Weight loss as a sign of cancer in the primary care population of Kaiser Permanente Washington – a retrospective cohort study using routinely collected electronic health record data.

Problem

In the United Kingdom (UK), the National Institute for Health and Care Excellence (NICE) recommends urgent referral for the investigation of cancer in patients with unexplained weight loss in primary care, but provide no guidance on the extent or timing of weight loss. Weight is not routinely recorded in UK primary care and is often missing in primary care databases. Consequently, NHS primary care data are unlikely to be provide rich enough weight information to derive the extent and timing of weight loss that is most indicative of an underlying malignancy. Unlike the UK, weight is measured routinely in the USA and weight data is sufficiently complete to allow detailed analysis of weight change. The overall aim is to provide the evidence to allow GPs to more effectively implement the NICE cancer guidance on the management of unexplained weight loss using USA primary care data

Approach

The study population comprised a dataset of 50,000 randomly selected US primary care patients aged >=40 years who were members of Kaiser Permanente Washington (KPW) on their date of birth in 2006 and resided in the local cancer registry catchment area. We examined associations between measured weight loss and a diagnosis of cancer. A multi-variable Cox proportional hazard model with time-varying covariates for percentage weight change, symptoms, and blood test results. The model was used to estimate the risk of cancer for combinations of weight loss and intervals of weight measurement in order to identify thresholds for further investigation.

Findings

Of 43,302 patients with measured weight and over 288,000 person-years of follow-up, 3,352 were diagnosed with cancer. The median (IQR) number of weight change measures of 13 (6, 24) at a median (IQR) interval of 39 (15, 104) days. Percentage weight change has a linear association with the risk of any cancer (Hazard ratio of 1.05, 95% CI 1.04 to 1.07 per unit % weight loss independent of age, sex, ethnicity, haemoglobin and platelets, previous cancer diagnosis, the time interval between consecutive weight measures and diagnoses of conditions associated with weight change. In men aged 60 with no existing comorbidities, weight loss of 5% over 6 months is associated with a risk of 2.3%, increasing to 3.0% if weight loss occured over a 1 month interval.

Consequences

Measured weight loss is independently associated with an increased risk of cancer. More weight loss means a higher risk of cancer independent of the time interval between weight measurements and a previous diagnosis of cancer. Diagnostic strategies necessitating that weight loss is observed over a fixed period may delay diagnostic action. Our initial analysis will be expanded to explore the utility of diagnostic strategies using increasing thresholds of measured weight loss and changes in combinations with changes in biomarkers for presentation at the conference.