I was diagnosed in March of 2018 with bone Mets on almost every bone. Starting PSA was 380 ish. I have been getting treated at MD Anderson from the beginning. Have had ADT, Taxotere, Zytiga, Apalutamide, Cabazitaxel, & carboplatin Combo. Show
My PSA got down to 3.9 and stayed there for one month and had been climbing for over a year. Looks like it is leveling off at 250 ish. My MO will not comment on my life expectancy, he just says each case is different. But I feel like I am running out of time based on the reading that I do on my own. Can anyone give me advice on what I should do? Or examples of cases that were able to turn their PSA around after such a swing as I have had? me. For those that have had Taxotere how long did you see your drop in PSA last after your last... Treated with Zytiga for 13 months. During my first 11 months the PSA has followed a decline in each... 56 years old, PSA =5.8, RP in July 2018 Gleason 3+4 (although original biopsy said Gleason 9!)... ago I had blood work done at my MO and PSA was at 0.09 . This week my urologist also did a PSA...
BackgroundMeasurement of blood concentration of PSA is a test that can identify men who have an increased probability of having an undiagnosed prostate cancer and, as a result, may identify cancers at a stage at which they are more likely to be curable than if they presented clinically. However, tests aimed at diagnosing cancer early are never perfect. Some fraction of tests done will produce false positive results, prompting diagnostic tests, usually invasive, that do not find cancer to be present. Some, perhaps most, tests for early cancer also bring to light some cancers that would otherwise never have become clinically evident in the patient’s lifetime. From a histopathological point of view, these are real cancers but they are either progressing slowly or not at all, such that, if left, they would have never bothered the patient. They are commonly referred to as overdiagnosed cancers and their detection by tests for early diagnosis of cancer is referred to as over-diagnosis. False positive tests and over-diagnosis both cause some harm, which varies from minor discomfort occasioned by conduct of a biopsy to death in the rare case, for example, that a man with an overdiagnosed cancer dies as a result of complications of surgery aimed at curing it. In making decisions about PSA testing, the balance of the anticipated benefit – better health and extension of life due to early diagnosis – against the inevitable harm must always be taken into consideration. It is of paramount concern in this section of the guideline. Strategies for PSA testing vary according to the age at which testing commences and ceases, the interval between tests, and the PSA threshold for further investigation (e.g. biopsy of the prostate). Protocols currently in use in Australia and elsewhere differ in all these variables. Simple evaluative measures, such as a higher cancer detection rate, a shift in the stage distribution of cancer towards earlier stages or longer survival of people whose cancer was detected using the test, cannot be used to infer that testing achieves a better outcome from the cancer. Only demonstration of a reduction in mortality from cancer in people to whom the test is applied can provide certainty as to its efficacy. Randomised controlled trials are the only way in which such a reduction can be demonstrated confidently. In principle, they also provide the best evidence as to the extent of the associated harm. A systematic review of the available randomised controlled trials was the primary source of evidence used to answer PICO question 3.1. Rigorous comparison of the performance of a range of different PSA testing strategies (e.g. with different age at testing, test interval, or biopsy criteria) to identify the optimal testing protocol would require many large randomised controlled trials with long follow-up periods. Since it is unlikely that such studies will be done, mathematical models have been developed that use information gained from the randomised controlled trials and other research to predict outcomes, both beneficial and harmful, of testing strategies that the randomised controlled trials have not evaluated specifically. We therefore also undertook a systematic review of relevant modelling studies to assist in answering PICO question 3.1. If it is accepted, on the basis of evidence from randomised controlled trials, that a test such as the PSA test is able to deliver the desired outcomes, studies of comparative test performance (e.g. sensitivity, specificity, and positive predictive value) are useful in evaluating different approaches to achieving the desired outcomes. Such studies were used to provide evidence that might assist in answering PICO question 3.2, and have been used in a later section to assess the likely benefit or harm from adding DRE to PSA testing in deciding which men are at high risk of having a cancer that is not yet causing symptoms. Once an efficacious test for early diagnosis of cancer is in widespread use in the community, observational epidemiological studies may be useful in evaluating its effectiveness in practice and in considering ways and means of improving its performance and achieving the best balance of benefits to harms. Such studies, however, are prone to a range of biases and should not be the primary basis for deciding whether or not to use such a test in the first place. Observational epidemiological studies were the main source of evidence reviewed for PICO question 3.3. Back to top EvidenceEffect of testing strategies on rates of prostate cancer-specific death and metastases at diagnosisProstate cancer death reported in randomised controlled trialsFour randomised controlled trials[1][2][3][4][5][6][7][8] and one pseudo-randomised trial[9][10] were identified that investigated whether prostate cancer mortality is reduced by PSA testing in men without a prostate cancer diagnosis or symptoms that might indicate prostate cancer. Three were judged to be at moderate risk of bias (the European Randomized Study of Screening for Prostate Cancer [ERSPC],[8] the Prostate, Lung, Colorectal and Ovarian Cancer Screening Trial [PLCO][3] and the Norrköping Randomised Controlled Trial of Prostate Cancer Screening[9]), and two were judged to be at high risk of bias (screening studies conducted in Stockholm[2] and Quebec[6]). The search strategy, inclusion and exclusion criteria, and quality assessment are described in detail in the Technical report. The largest of the trials was ERSPC,[8] a multicentre trial with seven centres. It found, in men aged 55–69 years, that PSA testing every 2–4 years (mostly without DRE and using a PSA level of > 3.0 ng/mL as an indication for biopsy), reduced prostate cancer-specific mortality compared with no testing (in reality background levels of testing): relative risk (RR) 0.79; 95% confidence interval (CI) 0.68–0.91 at a median of 11 years’ follow-up. The other four trials[10][2][3][6] reported RRs of 1.01–1.16 at follow-up of 8–20 years. The most recent of these and by far the largest, the PLCO,[3] reported an RR of 1.09 (95% CI 0.87–1.36). The five studies summarised above were also included in a contemporary meta-analysis of trials of PSA testing for prostate cancer.