Further to our recent publication of two papers in the BJGP,1,2 we have been asked to evaluate specifically whether dyspepsia is a significant independent predictor of upper gastro-intestinal malignancy (in other words, gastro-oesophageal and pancreatic malignancy) and to consider adding it to the models. These symptoms (heartburn or indigestion) were not included in the original analysis that had focused on more traditional alarm symptoms. We, therefore, undertook an analysis based on the original derivation cohort from the published studies and identified patients with new onset of (a) heartburn or (b) indigestion (other than where heartburn is explicitly mentioned). We determined the age–sex incidence rates. We added both factors to the Cox models and determined the hazard ratios adjusted for the factors in the original models. We tested for interactions between the new variables and age. We evaluated performance of the new models on the original validation dataset using published methods.
Figure 1 shows age–sex incidence rates of each symptom (where dyspepsia represents either heartburn or indigestion). The crude incidence rate for new onset heartburn in patients aged 30–84 years was 130 (95% [CI] 128 to 133) per 100 000 person years for males and 196 (95% [CI] 193 to 199) for females. The incidence rate for indigestion in males was 680 (95% [CI] 680 to 693) per 100 000 person years for males and 844 (95% [CI] 836 to 850) for females. Table 1 shows the hazard ratios for heartburn and indigestion in the new Cox models, adjusted for the other factors in the original models (see footnote). There were no age interactions for these symptoms. Both heartburn and indigestion were independently associated with risk of gastro-oesophageal cancer and also pancreatic cancer in both males and females. The adjusted hazard ratios associated with indigestion without heartburn were higher than those associated with heartburn. For example, females with heartburn had a 2.2-fold increased risk of gastro-oesophageal cancer and a 2.5 fold increased risk of pancreatic cancer. Females with indigestion without mention of heartburn had a 4.3-fold increase in gastro-oesophageal cancer and a 3.8-fold increase in pancreatic cancer. The pattern for males was similar. We therefore retained both heartburn and indigestion in both updated models for males and females. The performance of the updated algorithms on the validation cohort was equivalent to that of the original models for gastro-oesophageal cancer and marginally better for pancreatic cancer. The R2, D-statistic, and receiver operating characteristic statistics for gastro-oesophageal cancer were 71%, 3.2 and 0.90 for females, and 71%, 3.2 and 0.92 for males. The corresponding values for pancreatic cancer were 62%, 2.6 and 0.84 for females, and 64%, 2.7 and 0.86 for males.
In summary, we have identified and quantified two additional symptoms (heartburn and indigestion) that are predictive of both upper GI cancers. We have now included both symptoms in updated models at QCancer (http://www.qcancer.org). As with the other symptoms included in the models, it is important to remember that they represent symptoms that have been significant enough for a patient to present to their GP and for their GP to record. Not all patients with such symptoms will have attended their GP and not all such symptoms will be reported or recorded.
Acknowledgments
We particularly thank Professor Sir Mike Richards (Department of Health cancer tsar) and Ms Ali Stunt (CEO of pancreatic cancer action) for discussing and requesting the additional analyses.
Notes
Competing interests and financial disclosures (as per original paper reproduced here)
All authors have completed the Unified Competing Interest form at http://www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare: JHC is professor of clinical epidemiology at the University of Nottingham and co-director of QResearch® — a not-for-profit organisation that is a joint partnership between the University of Nottingham and EMIS (leading commercial supplier of IT for 60% of general practices in the UK). JHC is also a paid director and co-founder of ClinRisk Ltd that produces software to ensure the reliable and updatable implementation of clinical risk algorithms within clinical computer systems to help improve patient care. The software which implement the algorithms described in this paper are free for anyone to use under the terms of the GNU lesser GPL3. For those who wish to implement software in a closed source setting, then a license fee is payable to ClinRisk Ltd. CC is associate professor of medical statistics at the University of Nottingham and a paid consultant statistician for ClinRisk Ltd. This work and any views expressed within it are solely those of the co-authors and not of any affiliated bodies or organisations.
- © British Journal of General Practice 2012