Correspondence: Philipp Dahm, Department of Urology, College of Medicine, University of Florida, Health Science Center Box 100247, Room N-203, Gainesville, FL 32610-0247, USA.e-mail: firstname.lastname@example.org
What's known on the subject? and What does the study add?
Free to total PSA ratios are commonly used as an adjunct to total PSA levels to better define an individual's risk for prostate cancer; however, its strengths and weaknesses are not well understood.
This article illustrates the use of likelihood ratios that can be generated from the reported sensitivities and specificities from given free to total PSA thresholds in either increasing or decreasing an individual patient's probability of prostate cancer. Understanding the strengths and limitations of free to total PSA testing will help clinicians anticipate whether its use is indicated or not.
A 59-year-old Caucasian male patient is referred to your clinic with a serum PSA concentration of 6.0 ng/mL and a benign prostate examination. You have an extended discussion with him about our current understanding of the benefits and harms of prostate cancer screening during which he raises the question whether a ‘free PSA test’ might help him inform the need for a biopsy. You indicate to him that it has not been your practice to obtain free PSA levels and that a biopsy would be more definitive. However, you remember few details about the performance characteristics of a free PSA test. You therefore offer to review the current best evidence on this question and commit to calling him over the telephone to discuss the test when you have more information.
Using the PICO format, which stands for Patient, Intervention, Comparison and Outcome , you formulate the clinical question you are planning to answer: ‘In men with an elevated PSA level of 4–10 ng/mL, what is the diagnostic accuracy of a free PSA when compared with TRUS biopsy in detecting prostate cancer?’
Finding the Best Evidence
To find up-to-date and evidence-based information on this question you turn to DynaMed (http://dynamed.ebscohost.com), a comprehensive resource of pre-appraised evidence . There you find the statement ‘use of percentage of free PSA can reduce unnecessary biopsies with minimal loss in sensitivity in detecting cancer’ from a reference, which appears to exactly address your question . However, details about sensitivity, specificity and likelihood ratios are not provided; you therefore decide to retrieve the full-text study to appraise it in detail.
The study is titled ‘Use of the Percentage of Free Prostate-Specific Antigen to Enhance Differentiation of Prostate Cancer from Benign Prostatic Disease: A Prospective Multicenter Clinical Trial’  and reports the results of a prospective study using Tandem PSA and free PSA assays in 773 men aged 50–75 years with palpably benign prostate gland and a PSA level of 4–10 ng/mL, of which 379 had prostate cancer, and 394 had benign prostatic disease.
Before assessing the value of the free PSA test result, you should estimate the patient's pre-test probability for prostate cancer. You enter the patient's risk profile (PSA level, DRE findings, ethnic background, no prior history of biopsies, no family history of prostate cancer) into an online calculator (http://deb.uthscsa.edu/URORiskCalc/Pages/calcs.jsp) based on the prostate cancer prevention study , which estimates the patient's risk for harbouring prostate cancer at 43% (95% CI 41%–47%).
Sensitivity and Specificity
For patients aged 50–59 years, the study reports a sensitivity of 98% (95% CI 93%–100%) and a specificity of 11% (95% CI 6%–19%) with a 25% free PSA threshold. The 25% threshold was determined in an effort to have 95% sensitivity in the study population. In other words, the authors wanted a negative test (free PSA >25%) to help exclude prostate cancer. You take note however, that this study did not have the typical design of a diagnostic test study in which patients have diagnostic uncertainty about whether they do or do not have prostate cancer when the test is performed (in this case: free PSA). The retrieved study compares two clearly defined groups with and without prostate cancer. This raises concern that the study sample may not be applicable to clinic patients where indeterminate cases are included. The potential for this ‘spectrum bias’  may mean that the performance of free PSA in unselected clinic patients might be different than, typically inferior to, what is reported in the study.
Likelihood Ratio (LR)
Acknowledging the limitations of the study, you decide to use the reported sensitivity and specificity to calculate the LRs associated with the test when a threshold of 25% is used (Appendix). The positive LR (LR+) and negative LR (LR−) allow you to determine the change in the probability of a diagnosis based on a positive or negative diagnostic test result, respectively . While rarely reported in the urological literature, it can readily be calculated from the information that is provided either in the form of the associated sensitivity and specificity or data provided in a 2 × 2 table. In this scenario, the calculated LR+ (PSA ≤25%) is 1.1 and a LR− (PSA >25%) is 0.18.
Case Resolution – Applying the Results to the Care of Your Patient
At this point, you have the pre-test probability (0.43) based on the patient's characteristics, along with the LR+ (1.1) and LR− (0.18), which you can apply to a Fagan nomogram (Fig. 1). The post-test probability of having prostate cancer with a free PSA level of ≤25% (positive test) is ≈45%, which is essentially unchanged from the pre-test probability. The post-test probability of having prostate cancer with a free PSA level of >25% (negative test) is 12%, which is much lower.
Based on these calculations, it becomes apparent to you that when a free PSA level is >25%, it can provide reassurance that a prostate biopsy may not be needed, whereas a free PSA level of ≤25% does little to inform risk. With a free PSA level of >25%, the patient's decision to forego a biopsy will still hinge on his willingness to accept a residual risk of 12%; if that risk is still too high, there is no value in ordering this extra test and instead the patient may skip this test and choose to have a prostate biopsy.
This study shows the value of free PSA using a threshold of 25%. As this example exemplifies, it is most helpful as a diagnostic test when the result is >25%. Estimating the post-test probability for a given patient, if a given test were to be positive or negative, allows you to make a rational decision of whether or not a free PSA test will be helpful.
Conflict of Interest
Source of Funding
Funding provided is Departmental.
Appendix: Appendix: Calculation of LR for Percentage Free PSA and Prostate Cancer