[11] The authors reported a summary relative risk of death from prostate cancer in men randomised to PSA testing of 1.00, 95% CI 0.86–1.17. They concluded that a pooled meta-analysis of the five included studies in this review identified that screening did not significantly decrease prostate cancer-specific mortality and is associated with a high degree of over-diagnosis, treatment and screening-related harms. They noted the overall heterogeneity in quality and study design of the five studies and gave greater weight to the four studies that did not find evidence of reduction in prostate cancer mortality than to the one study that did (ERSPC) in framing their conclusion. Taken together, the results of the PLCO,[3] Norrköping,[10] Stockholm[2] and Quebec[6] trials are statistically incompatible with those of the ERSPC[8], either as used in the 2013 meta-analysis[11] (PLCO results from Andriole et al 2009[12] and ERSPC results from Schroder et al 2009[13]) or when updated with further experience of PLCO[3] and ERSPC[14]. A fixed effects meta-analysis of the PLCO, Norrköping, Stockholm and Quebec trial results from Figure 2 of Ilic et al (2013)[11], the four-studies’ results to which Ilic et al gave greater weight in reaching their conclusion, gives an RR of 1.09, 95% CI 0.94–1.27 (p-value for heterogeneity among studies 0.91) for the risk of prostate cancer death in those offered testing relative to those not offered testing. This result compares with an RR of 0.84, 95% CI 0.73–0.95 from the ERSPC 2009 results as included in Ilic et al[11]. Note that the upper 95% confidence bound of the ERSPC estimate just overlaps the lower 95% confidence bound of the pooled four-studies results. Moreover, if the ratio of the four studies RR to the ERSPC RR is calculated, using the method of Altman et al[15], the value obtained is 1.30, 95% CI 1.06–1.58, which provides clear evidence that the results of the four studies are not statistically compatible with the ERSPC results. If we use the 2012 results of PLCO and ERSPC in these calculations instead of the 2009 results, the incompatibility is greater: the four studies RR of death from prostate cancer in those offered testing compared with those not offered testing becomes 1.08, 95% CI 0.94–1.24, the ERSPC 2012 result is 0.79, 95% CI 0.68–0.91. The lower 95% confidence bound of the former does not overlap the upper bound of the latter and the ratio of the two is 1.37, 95% CI 1.12–1.67, which provides strong evidence against the identicality of the two RR estimates. Based on the above evidence that the results of the four studies and the results of the ERSPC are statistically incompatible, to proceed with formulation of a guideline for PSA testing the Expert Advisory Panel was constrained to assume that either the four studies were correct, or that the ERSPC was correct. The Panel preferred the ERSPC for the following reasons.
Should further research find that the ERSPC results are more unreliable than the Panel has judged them to be, it would have to reconsider its decision to prefer the evidence of the ERSPC and therefore this guideline. In this context, it is relevant to note that the ERSPC published results up to 13 years of follow-up (previously 11 years) after the last date for the literature searches that contributed to the systematic reviews for this guideline.[18] Key features of the results summarised above, which are based on 11 years of follow-up, and those based on 13 years of follow-up are shown in Table 2.1. Table 2.1. Summary of results of ERSPC study up to 11 years (as used for this guideline) and up to 13 years (published after last date of systematic review searches) in the core age group (55-69 years)
Sources: Schroder et al (2012)[8], Schroder et al (2014)[18]
Back to top Metastases at diagnosis reported in randomised controlled trialsThree trials (ERSPC,[14] PLCO[3] and the Norrköping[9] trial) considered metastatic prostate cancer at diagnosis as a trial outcome. Two of these trials reported a lower risk of metastatic prostate cancer at diagnosis in the intervention arm than in the control arm:
Systematic PSA testing in men without prostate cancer or its symptoms was not associated with reduced risk of metastatic prostate cancer at diagnosis in the Norrköping trial[9] (RR 1.12; 95% CI 0.63–1.99). In this trial, testing began at age 50 years and continued every 3 years for 12 years. The first two tests consisted of DRE alone, and the third and fourth test included the combination of DRE and PSA testing (with PSA > 4.0 ng/mL as the indication for biopsy). Overall, there is moderately consistent evidence that PSA testing, according to the range of strategies used in these trials, reduces the incidence of metastatic prostate cancer at diagnosis. The lower RR seen in the ERSPC trial,[14] compared with the PLCO[3] and Norrköping[9] trials, might indicate superiority of the PSA testing strategies used in the four ERSPC component studies analysed, which differed from the PLCO[3] and Norrköping[9] trials mainly in use of a PSA threshold for biopsy of > 3.0 ng/mL, not > 4.0 ng/mL. Back to top Interpreting the randomised controlled trial findingsGiven that greater reliance is being placed on the finding of the ERSPC[14], and that this trial showed a benefit for systematic PSA testing in men without prostate cancer or its symptoms, detailed consideration was given to the protocols followed to gain the observed effect. While the ERSPC centres varied in the detail of their testing protocols, they shared the following features:
