We’re back with our third blog in our pancreatic cancer series. In addition to an update on our methods, we are pleased to begin sharing our results. Keep reading to see what we’ve learned so far, and what’s still to come…
In order to restrict our analysis to adults (and adult reference range values for the tests) to aid interpretation of the results, we set a lower limit of age for the study population at 23 years in 2016. This change results in the exclusion of 30 patients from our original population, new N = 11,541.
In addition, we excluded tests from the results analysis if the result was invalid, meaning that the test was ordered but not conducted for various reasons. This resulted in the exclusion of 1,014 tests (0.5%). We also looked at tests by year conducted, not by month in that year. Therefore, a more granular timeline of testing frequency might yield more insights, and could be a topic for future analysis.
We also note that using the ICD-10 code for pancreatic cancer used to identify patients does not allow for the histological type of cancer.
We have divided up our results, and will present a second blog post of results next week. In this blog post, we will look at testing frequency, and the rates of abnormal high and low test results from 1 to 5 years prior to diagnosis of pancreatic cancer in our population in 2016.
Figure 1 identifies the percent of patients that were tested for each of the selected tests stratified by year and test for the 5 years leading up to their cancer diagnosis in 2016.
As you would expect, common serum tests appear frequently, including alkaline phosphatase, platelets and hemoglobin. But the graph shows that the testing frequency slope increases for alkaline phosphatase across the 5-year period, with steep increases at years 4 and 2 prior to diagnosis. Similarly, the slopes of testing frequency of hemoglobin and platelets also rises consistently across the lookback period, with steep increases in slope for both tests at 2 years prior to diagnosis. It is notable that testing for CA19-9, a marker of abdominal tumors, shows a low frequency until 2 years prior to diagnosis, when the slope of the frequency of conducting this test rises fast. Of the remaining tests in our analysis, there is a slow rise in the frequency of HbA1c testing in the 5 years prior to diagnosis, and a virtual flat line for the frequency of amylase and CA125 testing.
We looked at the frequency of abnormal high results on the tests included in the analysis, and in line with previously published reports, including our ASCO abstract from 2015, a high frequency of abnormal high results were seen for alkaline phosphatase and CA 19-9 prior to diagnosis (Figure 2). We also found that high HbA1c results were common in this patient cohort in the years prior to diagnosis. The slopes of the rises in the test results over time for these three tests are of interest. The test result values for alkaline phosphatase and CA19-9 increase steeply 2 years prior to diagnosis. In contrast, the slope of HbA1c test results rose gradually over the 5-year period prior to diagnosis.
The rates of detection of abnormal high test results in the other four tests included in the analysis were relatively flat, with small rises seen in rates of high values for amylase and CA 125 in the year prior to diagnosis. A rise in the frequency of abnormally high values of hemoglobin and platelets would not be expected in this patient cohort, and the data supports this hypothesis.
The frequency of abnormal low results on the tests included in this analysis shows that there is a slow rise in the rate of abnormal low values for hemoglobin across the 5 years prior to diagnosis, with a sharp rise 2 years prior to diagnosis, when the frequency doubled. There is a flatter slope to a rise in low values for platelets until 2 years prior to diagnosis, when the slope rises more sharply. In contrast, frequencies of abnormal low results for alkaline phosphatase rise gradually and then fall, starting at 2 years pre-diagnosis. The frequency of abnormally low HbA1c remained flat across the lookback period.
There was no rise seen in the frequency of abnormal, low results for the following tests: serum amylase, CA19-9, CA125 and fasting glucose.
Our analysis indicates that:
These findings suggest that there may be a trend in test results in the years prior to diagnosis of pancreatic cancer, particularly at the 2-year prior mark. In the next blog post, we continue our presentation of the results of this analysis. Tune in as Dr. Jason Bhan, Prognos Cofounder and Chief Medical Officer shares his thoughts.