1Survival has improved in many high-income countries to over 90% in recent years, 2 resulting in a growing population of childhood ALL survivors at risk of late effects. 3,4Among these, overweight and obesity are of particular concern since they potentiate the risk for cardiovascular disease 5 -the most common non-malignant cause of death among childhood cancer survivors. 6][9][10][11] Cranial radiation therapy (CRT) has been described as a risk factor for obesity in North America 8,9,[12][13][14] and Switzerland, 7,15 particularly among females treated at young age (0-4 years).
14]16,17 Therefore, comparison across studies and countries is complex.We hypothesize geographically specific factors driving overweight and obesity, necessitating different strategies to identify survivors at risk for weight problems and introduce interventions early in follow-up care.Therefore, we analyzed linked data from the North American Childhood Cancer Survivor Study (CCSS) and the Swiss Childhood Cancer Survivor Study (SCCSS) to determine whether the prevalence of overweight or obesity differs and to identify geographically specific risk factors for overweight and obesity.and malignant solid tumors.Questionnaires from CCSS (https://ccss.stjude.org)and SCCSS (https://www.swiss-ccss.ch)include similar questions about health outcomes and explanatory variables. 18,19CSS is a study of 31 institutions in the United States and Canada of 5-year survivors of childhood cancer diagnosed between 1970 and 1999. 19For our analyses, we included survivors of childhood ALL diagnosed between 1976 and 1999, aged ≥18 at time of follow-up who completed either baseline (1992-2001), expansion baseline (2002-2017), follow-up 2 (2001-2005), follow-up 4  (2007-2009), or follow-up 5 (2014-2016) questionnaires who provided informed consent.Although we refer to CCSS as a North American cohort, for our analysis survivors from Canada represented only 3% (44/1287).However, since the prevalence of overweight and obesity was similar between survivors in the United States and Canada, we did not exclude survivors in Canada (Table S1).CCSS is registered at Clinc ialTr ials.gov(identifier: NCT01120353) and approved by relevant institutional review boards.We list participating institutions in Supplemental Methods.
SCCSS is a population-based study of all children diagnosed with cancer in Switzerland.All children are treated in one of nine pediatric oncology-hematology centers and registered in the Swiss Childhood Cancer Registry (www.childhoodc ancer regis try.ch). 18The registry includes all children and adolescents diagnosed with cancer prior to age <21 years in Switzerland since 1976. 20It has a high case ascertainment of >95% for individuals diagnosed younger than age 16. 21We included ALL survivors diagnosed between 1976 and 1999, aged ≥18 at the time of follow-up with completed baseline 1 (2007-2013), baseline 2 (2015-2016), or follow-up 1 (2017) questionnaires who provided informed consent.SCCSS is registered at Clinc ialTr ials.gov(identifier: NCT03297034).Ethical approval was granted by the ethics committee of the canton of Bern, Switzerland (KEK-BE: 166/2014 and 2021-01462).We list participating institutions in Supplemental Methods.

| Sibling comparison group
We used siblings as a comparison group.During CCSS and SCCSS baseline questionnaire collection, survivor participants were asked for consent to contact siblings and their contact information.For CCSS, a random selection of nearest-age siblings were identified.For SCCSS, all siblings of survivor participants were contacted.Siblings received the same questionnaires as survivors without questions about cancer history.

| Outcome
We collected self-reported body weight without clothes and height without shoes from the most recent questionnaire and calculated body mass index (BMI) by dividing weight by height in meters squared (kg/m 2 ).We used BMI as a continuous and categorical variable with the following cutoffs: underweight (<18.5 kg/m 2 ); normal weight (18.5-24.9kg/m 2 ); overweight (25-29.9kg/m 2 ); obesity (≥30 kg/m 2 ). 22

| Statistical analysis
We weighted siblings so they became representative of survivors regarding the distribution of key sociodemographic variables (sex, attained age, and ethnicity).We fitted a logistic regression with survivorship status (survivor vs. sibling) as the outcome and the key sociodemographic variables as predictors.We calculated analysis weights for siblings as the inverse probability of being a survivor estimated from this regression.We matched ALL survivors in North America and Switzerland based on sex (exact) and attained age (±2 years) on a 3:1 ratio.We matched siblings in North America with siblings in Switzerland in the same way.We analyzed CCSS and SCCSS datasets separately for the comparison between ALL survivors and siblings.We handled missing values with multiple imputation by chained equations assuming missing at random, 24 generating 10 imputed datasets, and pooling the results according to the Rubin's rules. 25e used univariable and multivariable multinomial logistic regressions (BMI categories) to identify factors associated with overweight and obesity in the pooled CCSS and SCCSS datasets and performed interaction tests to see whether effect estimates of risk factors differed between cohorts.Since BMI is a continuous rather than categorical trait, we ran sensitivity analyses investigating factors associated with BMI in a multivariable linear regression (BMI continuous).We used STATA software (version 16, Stata Corporation) for all analyses.

