Insights

Our Pancreatic Cancer Analysis: How We Did It - Prognos Health

Written by Nick Zgorski | Jun 1, 2017 7:54:01 PM

The Prognos Study Method

By Carol Smyth, MD & Brian Saindon

Background

About 80% of pancreatic cancers are found late, due to the lack of symptoms associated with the tumor until its late stages. Finding the tumor late results in poor survival rates. There is currently no specific blood test to screen for pancreatic cancer, and researchers around the world are working on identifying tests that would help to identify early pancreatic tumors. A 2009 study by Harsha et al found that over 1,000 genes can be overexpressed in pancreatic cancer, implying that there are many potential biomarkers waiting to be discovered (http://journals.plos.org/plosmedicine/article?id=10.1371/journal.pmed.1000046).

Method

Using the Prognos database of lab tests, which currently includes 9 billion tests from over 160 million patients, we identified patients at first diagnosis of pancreatic cancer via ICD-9 and ICD-10 codes in 2016. Tracking their lab results retrospectively, we looked back by 5 years to determine if there was a pattern in the “noise” of the lab tests, to see if there are potentially abnormal results in commonly conducted lab tests that may indicate an increased risk of pancreatic cancer on a population basis. This type of retrospective analysis has not been published to date in pancreatic cancer, to the best of our knowledge/research. Because there are many tests that were candidates for this analysis, we elected to focus on the following lab tests and results:

  • CA 19-9 and alkaline phosphatase because our previous research suggested that these were potential biomarkers (see http://abstracts.asco.org/156/AbstView_156_153453.html). In addition, CA 19-9 is the only biomarker approved by the FDA for pancreatic cancer (http://www.sciencedirect.com/science/article/pii/S1590865815006908).
  • Amylase – levels found to be high in less than 50% of patients with operable pancreatic cancer tumors (http://emedicine.medscape.com/article/280605-overview)
  • HbA1c and fasting glucose – due to the high rate of diabetes in patients with pancreatic cancer –  new onset diabetes is common in the year prior to diagnosis of pancreatic cancer
  • CA 125 – we included this test because previous studies suggest a rise in CA 125 is present up to 2 years prior to pancreatic cancer diagnosis (https://www.pancreaticcancer.org.uk/latest-news/2015/february/tumour-biomarkers-research-shows-promise-for-early-detection/)
  • Hemoglobin and platelets – these test results are often low in patients with pancreatic cancer

In the Prognos database, we identified 11,571 patients identified with ICD codes for pancreatic cancer in the calendar year 2016. Demographics of this patient group follow. We note that the group is approximately equal between genders, and that the majority of patients (>60%) are between the ages of 61 and 80 years old. These findings are in keeping with data published by the National Institutes of Health (https://seer.cancer.gov/statfacts/html/pancreas.html).

Cohort Descriptives

Measure Frequency (n) (%)
Patient Gender
Male 5,672 49.02%
Female 5,898 50.97%
Unknown 1 0.01%
Patient Age Distribution
Patient Age Group (years)
0-10 10 0.09%
11-20 11 0.10%
21-30 45 0.39%
31-40 170 1.47%
41-50 529 4.57%
51-60 2014 17.41%
61-70 3814 32.96%
71-80 3296 28.49%
81-90 1497 12.94%
91-100 182 1.57%
>100 2 0.02%
Unknown 1 0.01%

Patients with First Pancreatic Cancer Diagnosis in 2016 (n=11,571)

In the next blog, we will report on the analysis of our retrospective look at the common lab tests listed above, to see if there is a pattern in abnormal lab results over the 5-year period prior to diagnosis of pancreatic cancer in the 2016 patient cohort. Our hope is that identification of a pattern, if any, might lead to identification of a panel of common tests that might be used for screening for this lethal tumor.

Reading List: