It Takes One Line (of SAS) to Analyze Your Data, but it Takes Many - - PowerPoint PPT Presentation

It Takes One Line (of SAS) to Analyze Your Data, but it Takes Many Lifelines to See it (Through)

Shawn Murphy, Victor Castro, Wannapa Mahamaneerat, and Taowei David Wang Partners HealthCare October 30th, 2012

Does Glargine Increase Cancer Risk?

• A large German study on diabetes database

suggests so

– Controversial, but very large sample – Other Countries did the same study – Other Countries did the same study

• Sweden, UK, Scotland

– Conflicting results

• What about our own data?

– Focus on glargine and pancreatic cancer – Performed 2 studies

Cohort Study from EHR

All Patients with some 70,772 Patients 1980-2010 with some form of Insulin Therapy

All Patients with some Glargine Insulin

Cohort Study from EHR

70,772 Patients 2135 Glargine only 51177 Other insulin only 17460 Both with some form of Insulin Therapy Other Insu Both

All Patients with some

Statistical Significance

Glargine Insulin

Cohort Study from EHR

with some form of Insulin Therapy

Statistical Significance

Other Insu Both

All Patients with some

Statistical Significance

Glargine Insulin

Cohort Study from EHR

with some form of Insulin Therapy But in Lifelines2, we see that Order of Cancer and Glargine is not enforced.

Statistical Significance

Other Insu Both

cancer glargine

All Patients with some

Statistical Significance

Glargine Insulin

Cohort Study from EHR

with some form of Insulin Therapy But in Lifelines2, we see that Order of Cancer and Glargine is not enforced. Nor is the time taken from Glargine to Cancer taken into account.

Statistical Significance

cancer glargine

Other Insu Both

All Patients with some All Patients with Pancreatic Cancer after Exposure Glargine Other Insulin

Cohort Study from EHR

with some form of Insulin Therapy Exposure to Insulin Therapy Both

All Patients with some All Patients with Pancreatic Cancer after Exposure

Much smaller Statistical Significance!

Glargine Other Insulin

Cohort Study from EHR

with some form of Insulin Therapy Exposure to Insulin Therapy Both

Case-Control Study in EHR

8000+ pancreatic cancer patients (cases) All Patients with Pancreatic Cancer

All Patients with Pancreatic Cancer 8000+ pancreatic cancer patients (cases)

Case-Control Study in EHR

Matched Controls: Patients without Pancreatic Cancer 25000+ pancreatic cancer patients (controls) Matched by Age Gender Fact Counts Start and End Dates

All Patients with Pancreatic Cancer 8000+ pancreatic cancer patients (cases)

Case-Control Study in EHR

Matched Controls: Patients without Pancreatic Cancer Fact Counts But… Fact distribution can be very different

All Patients with Pancreatic Cancer 8000+ pancreatic cancer patients (cases)

Case-Control Study in EHR

Matched Controls: Patients without Pancreatic Cancer Case Control spike Fact Counts But… Fact distribution can be very different

All Patients with Pancreatic Cancer 8000+ pancreatic cancer patients (cases)

Case-Control Study in EHR

Matched Controls: Patients without Pancreatic Cancer Fact Counts But… Fact distribution can be very different We used Fact Count/Year. No Statistical Significance

Summary

• Running statistical analysis is easy
• Ensuring our data is

– Clean, representative, and suitable for the type of analysis is not analysis is not

• Lifelines2 is helpful

– Check if our data is clean – See if matching criteria are appropriate – Present patient data to colleagues

References and Links

• Lifelines2 in i2b2 (TimeAlign) https://community.i2b2.org/wiki/display/timealign/TimeAlign
• Hemkens, L., et al., Risk of malignancies in patients with diabetes treated with human insulin or insulin analogues: a cohort
• study. Diabetologia, 2009. 52(9): p. 1732-1744. http://dx.doi.org/10.1007/s00125-009-1418-4
• Colhoun, H., Use of insulin glargine and cancer incidence in Scotland: a study from the Scottish Diabetes Research Network

Epidemiology Group. Diabetologia, 2009. 52(9): p. 1755-1765. http://dx.doi.org/10.1007/s00125-009-1453-1

• Currie, C., C. Poole, and E. Gale, The influence of glucose-lowering therapies on cancer risk in type 2 diabetes. Diabetologia,
• 2009. 52(9): p. 1766-1777. http://dx.doi.org/10.1007/s00125-009-1440-6
• Jonasson, J., et al., Insulin glargine use and short-term incidence of malignancies—a population-based follow-up study in
• Sweden. Diabetologia, 2009. 52(9): p. 1745-1754. http://dx.doi.org/10.1007/s00125-009-1444-2
• Sweden. Diabetologia, 2009. 52(9): p. 1745-1754. http://dx.doi.org/10.1007/s00125-009-1444-2
• Rosenstock, J., et al., Similar risk of malignancy with insulin glargine and neutral protamine Hagedorn (NPH) insulin in

patients with type 2 diabetes: findings from a 5 year randomised, open-label study. Diabetologia, 2009. 52(9): p. 1971-1973. http://dx.doi.org/10.1007/s00125-009-1452-2 (letter to the editor)

• Smith, U. and E. Gale, Does diabetes therapy influence the risk of cancer? Diabetologia, 2009. 52(9): p. 1699-1708.

http://dx.doi.org/10.1007/s00125-009-1441-5 (editorial)

• Pocock, S.J. and L. Smeeth, Insulin glargine and malignancy: an unwarranted alarm. The Lancet, 2009. 374(9689): p. 511-513.
• Gale, E.A.M., Insulin glargine and cancer: another side to the story? The Lancet, 2009. 374(9689): p. 521-521.

Backup Slides

Cohort

• Start with people with Glargine and people

with other insulin and follow them longitudinally to see who develops cancer.

• Relative risk.
• Relative risk.

Case-Control

• Focus only on patients who have pancreatic

cancer (case) and select similar patients to compare with (control)

• Odds-ratio
• Odds-ratio
• Better-suited because pancreatic cancer is a

relatively rare condition. This way we narrow the scope

Relative Risk

Glargine Pancreatic Cancer No Pancreatic Cancer a b No Glargine c d Relative Risk = (a/(a+b)) / (c/(c+d))