| Study population
We included 1287 ALL survivors in North America (1243 United States; 44 Canada) and 429 from Switzerland; and 2034 siblings in North America and 678 in Switzerland (Figure S1).The mean attained age was 30.5 years (standard deviation [SD] 7.6 years) for ALL survivors of both cohorts after matching (Table 1; Tables S2 and S3).The mean age at diagnosis was 7.5 years (SD 4.8) for participants in North America and 6.2 years (SD 4.0) for participants from Switzerland (Table 2).More participants from North America reported receiving CRT ≥18 Gy than from Switzerland (38% vs. 14%).
In the two cohorts, we found associations with demographic, socioeconomic, and clinical factors followed the same direction and of comparable strength (all p-values for interactions ≥0.05), suggesting drivers of obesity are the same in North America and Switzerland (Table S8).The only difference was the association with CRT, which was weaker in North America (OR = 1.14, 95% CI: 0.82-1.58)than in Switzerland (OR = 3.10, 95% CI: 1.08-8.89)(Table S6).The number of survivors with obesity who received ≥18 Gy CRT was small (11 of 30 in Switzerland and 156 out of 375 in North America (Table S4).
The multivariable linear regression models-which modeled BMI as a continuous outcome-identified the same predictors as the logistic regression models.In fact, associations were stronger, in particular for young age at diagnosis, physical activity, and CRT (Table S7).

| DISCUSSION
Our collaborative analysis found overweight and obesity more common among survivors of childhood ALL and their siblings in North America than Switzerland.Although our results demonstrate an increased risk for obesity among survivors compared with siblings, the main risk factors for overweight and obesity in both cohorts were sociodemographic-not treatment-related-factors.
7][28][29][30] Our findings contribute to understanding the differential development of obesity globally after ALL treatment during childhood.First, we found evidence of a treatment exposure role.4,17 However, such finding is largely of historical interest since CRT is rarely used in contemporary treatment of ALL, and the effect size was relatively small.Second, we found risk for obesity was higher among children diagnosed and treated for ALL at young age (0-4 years). It is posible chemotherapy and radiotherapy effects on growing bodies are stronger during this developmental window.Alternatively-and perhaps more likely-the long, intensive treatment possibly disturbs the development of individual patterns of physical activity occurring at this age.For two or more years, children diagnosed with ALL are repeatedly hospitalized, receive chemotherapy, and experience restrictions of social contacts to reduce the risk of infection.So, opportunities to socialize, play, and run around with peers are limited and parents may be overprotective; it is possibly more relevant during preschool when activity patterns are consolidated.31,32 Physical activity habits acquired during early childhood strongly track throughout life.33 For children diagnosed later, it might be easier to return to previous physical activity and lifestyle habits after treatment ends, while children diagnosed when preschoolers likely do not remember a physical activity period from before.In our study, we found no evidence that treatment with steroids was a strong predictor of overweight and obesity in longterm survivors of childhood ALL, as shown before within the SCCSS.15 During and shortly after treatment, exposure to steroids can lead to overweight and obesity, but in the long-term we found that sociodemographic factors seem to play a more important role.However, within our study, we were not able to compare different doses of steroids, so we cannot exclude effects of higher doses of steroids.The dramatic differences between the prevalence of obesity between participants from North Americasurvivors and siblings-and participants from Switzerland shows how strongly lifestyle factors, such as diet and physical activity, influence overall risk.In the United States general population, about 30% of adults aged 20-39 are obese 34 ; in Switzerland, the proportion is 11%.7,35 Our findings for survivors with obesity-29% in North America and 7% in Switzerland-is comparable.It means the risk of survivors developing obesity is not an unavoidable result of cancer treatment.Instead, it reflects the general population and suggests it is potentially avoidable with health policy changes and early lifestyle interventions at the population level.In fact, our study shows household income and socioeconomic factors affecting physical activity and diet-with a much stronger impact on overweight than cancer treatment when comparing survivors with siblings, which was consistent across cohorts.
The associations found with age reflect findings from the general population. 34,35In both cohorts, males showed higher risk for overweight and obesity than females.Although it contrasts with previous CCSS publications, 8,9,11,12,36 it aligns with an SCCSS study 7 and findings from general Swiss and Non-Hispanic, White US populations. 34,35A meta-analysis of 10 studies from the United States, The Netherlands, and the United Kingdom found no risk difference by sex. 10 Thus the higher risk for males with ALL likely reflects the higher background risk in the general population of Switzerland and North America. 34,35he comparability of the two cohorts demonstrates the main strength of our study, which share the same design, methodology, and questionnaires.Inclusion criteriaera of cancer diagnosis (1976-1999), sex, and attained age-were identical and prevented biases and difficulties other studies encountered when interpreting results.Our study also has limitations.Since participants self-reported height and weight in both cohorts, survivors possibly underestimated weight or overestimated height due to social desirability bias.However, we expect the degree of error of BMI assessment to be non-differential, that is, similar for survivors and siblings in both cohorts.Cumulative doses of CRT were assessed by an intention-to-treat approach for SCCSS and by detailed dosimetry in CCSS. 37Thus, the measurement error for CRT was larger for survivors in Switzerland, which possibly attenuated effect estimates.Furthermore, we could not investigate risk factors for overweight since both cohorts lacked information, namely physical functioning, BMI before or at diagnosis, parental BMI, and genetic factors. 13,14,38