Therefore, ERSPC results can be taken as indicative of the outcome of a policy of 2- to 4-yearly testing of men aged 55–69 years, referring men for biopsy when total PSA was > 3.0 ng/mL, and ceasing testing at age 70–75 years. While the published results of different ERSPC centres generally give little indication of consistent variation in effect due to variation in the testing protocol, the results from the Goteborg centre, which differed in offering testing at 2-year intervals from age 50 years, suggest that an earlier start and more frequent testing might be preferable to testing at 4-year intervals from age 55. In addition, in an all ages analysis of the ERSPC (Schroder et al 2012, Supplementary Appendix Table 5A), there was nothing to suggest efficacy of testing in men 70+ years of age (RR 1.18, 95% CI 0.81–1.72), although the confidence interval was wide. Back to top Modelling studiesIn addition to the evidence from randomised and pseudo-randomised controlled trials, three modelling studies[16][19][20][21] met the inclusion criteria for this review. They were studies in which participants had no history of prostate cancer or symptoms that might indicate prostate cancer at baseline (or that used state-transition models), and which compared two or more PSA testing strategies and reported benefits (e.g. prostate cancer-specific mortality, lives saved from prostate cancer or incidence of metastatic cancer at diagnosis) and harms (e.g. false positives or over-diagnoses of prostate cancer). All three modelling studies were in English and published before 1 March 2014 (see Technical report). One study was based on the MISCAN model of cancer screening[20][21] and two were based on the Fred Hutchinson Cancer Research Center (FHCRC) microsimulation model of prostate cancer screening.[16][19] None of these studies was developed and calibrated for the Australian context, or validated in Australia. The MISCAN model was based on the Dutch population and calibrated mainly to Dutch and other European data, and levels of participation in testing were assumed to be 100%[19] and 80%.[20] The FHCRC studies were based primarily in the US population and were calibrated to US data, although one study[19] used initial treatment data for British Columbia, Canada. While not explicitly stated, it appears that both assumed 100% screening participation. Their simulated populations were, respectively, men with age distribution according to the European Standard Population,[21] men aged up to 100 years with age distribution according to the European Standard Population,[20] contemporary men in the USA aged 40 years,[16] and men in British Columbia aged 40 years.[19] Each model was expertly assessed as to its strengths and limitations across the domains of specifications, natural history, screening or triage recommendations and behaviours, diagnostic pathways, invasive cancer (survival, treatment) and costs (reference to rating scale). The strengths of both models, which included well-documented and relevant data sources and independent validations, were considered to outweigh their limitations, such as inadequate sensitivity analyses. As such, both models were found to adequately simulate prostate cancer incidence and mortality, with the caveats that neither model incorporated realistic screening behaviours (80% or 100% participation was assumed) and that the health outcomes presented for the MISCAN prostate cancer model were not adequately discounted in the assessment of quality-adjusted life years gained or lost. Back to top Modelling to predict effect of testing protocols on outcome death from prostate cancer and balance of benefits and harmsTables 2.2–2.4 describe the 47 different PSA testing protocols, with more than one protocol modelled in each of the three studies, and present the following outcomes:
These modelled outcome estimates provide a basis for selecting the protocol that, on present evidence, achieves the best balance between benefits and harms of PSA testing. Prevention of death from prostate cancer – the primary aim and main benefit of testing – is indicated by the probability that prostate cancer death is prevented. The harm to men who are tested is indicated by the probability of one or more false positive PSA tests and the probability of having an overdiagnosed cancer. ‘Mean months of life gained per man diagnosed’ measures the balance of benefit (life gained) to harm (over-diagnosis) as does, inversely, the ratio ‘number of men overdiagnosed with prostate cancer per prostate cancer death prevented’, which has been added in Table 2.4. Mean months of life gained per man diagnosed can also be interpreted as the expectation of life gained by each man diagnosed with and treated for prostate cancer as a result of PSA testing. It is strongly influenced by the probability of over-diagnosis; the more men there are over-diagnosed the more there are to ‘share’ the expectation of extension of life with men who actually experience the extension due to early diagnosis and treatment of a cancer that would otherwise have killed them. To assist in assessing the trade-offs between these outcomes, the testing protocols have been sorted in descending order by the probability that prostate cancer death is prevented. In addition, the testing protocol most like that of the ERSPC has been highlighted in each table to provide a directly evidence-based reference point with which to compare the possible alternative protocols. Back to top Making protocol choicesTable 2.2 summarises the three alternative protocols based on the MISCAN model.[20] A change from 4-yearly to annual testing in this model predicts a 50% increase in probability of prevention of death from prostate cancer which is accompanied by a 22% increase in men with more than one false positive, a 55% increase in probability of over-diagnosis and a minimal fall in mean months of life gained per man diagnosed. Thus, the increase in benefit from the increase in testing frequency would appear to outweigh the additional harm. Table 2.3 summarises protocols from the Pataky et al (2014)[19] model. Broadly it suggests that all protocols with higher probability of prevention of death from prostate cancer (up to 27% higher) achieve that at the cost of an increase in the percentage of men with more than one false positive, an increase in the probability of over-diagnosis and a reduction in means months of life gained per man diagnosed. Protocol 29 is an exception, however, where addition of testing in men 70–74 years, using a criterion for further investigation of 4.0 ng/mL instead of 3.0 ng/mL, is accompanied by a higher probability that death from prostate cancer is prevented, a fall in the percentage of men with more than one false positive, a fall in the probability of having an overdiagnosed prostate cancer and quite a small fall in mean months of life gained per man diagnosed. Table 2.4 summarises the much larger number of protocols examined by Gulati et al (2013).