Discussion
This analysis of participation rates in late effect studies after childhood cancer illustrates potential factors influencing participation rates. We included 879 survivors of childhood cancer characterized by universal access to tax-payed health care, low social inequality, access to population-based cohorts and complete follow-up using highly valid and complete registers by use of the civil registration number assigned to all citizens. We applied a systematic search strategy and sought additional information from the corresponding authors in case of missing data.
All studies reported SPR or provided information sufficient to calculate SPR. Three studies did not clarify study dropouts. Less than three studies had missing data on age distribution. Study populations ranged from 15-161 participants, and the outcome measures varied with diverse clinical examinations such as exercise programs, endocrinologic or audiologic evaluations, radiologic imaging and neurocognitive examinations. SPR ranged from 27-100%. The two studies with complete participation were rather small studies each including 15 patients.
Importantly, SPR varied between diagnostic groups being lowest when inviting survivors of central nervous system tumors (73%; 95% CI [0.63-0.81]) or hematologic malignancies (67%; 95% CI [0.52-0.79]). A tendency towards a higher SPR in studies of solid tumors compared to hematological malignancies may be due to a different burden of late effects. Although survivors of solid tumors have an increased risk of late effects [30, 31], a higher risk is reported in survivors of hematologic malignancies [32, 33] or central nervous system tumors [34]. Although the extent of cognitive impairments can be different in the two patient groups, attention deficits and impairments in executive functioning among survivors of pediatric acute lymphoblastic leukemia and central nervous system tumors have been identified [35, 36]. Executive functions are cognitive high-level skills, controlling, organizing and planning neurocognitive activity such as behavioral actions and social learning. Thus, pediatric cancer survivors may be challenged to read an invitation letter and take the initiative to reply, due to impaired planning and organizing skills; this could lead to a lower participation rate.
Attendance in late effect clinics has been widely investigated [37, 38]. Nathan et al. [39] found increased rates of attendance associated with female gender and higher socieoeconomic status. In our cohorts, information on sex distribution in the invited participants was not available, and therefore we could not estimate the sex distribution in invited compared to the sex distribution of participants accepting to be enrolled in the study. Further, as recruitment methods were inconsistently reported, we were not able to investigate if contact by telephone or letter influenced the participation rate. Harlan et al. [7] described that extensive efforts were necessary to recruite adolescents and young adults to complete a questionaire. In the future, we need to consider more personalized recruitment strategies, as study populations in late effect studies are heterogenous also regarding age.
We divided the examinations in the studies to be of either short (less than three hours) or long (more than three hours) duration to evaluate if a long examination was a potential barrier for participation. Studies on solid tumors had small samples sizes and always a short duration of clinical examination and either a long or missing follow-up time. This might result in discrete changes in small sample sizes being possibly of disproportionate importance. Five studies included hematologic cancers with a long duration and long median follow-up time of more than ten years. This implies that marginal analyses considering one factor at a time may be explained or confounded by other factors. The five studies of mixed tumors were quite heterogenous considering participants and duration of clinical examinations resulting in a more imprecise pooled estimate for this group.
We did not find any statistically significant differences in time since diagnosis and duration of examination in our study. The majority of participants were invited more than ten years after being diagnosed, and we found that neither time since diagnosis nor the duration of the examination were challenging in recruitment.
Underreporting of recruitment strategies also made it difficult for Hudson et al.[40] to evaluate the impact of different recruitment strategies. They encouraged transparency in the reporting of participant identification, invitation and consent to enable researchers to understand research implications, risk of bias and to whom results apply [40]. This is additional information compared with the current structured guidelines for observational studies [41].
Although it has been clear for more than four decades that lifelong survivorship care is needed for most survivors, there is a great variability in the provision of long-term follow-up programmes for childhood cancer survivors across countries and thus still an urgent need for improvement in long-term follow-up care of survivors [42, 43]. Even in the Nordic countries, where late effect clinics are more widely distributed than in many other countries, a low participation rate might lead to little or no follow-up of childhood cancer survivors and thus the risk of an increased morbidity in childhood cancer survivors.
We encourage future clinical late effect studies to describe the recruitment procedure more thoroughly including sex distribution of the invited patients, how patients were contacted and by whom. We believe that this will lead to a better understanding of the factors influencing participation rates and thus improved participation.