Self-reported late effects and long-term follow-up care among 1889 long-term Norwegian Childhood, Adolescent, and Young Adult Cancer Survivors (the NOR-CAYACS study)

Participants eligible for the Norwegian CAYACS study (NOR-CAYACS) were identified through the CRN [32]. We included survivors who were diagnosed between 1985 and 2009, with any childhood cancer (CC, excluding central nervous system tumors) at ages 0–18 years or at the age 19–39 years with one of the following diagnosis: breast cancer (BC, ICD-10: C50, stage ≤ III), colorectal cancer (CRC, ICD-10: C18–20), non-Hodgkin lymphoma (NHL, ICD-10: C82–85), leukemias (LEUK, ICD-10: C91–96), and a random subsample of malignant melanomas (MM, ICD-10: 43). These diagnostic groups were chosen on the background of relatively good prognosis, expected risk of late effects, and lack of participation in other studies at our department. Other common AYA diagnostic groups, such as Hodgkin lymphoma and cervical and testicular cancer, were not included due to inclusion in other ongoing studies at our department, to avoid double publications and participation exhaustion. We excluded survivors who in the past 5 years preceding our study were diagnosed with a new cancer diagnosis as registered in the CRN, and/or those who self-reported relapse, and survivors self-reporting currently receiving oncological treatment. Eligible survivors were mailed an invitation to participate from the Oslo University Hospital including study information, an informed consent form, the questionnaire, and a pre-addressed pre-paid return envelope. No incentives to participate were offered. A reminder was sent by mail to survivors who had not responded after 5 months (more details about the NOR-CAYACS questionnaire are described in a previous paper [32]).

We used descriptive statistics with numbers and proportions or means with standard deviations (SD) to present the characteristics of the participating survivors. To compare responders with non-responders, univariable logistic regression models with response status as outcome were used.

For aim 1, we used numbers and proportions to describe self-reported late effects overall and by cancer diagnosis and treatment. Chi-squared statistics was used to compare the late effects across the cancer types. For aim 2, we used numbers and proportions to describe follow-up care, stratified by cancer diagnosis, treatment, and number of late effects. For aim 3, we used univariable and multivariable logistic regression models with having received/receiving follow-up care as outcome and clinical and sociodemographic factors as exposure variables. For this analysis, we only included survivors who reported at least one late effect. Factors associated with the outcome in the univariable model with a p value < 0.05 were entered in the multivariable model. Gender and age at survey were kept a priori in the multivariable model. Because of multicollinearity between diagnosis and treatment, only treatment was included in the final model. A p value < 0.05 was considered significant and all tests were two sided. All analyses were performed in SPSS (IBM SPSS Statistics 24).

The study was granted concession by The Norwegian Data Protection Authority (15/00395-2/CGN) and approved by the Regional Committee for Medical Research Ethics (2015/232 REK sør-øst B), and the Data Protection Officer at Oslo University Hospital and the Norwegian Cancer Registry.

Of the 5361 survivors invited, 2104 (39%) responded and 1889 (35%) were included in the current study (Fig. 1). In brief, responders were more likely to be females, slightly older at diagnosis and survey, longer time since diagnosis, and be diagnosed with BC, CRC, NHL, or LEUK than non-responders (data provided in Supplementary Table 1). This is in line with the non-response bias analysis done previously in a larger sample from the NOR-CAYACS study [32].

The sample consisted of 31% CC, 26% BC, 8% CRC, 12% NHL, 7% LEUK, and 16% MM survivors (Table 1). Two-thirds of the participants were female (68%) and the majority (57%) had higher education. Multimodal treatments were received by 58% of the survivors. Mean age at survey was 30 years among the CC and 49 years among the other diagnostic groups, diagnosed on average at the age of 11 and 33 years, with a mean of 19 and 15 years since diagnosis, respectively. Nine percent (n = 171) of the survivors had experienced a relapse before 2010 (5 years prior to study inclusion).