| CONCLUSIONS
Overall, our study demonstrated overweight, and obesity are highly prevalent and thus a major health concern with   5 which is deleterious for childhood ALL survivors with already high burdens of cardiovascular disease and related mortality from cancer and treatment with cardiotoxic anthracycline chemotherapy. 6,39e recommend identifying survivors at risk for cardiovascular disease early and offering support with targeted interventions. 40,415][46][47] Some children need special support because of musculoskeletal or medical impairments. 48,49Our findings further suggest that introduction of physical activity-as a regular supportive treatment during acute cancer treatment-should be considered as interventions for evaluation among preschoolers since activity patterns acquired early track during the lifetime. 31,32The German Network ActiveOncoKids implements such a physical activity program as usual care for pediatric and adolescent patients during and after cancer treatment.Sports scientists offer individualized trainings on a daily basis in addition to medical services, such as physiotherapy. 50besity is multi-factorial in origin and survivors of ALL share most risk factors with the general population.Policy measures and structural changes aiming to reduce obesity in the population at large also automatically improve the situation for ALL survivors.Such measures possibly include more physical activity lessons at school, availability of safe outdoor spaces for physical activity, appropriate food labeling, access to healthy food choices, sugar taxes, close monitoring of weight trajectory and nutrition, and physical and behavioral counseling by pediatricians and general practitioners.
Our study confirmed strong effects of socioeconomic status on overweight, namely income and ethnicity.Thus, national policies to reduce social differentials in health and income in Switzerland and North America are essential for improving the situation.An international comparison of socioeconomic inequalities in adolescent health among 34 North American and European countries found inequalities between socioeconomic groups increased over time and confirmed physical activity levels and health as related to average per-person income and income inequality within a country. 28ur study confirmed obesity as a prevalent health hazard among survivors of ALL and their siblings mainly driven by sociodemographic factors, like in the general population.Since cancer survivors are particularly susceptible to cardiovascular disease and other late sequelae of overweight and obesity, we recommend a two-pronged approach: health policies for reducing overweight among the general population and interventions targeting physical activity and diet during and after cancer treatment.

T A B L E 3
Overweight and obesity among ALL survivors compared with siblings from CCSS (North America) and SCCSS (Switzerland) (referent: normal BMI): a multinomial logistic regression analysis.
Demographic, socioeconomic, and lifestyle characteristics of ALL survivors and siblings comparing CCSS (North America) with SCCSS (Switzerland).
T A B L E 1 Clinical (diagnosis and treatment) characteristics of ALL survivors comparing CCSS (North America) with SCCSS (Switzerland).
Abbreviations: ALL, acute lymphoblastic leukemia; BMI, body mass index; CCSS, Childhood Cancer Survivors Study; IQR, interquartile range; SCCSS, Swiss Childhood Cancer Survivors Study; SD, standard deviation; sd, standardized.The italic values give insights about the number of missing values for each characteristic, country, and population group (survivors vs siblings).aSiblingsstandardizedbysex,attainedage,and race/ethnicity to ALL survivors by cohort.bWematchedALLsurvivors/siblingsfromNorthAmerica with survivors/siblings in Switzerland on a 1:3 ratio based on sex and attained age.cHighestdegree of education level is categorized as lower than college graduate/post graduate level and college graduate/post graduate level.dHouseholdincome(income per year) is categorized as low: CCSS baseline (1992-2001): <$20,000, expansion baseline (2002-2017), follow-up 2 (2001-2005), follow-up 4 (2007-2009), and follow-up 5 (2014-2016):<$40,000, SCCSS: ≤54,000 Swiss francs; middle: CCSS: baseline: $20,000-60,000, other questionnaires: $40,000-100,000, SCCSS: 54,000-108.000Swissfrancs; and high: CCSS baseline: >$60,000, other questionnaires: >$100,000, SCCSS: >108,000 Swiss francs.eAlcoholconsumption is categorized as never/rarely; weekly, ≥1 standard drink/week; daily, 1 standard drink/day; frequently, >1 standard drink/day.fPhysicallyinactive is defined as fewer than 150 min of activity per week; physically active is defined as 150 min or more of moderate or 75 min of vigorous physical activity, or a combination of moderate and vigorous-intense physical activity per week.T A B L E 1 (Continued)T A B L E 2 Siblings are standardized by sex, attained age, and race/ethnicity to ALL survivors by cohort.Predictors for overweight and obesity among ALL survivors (retrieved from multivariable multinomial logistic regression a ; referent: normal BMI).
Abbreviationsa Adjusted for all variables listed.b c d e