Development and validation of a risk assessment tool to improve early detection of childhood, adolescent and young adult cancer

Childhood and teenage cancer ranks as the 6th leading cause of total cancer burden worldwide and is associated with significant long-term morbidity (1). Cancer is the commonest cause of mortality by disease among children and young people in the United Kingdom (UK) (2). Delayed cancer detection is a known contributing factor(3), with presentation at advanced stages recognised to reduce survival(4). The UK has longer time-to-diagnosis across childhood cancers compared to other high-income countries(5, 6), as well as higher mortality rates across teenage and young adult cancers(7). This is an ongoing concern for young people with cancer, who, in a recent national survey, highlighted early diagnosis research as one of their top ten research priorities(8). This collectively highlights a clear health challenge and there is a pressing need to improve early detection of these cancers in the UK. Indeed, this is in line with the National Health Service (NHS) Long Term Plan (2019)(9), UK Cancer Reform Strategy (2015)(10) and Childhood Cancer and Leukaemia Group (CCLG) Strategic Plan (2020)(11).
The non-specific presentation and relative rarity of childhood, adolescent and young adult cancers (TYA) pose difficult diagnostic challenges to clinicians(12) and increase the possibility of delays. This is particularly relevant to general practitioners (GPs) who encounter childhood and TYA cancer patients at the earliest stages of the disease. National awareness initiatives, such as HEADSMART(5), have been employed to address this challenge in the UK. Although this initiative contributed to substantial improvements in diagnostic intervals in central nervous system (CNS) tumours, the national time-to-diagnosis target of 4 weeks has not been reached for all age groups(13). Furthermore, GPs remain unconfident in diagnosing childhood cancers even after taking part in this initiative(5). Similarly, recent findings of the “Accelerate, Coordinate, Evaluate” (ACE) programme demonstrated ongoing delays in referrals and cancer diagnosis in TYA(14). Clearly, novel approaches need to be explored to supplement current pathways.

Computer-based clinical decision tools (CDTs) are increasingly being used in clinical settings, supporting medical decision-making where challenges such as diagnostic uncertainty are present(15). Overall, CDTs have been reported to reduce diagnostic errors(16), improve clinical practice and patient care(15, 16). In primary care settings, recent evidence suggests that technology-based CDTs provide the most successful interventions in reducing diagnostic inaccuracies(17).The potential for CDTs in cancer diagnosis have been highlighted as an “area of extraordinary opportunity”, with promising developments seen in several adult cancers(18). A recent systematic review(19) has shown that these tools for cancer risk prediction have the potential to improve decision-making and clinical service outcomes, as well as one study showing reduction in time-to-diagnosis. Despite these advancements in adult cancers, CDTs have not been explored in childhood and TYA cancers.

Accordingly, through this study, we plan to use QResearch Database, the largest GP electronic health record database in the UK, to explore ways to detect childhood and TYA cancers earlier. We plan to achieve this through further studying early symptoms and signs in primary care, and subsequently using our findings to develop a novel GP-based risk prediction tool.

For identifying early pre-diagnostic clinical presentations of childhood and TYA cancer (acute lymphoblastic leukaemia, lymphoma and central nervous system tumours) we will specifically require GP data on:

• Demographics (year of birth, sex, ethnicity, Townsend Deprivation Score, GP-practice, date of registration, and date of leaving practice)

• Coded symptoms occurring at >1% frequency in cases or controls. This will be identified by selecting all symptoms and examination findings in sections 1 and 2 in the QResearch Read Codes. This will be including but not limited to the following symptoms and examination findings:

Abdominal mass, enlarged abdominal organ
Abdominal pain
Hepato-splenomegaly
Hepatomegaly
Splenomegaly
Bleeding
Bruising
Petechiae
Lymphadenopathy, enlarged lymph node, painless lymph node
Shortness of breath and lymphadenopathy/splenomegaly
Persistent cough
Persistent wheeze
Blood in stool
Pallor, pale
Pruritus, itching

Bone swelling
Bone pain
Growing pains
Lump (increasing in size)
Fracture (recurrent)

Headache
Vomiting (persistent)
Abnormal gait
Loss of coordination
Back pain
Squint
Precocious puberty
Puberty delay
Changes in sight (visual problems) - double vision, blurred vision, loss of vision
Abnormal eye movement
Dizziness
Falls (recurrent)
Seizure
Developmental regression
Developmental delay
Change in behaviour
Increased head circumference
Wry Neck, head tilt, stiff neck
Change in taste and smell


Lethargy/fatigue/tiredness
Nightsweats
Weight loss
Change in appetite (not feeding)
Fever
Recurrent infections
Parental concern/parental anxiety
Changes in sleep pattern
Inconsolable crying


• Clinical diagnostic codes (ICD 9 & 10)

ICD10 (C70.0, C70.1, C70.9, C71.0-C71.9, C72.0-C72.9, C79.0-C79.9, C81.0-C81.9, C82.0-C82.9, C83.0-C83.9, C84.0-C84.9, C85.0-C85.9, C86.0-C86.6, C88.0-88.9, C91.0- C91.9 C91.A, C91.Z, C92.0- C92.9 C92.A, C92.Z, C93.0-C93.9, C93.Z, C94.0-4,6,8, C95.0 C95.1,C95.9)
ICD-9 (191 (all clauses), 192 (all clauses), 196-199 (all clauses), 200-208 (all clauses)).

• Number of presentations to general practice

• Referral type 
to secondary care

Professor Julia Hippisley-Cox - University of Oxford

Dr Jason Oke - University of Oxford

Professor Anthony Harnden - University of Oxford