[16] The most notable feature of these protocols is that use of > 95th percentile of PSA for age as the criterion for further investigation in place of a PSA > 4.0 ng/mL, with age range for testing and frequency of testing held constant, consistently results in a lower percentage of men with one or more false positive tests, a lower probability of having an overdiagnosed cancer and an appreciably higher mean months of life gained per man diagnosed, but with some reduction in the probability that death from prostate cancer is prevented. Therefore, there is a clear trade-off of reduction in benefit for reduction in harm with the use of > 95th percentile of PSA for age as the criterion for further investigation, but the generally high levels of mean months of life gained per man diagnosed when using these protocols suggest they may have a net beneficial effect. Thus, use of the > 95th percentile for age as the criterion for further investigation might be considered. If we consider the ERSPC results as providing the best empirical evidence of which PSA testing protocol (if any) is efficacious in reducing mortality from prostate cancer, then we are left making choices between 55 and 50 years as the age at which to first offer a man PSA testing, offering testing at intervals of 4 or 2 years and ceasing to offer testing at 70 or 75 years of age. To aid in these choices we have extracted from Tables 2.2 to 2.4 comparisons of protocols that provide, most directly, the information we need to make those choices; these comparisons are in Table 2.5. In addition, to aid in the comparison, we have added to Table 2.5 comparative data for each pair of compared protocols, namely the difference in the percent of men having ≥ 1 false positive test and having an overdiagnosed cancer, difference in the percent of men having death from prostate cancer prevented, difference in the mean months of life gained per man diagnosed and the number of extra overdiagnosed cancers diagnosed per extra prostate cancer death prevented in going from the “less aggressive” (listed first in the pair) to the “more aggressive” protocol (listed second). Back to top Beginning testing at 55 or 50 years of ageOnly Pataky et al offer a comparison between a protocol beginning at 55 years of age and a protocol beginning at 50 years of age (Table 2.5), and in this comparison a change in testing frequency, from every 4 years to every 2 years, accompanies the change in age. Thus, while an unambiguous comparison between starting ages of 55 years and 50 years is not possible, the comparison made is advantageous because it compares the Goteborg protocol (starting at age 50 years and offering testing every 2 years) with the protocol followed by the other ERSPC centres (starting at age 55 and testing every 4 years). In summary, the Pataky et al model estimates that a change in starting age from 55 years to 50 years and an increase in testing frequency from every 4 years to every 2 years increases the probability of >1 false positive by 3.6%, increases the probability of over-diagnosis by 1% increases the number of prostate cancer deaths prevented by 18 per 10,000 (0.18%) and reduces the mean months of life gained per man diagnosed by 10.2 months. The number of extra overdiagnosed prostate cancers per extra prostate cancer death prevented is estimated at 5.6. It is not possible, in this comparison, to say whether this higher cost in overdiagnosed cancers is mainly due to the change in age, the change in frequency of testing or largely shared between the two. Examination of the effects of change in frequency (4 years to 2 years) in Table 2.5, however, suggests that the change in age may be the dominant factor. Either way, this protocol change has, with the separately assessed change from testing every 4 years to testing every 2 years, the best balance of additional benefit to additional harm of the protocols compared in Table 2.5. While the reduction in mean months of life gained per man diagnosed, 10.2 months, is quite high, the mean months of life gained per man diagnosed for the protocol starting at 50 years of age, 34.1, remains reasonably high. Back to top Extending testing from 69 to 75 years of ageThe three relevant protocol pairs closest to the ERSPC protocol are summarised in Table 2.5. The pairs differ only in their PSA criteria for further investigation. Each protocol pair showed modest increases in the probabilities of ≥ 1 false positive test (3% to 6%), over-diagnosis (1.1% to 1.8%), and prostate cancer death prevented (13 to 20 per 10,000) when going from the cessation of testing at 70 to cessation at 75 years of age (the more aggressive option). The numbers of extra over-diagnosed cancers per prostate cancer death prevented, however, were high, 7 to 9, and are reflected in appreciable falls in the mean months of life gained per man diagnosed, -9.1 to -18.7, to comparatively low absolute levels, 22.1 to 29.1. Back to top Testing every four years or every two yearsThe one model[19] that reported the impact of change in testing interval from 4 years to 2 years (in men aged 50–74 years, not 50–69 years) showed only small effects of the change. The proportion of men with ≥ 1 false positive test increased 0.7%, those with an over-diagnosed cancer also increased 0.7%, and there was a moderate increase in probability that prostate cancer death is prevented, 13 per 10,000 (Table 2.5). These results translate into in an estimated 5.4 extra over-diagnosed cancers per extra death from prostate cancer prevented by the change to the shorter interval. There was, however, little change, -0.5, in the mean months of life gained per man diagnosed. It appears, therefore, that the increase in prostate cancer deaths prevented by using a 2-year interval rather than a 4-year interval is well balanced against the increase in harm from false-positive PSA tests and over-diagnosis of prostate cancer. Back to top Beginning testing at age 40 yearsWhile not raised by variability in the ERSPC protocol, whether to offer testing first at 40 years of age (to obtain a PSA-based estimate of later risk of prostate cancer or to initiate regular testing) is a live issue. Gulati et al evaluated four protocols in which outcomes of testing from 50-69 and 40-69 years of age were compared at two different PSA criteria for further investigation, > 4 ng/mL and > 2.5 ng/mL (Table 2.5). For protocols testing men aged 40–69 years, the key outcomes (the probabilities of one or more false positive tests, over-diagnosed cancer, and prostate cancer death prevented, and the mean months of life gained per man diagnosed), were generally similar to those for protocols testing men aged 50–69 years. The increase in the probability that prostate cancer death is prevented by beginning testing at 40 years was small, at 2 to 3 in 10,000, and there were 5-7 extra overdiagnosed cancers per death prevented. In addition, because the increase in underlying prostate cancer mortality over 10 years from age 45–49 (7.98 per 100,000) is three times greater than that from age 40–44 (2.34 per 100,000), most of the small extra benefit would be gained by testing from age 45 (Table 2.7). Back to top Heijnsdijk et al (2009)[21] modelled the effects of different test protocols on initial treatments, including palliative therapy, which can be taken as an indicator of metastatic disease present at the time of diagnosis. Relative to no testing, testing every 4 years from ages 55 to 70 years using a PSA threshold of 3.0 ng/mL resulted in a reduction of 2.1 men per 1,000 with metastatic disease at diagnosis at a cost of 150 unnecessary biopsies per 1000 men tested. With testing from 55 to 75 years every 4 years, the reduction in metastatic disease