Search

Additional file 1: Supplementary Table 1. ANCOVA Model tables for each outcome verbal fluency outcome measure. Supplementary Table 2. Sensitivity and Specificity of cut off points on receiver operating curve forsemantic verbal fluency.

Timely diagnosis of a brain tumour is an important aspect of health care provision. There are no known risk factors for brain tumours in the majority of patients and most brain tumours are diagnosed following symptomatic presentation. The nature of presenting symptoms is critical to the speed of diagnosis. Understanding how symptoms are associated with the length of diagnostic time and its prognostic effect is a priority for early diagnosis research which could inform strategies to achieve an earlier detection of brain cancer. This thesis focuses on four key aspects in the diagnosis, namely 1) assessing the usefulness of existing primary care referral guidelines for an urgent brain imaging; 2) describing clinical and diagnostic pathway of adult patients with any type of brain tumour prior to a diagnosis; 3) assessing the association between presenting symptoms and diagnostic intervals and their impact on clinical care and survival in patients with the most malignant brain tumour, and lastly 4) assessing the value of a brief cognitive assessment in identifying which symptomatic patients are most likely to have a brain tumour. In the first study, a total of 3257 referrals for direct-access CT (DACT) head imaging were retrospectively identified. Diagnostic yield for a brain tumour was 1.8% (N=53 scans). There were no false negative scans. Referral symptoms were categorised based on Kernick’s symptom classification and NICE 2005 referral guidelines. Only referrals for “symptoms related to the CNS” achieved positive predictive value (PPV) of 3%. Nearly 80% of patients whose CT revealed a brain tumour had symptoms grouped under the “symptom related to the CNS” category. A large proportion of referrals were for a ‘simple’ headache (27%) but few of these patients had a tumour. Improving guidelines to better identify patients at risk of a brain tumour should be a priority, to improve speed of diagnosis and reduce unnecessary imaging and costs. However, as I explore, this will not be not straight forward, in particular because of the lack of specificity of symptoms associated with a brain tumour. The second prospective study included 180 patients with a heterogenous range of brain tumour types where I describe the clinical and diagnostic pathway to diagnosis. Patients symptoms were grouped as episodic attacks of consciousness (seizure), alarm or vague symptoms, and were recorded temporally along the patient’s diagnostic journey, occurring as first, second or at any point before the diagnosis. Vague, non-specific symptoms, were the most common symptoms across the pathway and were associated with increased diagnostic interval compared to episodic attacks of consciousness (seizure) (median 96 days vs 22 days, respectively, adjusted odds ratio 3.0, 95%CI 1.2-7.9) and a higher likelihood of more than 3 GP visits prior to a diagnosis (aOR 5.6, 95%CI 1.2-26.1). Interestingly, over half of patients with a seizure resulting in emergency admission actually reported initial vague non-specific symptoms. In a third study, 354 patients with a high-grade glioma were retrospectively identified from local multi-disciplinary team minutes between July 2010 and March 2015. Data on primary symptoms and time to diagnosis were corroborated by contacting deceased patients’ General Practitioners. Type of presenting symptoms was associated with diagnostic timeliness and overall survival. Multivariate models showed that development of cognitive symptom was an independent adverse predictor for 12-month survival (multivariate HR 1.8, 95%TCI 1.2-2.6) but was not associated with the type of surgery performed (ie biopsy versus debulking). Earlier diagnosis did not have an impact on survival. Finally, in a proof of concept study, performance on a 1-minute semantic and phonemic verbal fluency (VF) test was significantly poorer in patients with a brain tumour (n=180) compared with patients referred from primary care for brain imaging, but without a brain tumour (n=90). The biggest difference in scores between the groups was observed for semantic total VF scores compared to phonemic total VF scores (Cohen’s d = -0.97 and -0.47, respectively). A score of 14 animals on SVF test was associated with an 84% sensitivity and 54% specificity for not having a brain tumour (AUC 0.75 p