Our results show that FP results affect every eighth man in repeated screening for PC with PSA even with a relatively high cutoff level of 4.0
. More than a quarter of the men with FP results are subsequently diagnosed with PC, although most of these cancers are localised and of low grade and have similar characteristics as cancers in men with a previous negative screening test. More than half of these men have persistent high serum PSA levels resulting in repeated FP results and biopsies. They are also at high risk of dropping out of subsequent screening.
The Finnish Prostate Cancer Screening Trial is part of the ERSPC study. There are some differences between the ERSPC centres in, for example, the mode of recruitment, screening interval, invitation procedures and the PSA threshold leading to biopsy. The Finnish trial is population-based and the largest of the ERSPC centres. A population-based study design ensures good generalisability at the population level.
The ERSPC study recently showed preliminary mortality results indicating a 20% relative decrease in mortality in the screening arm (Schröder et al, 2009
). This was the first evidence for benefits from screening for PC with PSA. However, as shown by the ERSPC trial, 1410 men would have to be offered screening and 48 PCs treated to prevent one PC death during a 9-year period. In addition, the negative consequences of screening (adverse effects, including overdiagnosis, overtreatment and costs) still need to be carefully evaluated to allow assessment of the balance between benefits and harms before evidence-based decision-making concerning provision of screening can be made. This analysis contributes to that requirement.
Our study presents a similar proportion of FP results per screening episode as a previous Prostate, Lung, Colorectal and Ovarian (PLCO) Cancer Screening Trial study (Lafata et al, 2004
), but also provides longer follow-up and information on the relation between FP results and several clinically important characteristics, such as PC, BPH medication, age and PSA level. Cumulative rates of FP results in repeated screening for several screening modalities were recently reported from the PLCO trial (Croswell et al, 2009
). The authors showed a risk of 10.4% for at least one FP result in annual PSA screening during the 3-year screening period. We found a 12.5% risk for at least one FP result in two to three successive screening rounds during a 12-year follow-up, but the probability varied strongly with age. In the youngest age cohort (screened initially at age 55), the risk was 9.0% and in the oldest men (first screen at age 67 with only two rounds of screening) it was 15.7%. As previously noted, the incidence of PSA-elevating diseases other than PC (prostatitis, BPH) increases with age (Koskimäki et al, 1998
; Rhodes et al, 1999
; Wright et al, 2002
) and explains the higher FP proportion in older men. This explanation is also consistent with the finding that men who used BPH medication had an increased risk for an FP result, despite the PSA-lowering effect of finasteride (Etzioni et al, 2005
). These men were also older than men without BPH medication.
In our study, a PSA threshold of 4.0
was used. In addition, during 1996–1998, men with a suspicious DRE finding and in 1999–2007 men with PSA 3.0–3.9 and free/total PSA ratio
16% were referred. The PSA threshold was chosen in 1996 when the study began. A study from the Prostate Cancer Prevention Trial reported that 24.7% of men in the placebo group with PSA 2.1–4.0 had PC when biopsied at the end of the study, although 50% of these men were older than 71 years and all the cancers were stage T1 (Thompson et al, 2004
). In another study with younger subjects (50–65 years), 11.3% (13 of 115) of men with PSA level 1.1–3.99 and F/T PSA
20% had cancer in biopsy (Rowe et al, 2005
). In our study with men aged 55–71 years, the proportion of PC from those biopsied varied from 26.6 to 32.9%. If the PSA threshold had been lower, for example, 2.1–3.9
, the proportion of PC in the biopsied men would probably be smaller, that is, the downside of the expected improvement in sensitivity would be a decreased specificity.
A screen-detected cancer was defined as a PC detected within 1 year from the PSA test in a man with a screen-positive result. On this basis, we defined an FP as a screen-positive result with no PC diagnosis within 1 year from the PSA test (excluding men without biopsy). Prostate-specific antigen predicts the development of PC by several years and there is no clear time as for the optimal definition of an FP result, but the proportion of de novo cases relative to those present at the screen can be anticipated to increase with time since the PSA test. If we had extended our 1-year limit to, for example, 3 years, the number of FP results would have decreased by 86 (6.5%), 88 (5.9%) and 47 (4.7%) men in the first, second and third rounds, respectively. These men were diagnosed with an interval cancer within 1–3 years from the PSA test. As the proportion of these men out of all FP men was relatively small (4.7–6.5%), using another definition would not be likely to materially affect our results.