at diagnosis was 3.0 men per 1000 at a cost of 230 unnecessary biopsies per 1,000 men tested; and with testing at 55-70 years and a testing interval of 1 year, the reduction in metastatic disease at diagnosis was 2.6 men per 1,000 at a cost of 185 unnecessary biopsies per 1000 men tested. Expressed in approximately equivalent terms to those of Table 2.3, increasing the frequency of testing from four-yearly to yearly increases the probability that diagnosis with metastatic prostate cancer is prevented by 0.06 percentage points (0.6 per 1,000) at a cost of increasing the probability of having an unnecessary biopsy by 3.6 percentage points, and extending the age range for testing to 75 years increases the probability that diagnosis with metastatic prostate cancer is prevented by 0.09 percentage points (0.9 per 1,000) at the cost of increasing the probability of having an unnecessary biopsy by 8.0 percentage points. Table 2.2. Modelled outcomes of a range of PSA testing protocols sorted in decreasing order of probability of death from prostate cancer prevented for protocols reported by Heijnsdijk et al 2012
Source: Heijnsdijk et al (2012)[20] The protocol that most closely approximates the ERSPC testing strategy is shown highlighted. The protocols above it appear to perform relatively better in preventing death from prostate cancer. ~ Approximately FP: false positive *Outcomes were calculated as follows: Probability of ≥ 1 FP % = percentage of men having one or more false positive tests over the age range of testing Probability of over-diagnosis % = percentage of men having an over-diagnosed prostate cancer during the age range of testing Probability that prostate cancer death is prevented % = percentage of men prevented from dying from prostate cancer from date of first testing to age 100 years[20] Mean months of life gained per man tested = total months of life gained by men prevented from dying from prostate cancer averaged over all men tested NND = Number of men needed to diagnose and treat for prostate cancer to prevent one death from prostate cancer (probability of over diagnosis % divided by the probability that death from prostate cancer is prevented %) Mean months of life gained per man diagnosed = Mean months of life gained per man whose death from prostate cancer was prevented by testing divided by the NND (calculated as mean months of life gained per man tested divided by probability that prostate cancer death is prevented % multiplied by 100 and the result divided by the NND). † Modelled protocols from all models were ranked in order of decreasing probability that prostate cancer death was prevented § Heijnsdijk et al (2012)35 did not provide an estimate of this value. It was estimated by using the following approach: life years gained (undiscounted) per 100 men tested multiplied by 12 and divided by 100.‡ Protocol 28 approximates the testing strategy used in the intervention arm of ERSPC.[8] Table 2.3. Modelled outcomes of a range of PSA testing protocols reported by Pataky et al 2014, sorted in decreasing order of probability of death from prostate cancer prevented
Source: Pataky et al (2014)[19] The protocol that most closely approximates the testing strategy used by the ERSPC is shown highlighted. FP: false positive *Outcomes were calculated as follows: Probability of ≥ 1 FP % = percentage of men having one or more false positive tests over the age range of testing Probability of over-diagnosis % = percentage of men having an over-diagnosed prostate cancer during the age range of testing Probability that prostate cancer death is prevented % = percentage of men prevented from dying from prostate cancer from date of first testing to age 9034 Mean months of life gained per man tested = total months of life gained by men prevented from dying from prostate cancer averaged over all men tested NND = Number of men needed to diagnose and treat for prostate cancer to prevent one death from prostate cancer (probability of over diagnosis % divided by the probability that death from prostate cancer is prevented %) Mean months of life gained per man diagnosed = Mean months of life gained per man whose death from prostate cancer was prevented by testing divided by the NND (calculated as mean months of life gained per man tested divided by probability that prostate cancer death is prevented % multiplied by 100 and the result divided by the NND). ‡ Protocol 32 approximates the testing strategy used in the Gøteborg centre of the ERSPC § Pataky et al (2014)[19] did not provide an estimate of this value. It was estimated by using the following approach: life years gained (undiscounted) per 100 men tested multiplied by 12 and divided by 100. Table 2.4. Modelled outcomes of a range of PSA testing protocols reported by Gulati et al 2013, sorted in decreasing order of probability of death from prostate cancer prevented
Source: Gulati et al (2013)[16] The protocol that most closely approximates the protocol used by the ERSPC is shown highlighted. FP: false positive vPSA: PSA velocity *Outcomes were calculated as follows: Probability of ≥ 1 FP % = percentage of men having one or more false positive tests over the age range of testing Probability of over-diagnosis % = percentage of men having an over-diagnosed prostate cancer during the age range of testing Probability that prostate cancer death is prevented % = percentage of men prevented from dying from prostate cancer from date of first testing to the end of life31 Mean months of life gained per man tested = total months of life gained by men prevented from dying from prostate cancer averaged over all men tested NND = Number of men needed to diagnose and treat for prostate cancer to prevent one death from prostate cancer (probability of over diagnosis % divided by the probability that death from prostate cancer is prevented %) Mean months of life gained per man diagnosed = Mean months of life gained per man whose death from prostate cancer was prevented by testing divided by the NND (calculated as mean months of life gained per man tested divided by probability that prostate cancer death is prevented % multiplied by 100 and the result divided by the NND). † Modelled protocols from all models were ranked in order of decreasing probability that prostate cancer death was prevented §95th percentiles were 2.5, 3.5, 4.5 and 6.5 ng/mL for ages 40–49, 50–59, 60–69 and 70–74 years, respectively. ‡ Protocol 28 approximates the testing strategy used in the Gøteborg centre of the ERSPC[8] Table 2.5. Comparisons of outcomes of testing using different ages at testing (55–69 years or 50–69 years; 50–69 years or 50–74 years; 50–69 or 40–69 years) and different intervals between tests (4 years or 2 years) with the PSA criterion for investigation and the other PSA testing protocol components (interval between tests or age at testing) held constant
