Why is the specificity of a biomarker important? PSA for prostate cancer as an example.

I’ve described what biomakers are here and how they are discovered here. I’ve spent so much time discussing biomarkers because this is one of the aspects of personalized medicine that you may have already encountered in your doctor’s office or will encounter soon.  Similar to how we understand risk, it’s important to understand biomarkers because many healthcare decisions will be based on the results of tests that look at the presence, absence, or quantity of biomarkers.

So how do you know what the results of a biomarker test mean or whether or not a biomarker is good or not?  Scientists have created two measurements that can quickly tell you how good a biomarker test is: sensitivity and specificity.  Before we talk about what those two measurements measure, let’s first talk about the different scenarios for a patient after getting the result of a test using a biomarker.

  1. The test is positive and the patient has the disease. This is a good scenario because then the patient can move forward with the appropriate treatment.
  2. The test is positive but the patient doesn’t have the disease.  This is what we call a “false positive” because the test is incorrectly showing up as positive.  This can cause huge issues because a patient will receive a diagnosis, follow-up tests or treatment even though they don’t have a disease.
  3. The test is negative and the patient doesn’t have the disease.  Again, this is a good scenario because the patient is a-okay.
  4. The test is negative but the patient has the disease.  This is a “false negative” because the test is falsely showing that the patient doesn’t have a disease when they actually do.  This can also cause issues because then a patient won’t be treated even though they should be.

sensitivity_specificitySensitivity and specificity measure the best case scenarios – sensitivity measures when the test is positive and the patient has a disease and specificity measures when the test is negative and the person doesn’t have a disease.  The ideal test has 100% sensitivity (all sick people are tested as being sick) and 100% specificity (all healthy people have a negative test).  But this ideal situation is difficult to achieve.  Let’s use Prostate Specific Antigen (PSA) as a biomarker test for prostate cancer as an example.

There are 241,740 new cases of prostate cancer each year, and it is the most common malignancy in men (29% of all male cancers).  PSA levels are screened in men over 50 for increased expression and over $3 billion per year is spent for this screening.   What is PSA? It’s a protein produced by the prostate gland and can be elevated in men with prostate cancer, which is why it has been used as a biomarker for prostate cancer.  However, PSA may also be elevated in men with other conditions such as prostatitis (inflammation of the prostate), benign prostatic hyperplasia (enlargement of the prostate), or urinary tract infections.  Because of this, the PSA test is highly susceptible to false positives and false negatives.  Typically PSA greater than 4 ng/mL (this means that there is 4 nanograms of PSA protein in 1 milliliter of urine) is considered a positive test result for prostate cancer.  The sensitivity at this level is 21%, which means 21% of patients that have prostate cancer have PSA levels  greater than 4 ng/mL.  Specificity of the test is 91%, which means that 91% of patients who test negative do not have prostate cancer. This is good – there aren’t many patients who don’t have prostate having follow-up test or biopsy because of a false positive (because of the high specificity).

Let’s see what happens if a lower concentration of PSA are used as a cut off to try to detect more patients with cancer.  If you look at patients with greater than 1.1 ng/mL, the sensitivity increases significantly to 83%, which means that more people with cancer are being detected (great news!).  The trade off is a specificity of 39%, which means that a huge number of patients will be incorrectly diagnosed as having cancer (high false positives).  This will result in follow-up tests and biopsies.  The effect of these tests and biopsies are both psychological (thinking you have cancer when you don’t) and physical (an increased risk of complications and side effects caused by the biopsy).

psa test

So is PSA a good test or a bad test?  For the patients who have received a positive PSA test and have aggressive prostate cancer, this test saves lives.  However,because this is a common screening test, it does cost a lot of money.  Because the sensitivity of the test at higher PSA concentrations is so low, some cancers get missed.  And if the cut off is decreased to catch more of these cancers, there’s a much higher number of false positives in men who do not have prostate cancer resulting in costly, stressful test that may have added complications.

What’s the solution?  In the case of PSA, doctors have started measuring PSA levels over time.  Increases in PSA levels over time are found more often in men with prostate cancer.  PSA also exists in a few different forms. PSA can either be attached to other molecules or not.  The form that isn’t bound to other molecules is called “free-PSA” and scientists have found that the amount of free-PSA compared to the total amount of PSA is reduced in men with prostate cancer.  These improvements have decreased false negatives and false positives, making it a much better test.

Overall, biomarkers have the potential to revolutionize medicine, and in so many cases they already have.  But for you as a patient, understanding the challenges and pitfalls of these tests will help you be a more empowered patient with the knowledge to ask key questions when you receive the results from one of these new tests.

 

How do you find a biomarker? A needle in the haystack.

biomarker_useBiomarkers are biological substances that can be measured to indicate some state of disease.  They can be used to detect a disease early, diagnose a disease, track the progression of the disease, predict how quickly a disease will progress, determine what the best treatment is for the disease, or monitor whether or not a treatment is working. Biomarkers have the potential to do so much, and identifying biomarkers for different steps in the health/disease continuum would help doctors to provide each individual with targeted, precision healthcare.  Biomarkers have the potential to save billions of healthcare dollars by helping prevent disease, by treating disease early (when it’s usually less expensive to treat), or by targeting treatments and avoid giving a treatment that won’t be effective.

spotthedifferencesWith all this potential, you would expect doctors to be using data from biomarkers to guide every single healthcare decision – but this isn’t the case quite yet.  First scientists have to find these biomarkers – a process often referred to as biomarker discovery.  I like to compare finding a biomaker to those “spot the differences” games where you have to look at two images and circle what is different in one picture compared to the other.  This is exactly what scientists do when finding a biomarker, except instead of comparing pictures, they are comparing patients.  And it’s not an easy game of “spot the differences” it’s complicated: the pictures are small and there are tons of details.

