Your Internet search history could predict your cancer risk

One’s internet search history can tell a lot about someone, and now in terms of cancer detection, too, according to Microsoft researchers. In their study, published this week in the Journal of Oncology Practice, they were able to detect those who had pancreatic cancer — even up to five months before a doctor’s diagnosis!

So what does this mean? It’s huge news for early pancreatic cancer detection. Pancreatic cancer is typically an aggressive cancer that’s harder to detect than others. However, if search history is monitored, it could help show early warning signs, which can lead to an earlier diagnosis and a better prognosis. “Because pancreatic [cancer] may progress from stage I to stage IV in just over 1 year, this screening capability could increase 5-year survival,” said the findings. “Screening for pancreatic adenocarcinoma aims to detect the disease at a preinvasive or early invasive stage when it is still curable by surgical intervention and chemotherapy.”

The researchers consisted of Columbia University graduate student John Paparrizos and Microsoft researchers Dr. Ryen White and Dr. Eric Horvitz. They used proprietary logs of 9.2 million web queries from Microsoft’s search engine, Bing, from October 2013 to May 2015. Their sample consisted of English-speaking people in the U.S., and the researchers analyzed their search and click activities.

They first looked at search terms that indicated a recent diagnosis, such as, “Why did I get cancer in pancreas,” and “I was told I have pancreatic cancer, what to expect.” From there, the researchers went back months to determine initial symptoms people had searched for. Fascinating, right?

They reviewed signs, symptoms, and risk factors, and developed a symptom set, which included some of the following: sudden weight loss, taste changes, light stool, and abdominal pain. Then, synonyms were identified, such as “belly pain” or “stomach ache” for “abdominal pain,” and risk factors were looked at, as well, such as alcoholism, and its associated symptoms.

“We showed specifically that we can identify 5% to 15% of cases, while preserving extremely low false-positive rates,” the researchers’ paper said.

Any detection sounds great to us.

And just think — if the above can help detect pancreatic cancer, imagine what it can do in terms of detecting other types of cancer and health ailments.