†Criterion for biopsy but not interval between tests held constant. Data source: Pataky et al (2014)34. ††Interval between tests and criterion for further investigation held constant. Data sources: Gulati et al (2013)31 and Pataky et al (2014)34. ‡Age and criterion for further investigation held constant. Data source: Pataky et al (2014)34. ‡‡Interval between tests and criterion for further investigation held constant. Data source: Gulati et al 201331. *Model results for ages 50–74 years are presented because results for 50–69 years have not been reported. †No additional protocols that would permit PSA testing interval to be held constant. Sources: Gulati et al (2013)[16], Pataky et al (2014)[19] (Data extracted from Tables 2.3 and 2.4 to facilitate the comparisons.) Back to top Effect of different testing strategies on rates of biopsy-diagnosed prostate cancerTo examine and quantify the effect of different testing strategies on rates of biopsy-diagnosed prostate cancer, a systematic review was done that encompassed studies of men with no history of prostate cancer who had undergone a prostate biopsy less than 1 year after a PSA test and were participants in a prostate cancer screening RCT or in an NHMRC level of evidence III-2 or higher fully paired diagnostic performance study that permitted comparison of the diagnostic performance of two or more different PSA thresholds ≤4.1ng/mL or two different prostate cancer screening protocols, and achieved specified minimum levels of diagnostic confirmation and results reporting. Seven level III-2 diagnostic performance studies met the inclusion criteria.[22][23][24][25][26][27][28] All were at moderate risk of bias. In addition results from an analysis of relevant ERSPC data[29] have been included for comparative purposes only; it did not meet all inclusion criterion as only men with an elevated PSA were biopsied and the biopsy was a sextant biopsy. In one study, the placebo arm of the Prostate Cancer Prevention Trial,[26] men were biopsied regardless of PSA level or DRE, enabling comparisons of sensitivity and specificity at different PSA thresholds. In this study, men with a normal DRE and PSA levels at baseline were tested annually for 7 years and offered a sextant biopsy at the end of the trial.[26] Potential verification bias was considered and shown not to be an issue.[26] The remaining studies were able to provide estimates only of increases in numbers of cancers detected and numbers of unnecessary biopsies with decreasing PSA thresholds.[22][23][24][25][27][28][29] In six of these studies all men underwent prostate biopsy if their PSA levels exceeded specified thresholds. Participants were diverse, ranging from men with lower urinary tract symptoms to asymptomatic participants in population-based screening programs.[22][23][25][27][28][29] In the remaining study, all men with a family history of prostate cancer and a PSA below a specified PSA threshold underwent prostate biopsy.[24] The published studies did not describe how the PSA assays used were calibrated. For two studies, World Health Organization (WHO) calibration could be inferred from information available on the assay (Izotope) manufacturer’s website.[23][27] Two studies did not report the PSA assay used.[22][28] Only one study compared yields stratified by Gleason score at different PSA thresholds.[26] Comparisons between studies in terms of absolute numbers were limited due to differing biopsy protocols, populations and PSA assays and their calibration. Therefore, this review focuses on the effects of varying thresholds within studies. In all studies, lowering the PSA threshold increased cancer detection at a cost of increased unnecessary biopsies[22][23][24][25][26][27][28][29] In six of the eight studies, the ratio of false positives to true positives increased as the PSA threshold changed from 4.0 ng/mL to 3.0 or 2.5 ng/mL (Figure 2.1). In two studies in which lower PSA levels were assessed, the ratio of false positives to true positives increased more rapidly as the threshold was reduced from 3.0 ng/mL to 2.0 ng/ml, and even more rapidly again as it was reduced from 2.0 ng/mL to 1.0 ng/mL. The ratio of false positives to true positives varied across the studies from 1.1 to 4.2 at a PSA threshold of 4 ng/mL (Figure 2.1). Lowering the PSA threshold from 4.0 ng/mL to 3.0 ng/mL resulted in estimates of 2.17–3.77 additional unnecessary biopsies for every additional cancer detected.[23][26][28][29]
Sources: Data from Postma et al (2007),[29] Park et al (2006),[23] Shim et al (2007),[27] Muntener et al (2010),[22] Kobayashi et al (2006),[25] Rosario et al (2008),[28] Thompson et al (2005),[26] Canby-Hagino et al (2007).[24]
The sensitivity for detecting higher-grade cancers increased when the PSA threshold was lowered from 4.0 ng/mL, and these increases were greater than those for the detection of any cancer:[26] lowering the PSA threshold to 3.0 ng/mL increased the sensitivity for identifying any cancer by 11.7 percentage points, whereas the sensitivity for identifying cancers with Gleason score > 6 increased by 17.2 percentage points, and for identifying cancers with Gleason score > 7 increased by 17.5 percentage points. Similarly, lowering the PSA threshold to 2.5 ng/mL increased sensitivity for identifying any cancer by 20.0 percentage points, whereas the sensitivity for identifying cancers with a Gleason score > 6 increased by 26.8 percentage points, and for identifying cancers with a Gleason score > 7 increased by 28.0 percentage points. Further reduction to 2.0 ng/mL did not result in greater increases in sensitivity for detecting higher grade disease.[26] Considerable weight has been given to the Prostate Cancer Prevention Trial study.[26] However, there are two caveats to the application of these results to population-based prostate cancer testing in Australia. First, participants had PSA levels of 3.0 ng/mL or less, a normal DRE and an American Urological Association symptom score less than 20 prior to the start of annual testing and, thus, may not represent a general population of men in the relevant age group. Secondly, Hybritech PSA assays were used and, while it was not reported how these assays were calibrated, Hybritech calibration was probably used. As PSA measurements vary with assay type and calibration, the absolute values for PSA measurements reported in the Prostate Cancer Prevention Trial study[26] may not be directly applicable to the Australian context, in which over 95% of laboratories use the WHO calibration and the most commonly used assays are the Roche and Abbott assays. Back to top Using a PSA test result at a particular age to inform subsequent PSA testingTwo level III-2 studies[30][31] reported the risk of prostate cancer mortality according to PSA levels in men younger than 56 years. One was a retrospective cohort study of participants in the Copenhagen City Heart Study.