Let’s imagine a scenario that a scientist might face when wanting to find a biomarker for the early detection of pancreatic cancer.   Cancer is caused by mutations in the DNA, so you decide to look for DNA mutations as your biomarker for pancreatic cancer. So how do you “spot the differences” to find DNA biomarkers for pancreatic cancer?  First, you will need patient samples – maybe tissue or blood samples from a biobank that already has samples from patients with pancreatic cancer.  If samples aren’t already available, you will have to initiate a study partnering with doctors to collect samples from pancreatic cancer patients for you.  You will also need the second “picture” to compare the pancreatic cancer “picture” to.  This second picture will be samples from people who don’t have pancreatic cancer (scientists usually call this group the “control” group).  Then you have to “look” at the two groups’ DNA so you can find those differences.  This “looking” is often done by some genomics method like sequencing the DNA. This is where a lot of the complication comes in because if you look at all of the DNA, you will be comparing 3 billion individual nucleotides (the A, T, G, and Cs we’ve discussed in earlier posts) from each patient to each of the controls.  Even if you just look at the DNA that makes proteins, you’re still comparing 30 million nucleotides per patient.  And you can’t just compare one patient to one control!  Each of us is genetically different by ~1%, so you need to compare many patients to many controls to make sure that you find DNA that is involved in the disease and not just the ~1% that is already different between individuals.  But wait, we’re not done yet!  The biomarkers that you identify have to be validated – or double checked – to make sure that these differences just weren’t found by mistake.  And before biomarkers can be used in the clinic, they need to be approved by the Food and Drug Administration (FDA)

biomarker_discovery

From http://www.pfizer.ie/personalized_med.cfm


Whew… that was a lot of work! And so many people were involved: lead scientists who directed the project and got the money to fund it, researchers who do most of the work, computer people who are experts at crunching all of the data, and maybe even engineers to help run the equipment. Finding the biomarker needle in the biological haystack is difficult and takes time, money, and lots of people.  This is one of the reasons why there are only 20 FDA approved biomarkers for cancer (data from 2014).  But just because it’s difficult, doesn’t mean it’s impossible.  Furthermore, this effort is necessary to improve healthcare and decrease healthcare costs in the future.  It just might take a bit more time than we’d all like.

If you want to read more about the challenges and some of the solutions to biomarker discovery in cancer, take a look at this scientific article.  Or read about some successes from right in our backyard at Arizona State University on identifying biomarkers for the early detection of ovarian cancer and breast cancer.

 

What is a biomarker? A cornerstone of personalized medicine.

What is a biomarker? Biomarkers are biological measures of health or disease and are a cornerstone for personalized medicine.Historically, diagnosing a disease was based on symptoms. This reminds me of a joke.  A patient goes to see the doctor and tells him “Doctor, I hurt everywhere.” The patients touches his head “I hurt here”, he touches his arm “I hurt here”, he touches his stomach “I hurt here” and on and on.  The doctor looks at him and says “I know what’s wrong with you!  You have a broken finger!”

No one wants to be diagnosed or misdiagnosed with a broken finger. This isn’t to say understanding symptoms and using this information to contribute to a diagnosis isn’t important.  But what if…
…symptoms don’t lead to an obvious diagnosis?
…two patients have the same symptoms, but different diseases?
…two patients have the same disease, but different causes – either the root cause is different or they both have lung cancer but the genetic mutations in each cancer is different.  In this case different treatments would be are needed.
…two patients have different diseases, but similar causes – maybe they both have the same genetic mutation in two different kinds of cancer –  so the same treatment can be used?

biomarkerThis is where biomarkers come in.  Biomarkers are things in the body that can be measured to give us information about a disease or other condition.  Biomarkers can be a variety of things including

  • Imaging methods
  • Genes (presence or absence)
  • Specific gene mutations
  • Proteins or antibodies
  • Metabolites
  • Microbes?

And these things can be measured to in some way indicate if the person is healthy or has a disease.  Other biomarkers may be used to detect a disease earlier than when the patient is showing symptoms.  Detecting a disease early may allow the patient to change a behavior to decrease the likelihood of developing the disease or to start treating a patient earlier when it is easier to successfully treat a disease. Biomarkers may also be used to determine the severity of a disease or whether or not the disease is progressing.

biomarker_types
Some biomarkers that you may be familiar with are cholesterol, temperature, and blood pressure.  There are a number of biomarkers for pregnancy. Home pregnancy tests look for the presence of the protein beta human chorionic gonadotropin (also called beta-HCG) in the urine.  This protein biomarker is in the blood after the zygote implants 6-12 days after fertilization.  Other biomarkers such as serum creatinine and liver enzymes are markers for kidney and liver function, respectively.

So what makes a good biomarker?  First, it needs to be different if the patients has a disease.  For example, higher than normal blood glucose levels may indicate that a patient has diabetes and these levels of blood glucose would not be found in a patient who didn’t have diabetes.  Second, the biomarker would have to correlate with the outcome.  What this means is that as the patient’s condition changes, the biomarker would also change. In the case of the patient with high blood glucose and diabetes, when the patient starts regulating their diet or taking insulin, the blood glucose levels will go down.  Third, biomarkers should be easy to access, and one of the main reasons for this is so that testing for the biomarkers isn’t too expensive.  Blood is a common location for biomarkers, including in our example of blood glucose levels.  Finally, biomarkers should be consistent.  It wouldn’t be useful to have a biomarker that changes based on whether it’s noon or midnight.  It needs to be dependent on the health of the patient.

Biomarkers are a cornerstone of personalized medicine because they allow clinicians to use symptoms along with measurable and quantifiable factors in the body (the biomarkers) to diagnose, track, and treat disease. Learn more about biomarkers in this YouTube video