Analysis of factors contributing to diet quality outcomes
Of the 146 participants, 114 had a BMI z-score and were included in the cluster analysis. Of these, 55 were CCS and 59 were age-matched controls. The cluster analysis used four factors: child BMI z-score, ACARFS score, child emotional overeating score and child picky eating score, and the generated hierarchy was accepted at the three-cluster level. The allocation of healthy controls and CCS to these three groups is shown in Table 6, with all three clusters being a mixture of participants.
The first cluster (normal eating behaviour) was characterised by children with a healthy BMI z-score, good diet quality and low altered eating behaviours (picky eating/emotional overeating) and included 15 (27%) CCS and 29 (49%) controls. The second cluster (low BMI z-score, altered eating behaviour and poor diet quality) had the smallest number of participants with 8 (15%) CCS and 4 (8%) control. This cluster was characterised by children with a low (underweight) BMI z-score, poor diet quality and altered eating behaviour, particularly picky eating. Cluster 3 (high BMI z-score, altered eating behaviour and poor diet quality) was characterised by children with a high BMI z-score (trending to overweight), poor diet quality and altered eating behaviour, particularly overeating (32 (58%) CCS and 26 (44%) controls (TABLE 6).
Further analysis of the survivor group was undertaken to describe the treatment received by each child in each group. Children who had Non-Hodgkin’s lymphoma and Wilms’ tumour were found to have the highest proportion of CCS in cluster 3, with brain tumour survivors having the highest proportion of children in cluster 2 (Supplementary Table 1).
Discussion
This study revealed that CCSs consume diets of poorer quality and have higher rates of picky eating than their age-matched peers in the control group. It also appears that CCS eating behaviour and diet quality may be influenced by treatment intensity and the child’s cancer diagnosis.
Both the CCS group and the control group were found to have poor dietary intake for vegetables, grains, meat protein and dairy, affecting overall diet quality. These findings are in line with global patterns of dietary intake in CCSs. Zhang et al (2015) reported young childhood cancer survivors had poor adherence to the US dietary guidelines and consequently poor diet quality. This poor diet quality of CCS is of concern, as CCS are at a higher risk for early mortality and CVD than their peers (16). There is also a known association between poor dietary intake and many chronic diseases such as cardiovascular disease, metabolic syndrome and type 2 diabetes in the general population(25).
This evidence of poor adherence to dietary guidelines in CCS needs to be addressed in young cancer survivors. The finding of low vegetable intake for childhood cancer survivors is of particular concern due to the importance of folic acid intake and homocysteine levels(17,26). Decreased folic acid intake through a reduced intake of green leafy vegetables can result in an increase in homocysteine levels which are associated with endothelial dysfunction, increased low-density lipoprotein levels and ultimately atherosclerosis which can lead to an increased risk of CVD(17, 26).
This is the first study to demonstrate higher rates of picky eating in CCS than their peers, especially in CCS diagnosed at a young age. Published research into the general population has demonstrated the negative impact of a disrupted learning process for eating behaviour, which may then lead to the development of eating behaviours such as picky eating (2-4). The knowledge that children often receive cancer treatment during this critical eating development phase disrupting the learning process, suggests that cancer therapy may have a lasting effect on a child’s eating behaviour after treatment completion. In the general population children with picky eating consume less grains, fruits and vegetables than non-picky children resulting in low levels of vitamin E, vitamin C, folate and fibre, greatly impacting their diet quality(32, 33). Eating behaviours such as picky eating in CCS, are also likely to result in limited dietary intake and poor diet quality. Considering the importance of a healthy diet to reduce the risk of CVD and MetS(34), education focusing on eating behaviours needs to be considered.
Our study found several factors that may drive the diet quality of CCS. Diet quality, picky eating, overeating and BMI z-score were each represented in the three distinct clusters, suggesting they all may have a role in body size outcomes after treatment completion. In addition, comparison of cancer diagnostic classification with cluster group identification, indicated a distinct difference between the diagnosis of the child and the cluster they were allocated. This suggests that diagnosis groups which received single treatment modality regimens, such as Non-Hodgkin’s lymphoma and Wilms’ tumour, may be more likely to have a higher BMI z-score, altered eating behaviour and poor diet quality. These results may indicate an influence of treatment regimen on the four underlying variables that drive the cluster groups. Diagnosis groups such as brain tumour survivors that can receive more than one treatment modality (chemotherapy and radiation) might be more likely to have a lower BMI z-score, altered eating behaviour and poor diet quality. The findings are consistent with previous studies that have shown an increase in energy intake and BMI of Acute Lymphatic Leukaemia survivors who are considered to have a less intensive treatment regimen than other diagnostic categories(35). The identification of this group through the use of cluster analysis provides a possible indication of which cancer diagnostic classification groups might be more likely to develop altered eating behaviour problems, poor diet quality and altered body size, allowing for targeted nutritional intervention after cancer treatment.
Limitations
This study was limited by a low response rate from CCS reducing the generalisability of results and allowing for potential bias within the small sample. However, low response rates are not unusual in the childhood cancer study population (36). The small and heterogeneous nature of the sample may also have limited our ability to detect significant relationships between certain variables. Secondly, the reliance on parent-reports of their own and their child’s height and weight resulted in a high proportion (22%) of missing and possibly inaccurate data for this variable. In addition, the use of BMI z-scores to represent nutritional status is not ideal and prone to error in children with cancer, with BMI being a poor representation of body composition (37). Finally, the study design involved parent reported measures for diet and feeding variables. Despite being validated tools, the potential for reporter bias is high, especially for diet intake of the child and a parent’s feeding behaviour with the child. This may have resulted in an over reporting of diet variety and under-reporting of negative parent feeding behaviours. The use of more rigorous dietary intake measures such as a weighed food diary might be required for a more accurate measure of dietary intake in future studies.
Conclusion
This study found that CCS who have recently completed cancer treatment have a poor diet quality compared to age matched controls and have high rates of picky eating. Given that dietary habits established early in life are known to transpire into adulthood, and poor eating habits are linked to increased risk of Mets and CVD, young children who have recently completed cancer treatment with apparent poor dietary quality and altered eating behaviour are in critical need of dietary intervention early after cancer treatment. A tailored intervention that addresses altered eating behaviours, parent feeding practices and how to manage altered dietary intake due to treatment side effects is needed to bring about sustainable change for children early after cancer treatment.