[30] This study was at moderate risk of bias for PSA levels at ages 45–49 and 50–54 years and at high risk of bias for PSA levels at ages less than 45 years. The second study was the larger Malmö Preventive Project,[31] which was at high risk of bias. It used a retrospective cohort design to assess the risk associated with PSA levels at age 51–55 years, and a nested case-control design to assess the risk associated with PSA levels at 37.5–42.5 years and 45–49 years. For the latter design, absolute risk was imputed and the imputation was validated in the cohort group. This review focused on men from approximately age 40–55 years at testing and a maximum of 20 years follow-up, since its primary purpose was to obtain data relevant to PSA testing over a period of approximately 20 years from first testing. In the Copenhagen City Heart Study,[30] blood was sampled in 1981–1983 and PSA testing introduced into clinical practice in Denmark in 1995. Thus, informal PSA screening was unlikely to have affected 10-year risks of prostate cancer mortality. In the Malmö Preventive Project[31] blood was sampled from 1974–1984 for the case-control study and 1980–1990 for the cohort study. On the basis of Swedish PSA testing data,[31] the authors assumed that testing rates remained low (up to 5%) up until 1998 (8 years prior to end of study) and therefore that it was unlikely that any informal or opportunistic screening could have substantively affected prostate cancer mortality 15 and 20 years after PSA measurement. Given their retrospective designs, baseline PSA levels could not have affected prostate cancer diagnosis in either of these studies.[30][31] The studies[30][31] took place in Danish and Swedish populations (not primarily high-risk populations) that were followed up primarily in the pre-PSA era, when more effective definitive treatments may have been less readily available or offered than in Australia today. However, given that these are populations of European origin, as are a majority of Australians, and that the studies relate primarily to the natural history of a disease in relation to a risk indicator, they may reasonably be taken to represent the evolution of prostate cancer risk in Australia in relation to PSA levels measured on blood taken prior to the beginning PSA testing for the early detection of prostate cancer. Table 2.6 summarises estimates of increments in absolute percentage cumulative risk of prostate cancer death above the risk at a baseline PSA of < 1 ng/mL[30] or the lowest quarter of the PSA distribution[31] by age, length of follow-up and baseline PSA level. While the Copenhagen City Heart Study[30] reported on cumulative risk for three additional PSA levels (from > 3.0 to 4.0 ng/mL, from > 4.0 to 10.0 ng/mL, and > 10.0 ng/mL), increments in risk at these levels are not shown because the lower bound of the top 10% of the PSA distribution in the Malmö Preventive Project[31] lay consistently in the range 1.0–3.0 ng/mL. The results in the table show the following:
Sources: Orsted et al (2012)[30], Vickers et al (2013)[31] Back to top PSA testing strategies in high-risk groupsThere is little or no empirical evidence to support any particular modification of a PSA testing protocol to apply to men at high risk of prostate cancer. The approach taken in most guidelines for PSA testing is to recommend that men at high risk for prostate cancer begin testing at an earlier age than men at average risk (typically at age 45 years), whereas men at average risk are advised to begin testing at age 50 years. This is a rational approach because men at high risk have, depending on their risk factors, an increased risk at each age that is likely to be a constant multiple (RR for the risk factor in questionii) of the risk in men at average risk. Therefore, it should be possible to identify an age earlier than 50 years at which risk in men with a particular risk factor would be the same as the average risk at age 50 years, and from which risk would be expected to evolve with age in the same way as it would evolve from age 50 years in men at average risk. In principle, by beginning PSA testing at this age, high-risk men could expect the same benefit, and probably the same harm, from testing as average-risk men starting testing at age 50 years. Using present incidence or mortality rates for prostate cancer, it is arguably not possible to identify accurately the age at which men at, for instance, twice the average risk of prostate cancer would have the same underlying risk of prostate cancer occurrence or death as average-risk men at age 50. This is for two reasons:
Each of these factors will have an effect on the relationship of age with prostate cancer incidence and mortality because of the strongly age-determined frequency of PSA testing. Therefore, in seeking to determine an age at which high-risk men might be advised to begin PSA testing that is equivalent to a recommended age of 50 years for average-risk men, we chose to focus on the annual average prostate cancer mortality rates for Australia in 1991 to 1995, the 5-year period of peak prostate cancer mortality. This peak occurred shortly after PSA testing began in Australia and, thus, rates for 1991–1995 are unlikely to have been influenced by PSA testing. Mortality is considered to be more relevant than incidence in this context, because it is the hazard that PSA testing aims to prevent. Table 2.7 provides estimates of the increase in prostate cancer mortality in average risk men over the succeeding 10 years of their lives from ages 40, 45 and 50 years (based on 1991–1995 Australian mortality rates, which are approximately those that obtained before PSA testing in Australia could have had an effect on mortality).[32] For ages 40 and 45 only, Table 2.7 also includes estimates for men with varying levels of higher than average risk of prostate cancer (RR 2.0–5.0). A period of 10 years of life was chosen because most recent included results of the ERSPC indicate that most of the mortality reduction achieved through PSA testing is evident at 10–11 years after start of testing.[8] Table 2.7 indicates that a 45-year-old man at three times the average risk of prostate cancer would have an increase in his annual risk of prostate cancer death of 23.9 per 100,000 over the next 10 years of his life from the very low rate at age 45 years. This increase is a little higher than the corresponding increase for an average-risk man starting PSA testing at age 50 years (22.7 per 100,000), and would therefore provide as much justification, in terms of risk of death from prostate cancer, for offering PSA testing to a 45-year-old man at three-times the average risk of prostate cancer as there is for offering it to a 50-year-old man at average risk of prostate cancer. For a man at 2.5 times average risk, the increase in annual risk of prostate cancer death over the next 10 years is 20.0 per 100,000, which is somewhat less than that for the 50-year-old at average risk, but probably sufficient to justify offering PSA testing to a 45-year-old at 2.5 times the average risk of prostate cancer. Following the same logic, in 40-year-old men, a case can be made for offering testing to those whose risk is 9–10 times average risk (corresponding to increases in annual risk of prostate cancer death over the next 10 years of life of 21.1 and 23.4 per 100,000 respectively) or more. Table 2.7. Estimated increase in prostate cancer-specific mortality rate (annual number of deaths per 100,000 men) over the next 10 years for Australian men aged 40, 45 and 50 years who are at average risk of prostate cancer, and those who are at two- to ten-fold increased risk of prostate cancer