In cancer screening, FP results are problematic for several reasons. Biopsies bring discomfort and often pain to the patient during the procedure (Mäkinen et al, 2002
). Waiting for the result is a psychological strain, which can have negative effects for at least a year even after a negative biopsy result (Fowler et al, 2006
). The economic impact of FP results has not been thoroughly analysed, but these men seem to receive more follow-up interventions such as PSA testing and re-biopsies, which add to the costs of screening (Lafata et al, 2004
). Biopsy – similar to any invasive procedure – involves risks for adverse health effects, such as bleeding, infection or abscess formation (Mäkinen et al, 2002
), although these complications are not very common.
There is previous evidence that FP men undergo more follow-up testing and biopsies than men with normal PSA (Fowler et al, 2006
). Our results show that men with FP results receive more biopsies than do men who are diagnosed with PC. On average, every third FP man undergoes two biopsies within 4 years from the screen. It has been previously reported that the risk of clinically significant cancer decreases after the second biopsy (Djavan et al, 2003
). Our study is likely to underestimate the average number of biopsies as we have no data on private sector visits and procedures and it is likely that some of the benign biopsies in the public sector are not reported to our database.
However, our findings indicate an increased risk for future PC with a history of one or several FP results. As many as 16% of FP men were diagnosed with PC in the next round. Most of the PCs were not aggressive, but, for example, in the third round as many as 29.9% of cancers were Gleason score 7 and 13% were advanced (T3-4NxM0 or TxNxM1 or Gleason score 8). Of the first round FP men, almost a third were diagnosed with PC during the 8–11 follow-up years. The proportion of PC diagnoses among the men with FP results at the second and third rounds were substantially lower (15.9 and 2.6%) – most likely because of a shorter follow-up. As previously mentioned, over 10% of men over 50 years of age can be diagnosed with PC even with low PSA levels (Rowe et al, 2005
). Therefore, if men with an FP result receive more biopsies in the follow-up period than men with a negative screen, they could be more likely to receive a PC diagnosis because of more frequent biopsying. In addition, in 2002 we started using 10–12 core biopsies instead of sextant biopsies, which could increase the chances of finding small, indolent lesions during the later follow-up period. Both these factors increase the PC risk in FP men.
When the men were stratified by serum PSA level, it was evident that at the first (prevalence) screen, high PSA level was clearly associated with PC and moderately increased serum PSA level with FP. At the second and third (incidence) screens, these differences evened out and the positive predictive value of high PSA for PC decreased. The most likely explanation for these trends is that at the first screen most of the high PSA cancers were ‘harvested' from the study population. Some of them were still detected at the second screen, but generally the cancers that produce high PSA were caught at the prevalence screen and few such cases arose de novo between the screening rounds.
In the PLCO trial, men with an FP result were almost twice more likely to decline subsequent screening compared with men with a negative screening result (Ford et al, 2005
). Our results are similar, with RRs varying from 1.5 to 2.0. There might be several reasons behind this. The FP men could decide not to participate because of the unpleasant experience of unnecessary biopsy procedures and the anxiety related to the fear of PC diagnosis. On the other hand, an FP man could sense relief after a benign biopsy and deem it unnecessary to participate in the next screening round. Also, receiving a positive screening result without a confirmed PC diagnosis may erode a man's perception of the effectiveness of screening.
These findings emphasise the paradoxical problem of FP results in PC screening. On the one hand, FP men frequently have persistently high PSA levels (>50% chance of having another FP result in the next round) and undergo several biopsies. On the other hand, they are more likely to be diagnosed with PC, either because of biological processes or more active diagnostic procedures. New approaches are urgently needed for improved risk stratification among these men, that is, to predict which of them may harbour a clinically significant PC, which may have an insignificant indolent PC and which may have other factors underlying the elevated PSA level.
There is one weakness in our study. In some cases, the follow-up time after the third screen was relatively short (
3 years), as the last men were screened in the end of 2007 and follow-up ended in 2007. Therefore, some post-screening cancers were lacking for the last screening cohort. However, we believe that the strengths of this study well outweigh this weakness.
In conclusion, we present data from a prospective randomised controlled PC screening study spanning 12 years and three screening rounds. We have analysed the FP screening results during these rounds and calculated that every eighth man screened is subject to an FP result at least once in repeat screening. The men who receive FP results are likely to have a subsequent FP result(s) later if screened again. Also, these men commonly drop out of subsequent screening rounds. This poses a difficult equation, as men with FP results are at increased risk of being diagnosed later with a PC. More research is needed to balance the sensitivity and specificity of PC screening to minimise the proportion of FP results.