A wide range of late effects were reported, with 61.5% reporting at least one late effect (median 2.4, 0–14) (Table 2). The most commonly reported late effects were concentration and memory problems (28.1%), fatigue (25.2%), psychological reactions (22.2%), and hormonal changes (20.5%), but they differed by cancer type (Table 2, Fig. 2a; p = < 0.001). When categorized, 23% reported 1–2 late effects, 18% 3–4 late effects, and 20% reported 5 or more late effects. Among the MM survivors, 15% reported at least one late effect compared with 81% of BC survivors (p = < 0.001) (Fig. 2a). Survivors after BC and NHL reported the highest number of late effects, of which 30% BC and 33% NHL reported 5 or more late effects.

Among the survivors who had received multimodal treatments, 55% reported 3 or more late effects compared with 30% for systemic treatment, 11% for local treatment, and 2% for minimal surgery (Fig. 2b; p = < 0.001).

About two-thirds of all survivors (69%) reported not having received/not receiving follow-up care for late effects, while 23% had received/were receiving such care and 8% were unsure (Fig. 2c). These proportions differed across diagnosis (p = < 0.001). Lowest proportions of receiving follow-up care were reported among the MM (3%) and CRC group (16%) and highest among the CC, BC, LEUK, and NHL group (26–31%).

Treatment intensity was positively associated with receiving follow-up care. Compared with the other groups, survivors treated with multimodal treatments reported the highest proportions of receiving or having received follow-up care for late effects (Fig. 2d; p = < 0.001). Similarly, increasing number of experienced late effects was associated with increasing proportion of receiving or having received follow-up care (Fig. 2e; p = < 0.001).

The follow-up care for late effects was most often provided by hospital doctor (64%) (physician at treating hospital (28%), other medical specialists (36%)), general practitioner (44%), and community cancer nurse (0.2%).

A major strength of our study is the unselected, large nationwide, population-based cohort of long-term survivors representing a range of diagnostic groups. The high quality of the CRN data and the use of unique national identifying numbers in Norway allowed us to track and contact the whole population of eligible survivors, including those who may have been lost and disengaged from care. As such, these data provide a unique “snap-shot” of generally hard to reach populations.

There are several limitations. First, the results are based on self-report by the survivors and do not always reflect objectively verified late effects. Even though the wording of the question specified the conditions as late effects after cancer treatment, there is a risk that health problems not directly related to the cancer trajectory might have been attributed as late effects. However, other studies on the occurrence of late effects among childhood cancer survivors have also been based on self-report [2, 7]. Second, although our modest response rate might rise concerns of non-response bias risk, we found limited evidence of such in the same cohort [32]. There might be a concern that survivors who experience late effects are more inclined to answer the questionnaire which could lead to an overestimation of the prevalence of late effects. Based on the demographic and clinical information provided by the CRN for the whole population, we found no risk for an overrepresentation of survivors with fewer or more late effects than the non-responders in the sample using an inverse probability weighted model of participation [32]. Third, the cross-sectional design does not allow for causal relationships to be investigated. Fourth, using number of late effects as a proxy for the overall burden of late effects is a crude measure because the late effects are not necessarily mutually exclusive, and they were not weighted for severity. However, to construct a list of mutually exclusive late effects which can be subjected to appropriate severity weighting requires a level of details and length of questionnaire beyond the scope of the NOR-CAYACS study. Although crude, this proxy measure of late effect burden correlates well with treatment intensity, a well-known risk factor for late effects, adding to its face validity. Fifth, we only assessed “follow-up care specifically for late effects” in addition to routine follow-up care in one question. The distinction regarding these types of care was explained in the questionnaire. As the aim was to assess how many had received such extra care, distinguishing between such care received in the past and care currently received was considered not to be essential, albeit represents a minor limitation of the study. Last, our numbers for new cancers as late effects may be under-reported given survivors with any new diagnoses of cancers in the last 5 years prior to the study were excluded.

There is currently a gap between survivors’ experience of long-term follow-up care as recommended in international literature and guidelines. Survivors with low educational levels seem to be at higher risk of not receiving adequate follow-up care compared with survivors with higher education. Our data indicate that long-term CAYACS at risk of developing late effects would benefit from a formal model of follow-up care to ensure they receive recommended follow-up care.