*This value is provided as a point of reference with which to compare the increases in prostate cancer mortality over the next 10 years in men aged 40 and 45 years at various degrees of increased risk of prostate cancer. Source: Data from Australian Institute of Health and Welfare (2014) [32]
Back to top Evidence summary and recommendations
For recommendations on further investigations, see 2.5 Testing with variants of PSA to improve sensitivity after an initial total PSA ≤ 3.0 ng/mL and 2.6 Testing with variants of PSA or repeat PSA testing to improve specificity after an initial total PSA > 3.0 ng/mL. Back to top Expected benefits and harms from recommended PSA testingInforming men of the benefits and harms of testing is a key component of the recommendations regarding PSA testing. To aid their use in practice, therefore, we have compiled Table 2.8, a quantitative table of estimated harms, benefits and measures of the balance and harms and benefits associated with two of the testing protocols, testing from age 50 or age 45 in average risk men. This table can be used when informing men of the benefits and harms of testing and the trade-offs that a decision in favour of testing would entail. It is based on results of the best available mathematical modelling studies, which we have used elsewhere in this guideline. Ideally, the results would have been produced especially for this guideline and based on an Australian model. This is not yet possible but will be soon. It was not considered to be possible to add the protocol for testing men at higher than average risk to Table 2.8 since this issue has not yet been dealt with in published reports of the adequate quality models. Table 2.8. Modelled estimates of harms, benefits and balance of harms to benefits of recommended PSA testing protocols
*Probability of harms is estimated over the duration of the testing protocol; benefits are estimated over the lifetime from the age testing started. †Estimates of harms, benefits and balance based on modelling results for this protocol were from Pataky et al (2014)[19] ‡Estimates of harms, benefits and balance based on averages of the above results for 50-69 years obtained by Pataky et al (2014)[19] and results for a protocol for testing men 40-69 years of age every 2 years with a criterion for further investigation of > 2.5 ng/mL obtained by Gulati et al (2013)[16]. Most likely effect of the lower criterion PSA value is to over-estimate the probability of > 1 false positive PSA test. Back to top Health system implications of these recommendationsClinical practiceDespite a recommendation by the Royal College of Pathologists of Australasia to repeat PSA testing at intervals of 2 years or 4 years, depending on the result,[34] it is probable that many men currently having PSA testing are tested annually. Therefore, the recommendation to offer PSA testing every 2 years in men aged 50–69 years who wish to undergo testing after being informed of the benefits and harms of testing could lead to less frequent testing and fewer false positive tests. Misuse or new safety concerns from these recommendations are not envisaged. An increase in litigation alleging malpractice is possible given the benchmark these recommendations provide and the known frequency of practice that does not align with them, particularly with respect to assurance that men tested have been informed of the benefits and harms of testing. This potential legal risk will be mitigated by robust efforts to ensure that knowledge of the guideline is disseminated to all relevant health practitioners and the development of aids that will assist them in practising according to the guideline. ResourcingImplementation of the recommendation for a 2-year interval between PSA tests for men aged 50–69 years who wish to undergo testing could reduce the costs of testing, reduce the frequency of false positive tests and reduce consequent investigation and its cost. Barriers to implementationNo barriers to implementation of these recommendations are foreseen. Back to top i Clinical questions were translated into the PICO framework: population, intervention (or exposure), comparator and outcome (see Appendix 3). ii In this section, RR refers to a presumed unbiased estimate of the RR for prostate cancer. As noted in Chapter 1, studies of risk factors that are strongly believed or well known to put men at high risk for prostate cancer, such as a family history of prostate cancer, are likely to produce positively biased estimates of RR of prostate cancer incidence because of a higher likelihood that men thought to be at high risk will request or be offered PSA tests, often starting at a younger age, and have a risk of incident prostate cancer that is boosted by over-diagnosis. Correspondingly, estimates of RR of prostate cancer mortality are likely to be negatively biased due to earlier diagnosis of otherwise potentially fatal prostate cancer, although probably less so. While these matters do not influence the logic of this section, they need to be taken into consideration when deciding whether or not a particular risk factor should lead to a change in the PSA testing protocol, as proposed in the recommendations arising from this chapter. The recommendation for PSA testing strategies in men at higher-than-average risk of prostate cancer (below) is based on evidence on the RR of prostate cancer mortality associated with family history of prostate cancer, not the RR of prostate cancer incidence associated with it (Chapter 1 Risk), given that the former is likely to be the less biased estimate of relative risk. iii This Consensus-based recommendation assumes testing with the criterion for further investigation a PSA of ≥ 3 ng/mL. This recommendation will be a high priority for reconsideration when the Australian model of PSA testing has been completed. For example, use of the 95th percentile for age in place of ≥ 3 ng/mL might improve appreciably the balance of harms to benefits of testing in men 70–74 years of age. References
Back to top DiscussionSupporting attachments
Back to top |