Table 2.
Summary of findings from pooled analysis of all eligible trials (n=5)
| Outcome and timeframe | Study results and measurements | Absolute effect estimates | Certainty in effect estimates (quality of evidence) | Plain text summary | ||
|---|---|---|---|---|---|---|
| No screening | PSA screening | Difference (95% CI) | ||||
| Mortality and cancer incidence at 10 years* | ||||||
| All-cause mortality | IR 0.99 (95% CI 0.98 to 1.01) based on data from 675 232 patients in 4 studies. Follow-up 10-20 years |
129/1000 men | 128/1000 men | 1 fewer (3 fewer to 1 more) | Moderate (serious risk of bias†) | PSA screening probably has little or no effect on all-cause mortality |
| Prostate cancer mortality | IR 0.96 (0.85 to 1.08) based on data from 721 718 patients in 5 studies. Follow-up 10-20 years |
3/1000 men | 3/1000 men | 0 (0) | Low (serious risk of bias and inconsistency‡) | PSA screening may have little or no effect on prostate cancer mortality |
| Incidence of prostate cancer (any stage) | IR 1.23 (1.03 to 1.48) based on data from 675 232 patients in 4 studies. Follow-up 10-20 years |
32/1000 men | 39/1000 men | 7 more (1 to 15 more) | Low (serious risk of bias and serious inconsistency (resulting in large imprecision)§) | PSA screening may increase detection of prostate cancer (any stage) |
| Incidence of localised prostate cancer (stages I and II) | RR 1.39 (1.09 to 1.79) based on data from 647 751 patients in 3 studies. Follow-up 10-20 years |
19/1000 men | 26/1000 men | 7 more (2 to 15 more) | Low (serious risk of bias and serious inconsistency (resulting in large imprecision)¶) | PSA screening may increase detection of localised cancer |
| Incidence of advanced prostate cancer (stages III and IV) | RR 0.85 (0.72 to 0.99) based on data from 647 751 patients in 3 studies. Follow-up 10-20 years |
13/1000 men | 11/1000 men | 2 fewer (4 to 0 fewer) | Low (serious risk of bias and inconsistency§) | PSA screening may slightly decrease detection of advanced cancer |
| Quality of life | ||||||
| Quality of life | Measured by SF-6D (scale 0-1, high better) based on data from 1088 patients in 1 study | 0.76 mean | 0.75 mean | 0.01 lower (0.01 lower to 0.02 higher) | Low (risk of bias and indirectness). Data from surveys on a subsample of patients invited from sample of the Finnish ERSPC trial** |
PSA screening may have little or no difference on quality of life |
| Complications and adverse effects | ||||||
| Complication rates per biopsy | Based on data from 1147 patients in 1 study. Follow up 35 days |
Most common complications were blood in semen (93%), blood in urine (66%), pain (44%), shivers (19%), and fever (18%). 1.4% (95% CI 0.8% to 2.4%) of men admitted to hospital (mostly for sepsis) |
High (high quality and representative cohort study; intervention arm of CAP/ProtecT trial) | Complication rates per biopsy, (regardless of whether diagnosed through screening) | ||
| Biopsy related complications | Follow up 35 days | Among 1000 men, between those screened v not screened, prostate biopsies resulted in about 94 more men with blood in semen, 67 more with blood in urine, 45 more with pain, 19 more with fever, and 1 more hospitalised for sepsis | Low (because of additional uncertainty due to estimating likelihood along the diagnostic pathway) | Absolute differences in biopsy related complications between those screened v not screened | ||
| Complication rates by prostate cancer treatment modality | Based on data from 1643 patients in 1 study. Follow-up 6 years |
At 6 years, rates of any pad use (urinary incontinence) for active monitoring, surgery, and radiation groups were 8%, 17%, and 4%. Rates of erections not firm enough for intercourse were 70%, 83%, and 73% respectively. | High (high quality prospective cohort studies of a 3-armed RCT) | Complication rates per treatment modality for prostate cancer (regardless whether diagnosed through screening) | ||
| Complications of subsequent prostate cancer treatment | Follow-up N/A | Among 1000 men, between those screened v not screened, subsequent treatment for prostate cancer resulted in about 3 more with urinary incontinence, and 25 more with an erection not firm enough for intercourse | Low (because of additional uncertainty due to estimating likelihood along the diagnostic pathway) | Absolute differences in treatment related complications between those screened v not screened | ||
| False positive screening results within 1 year | Based on data from 61 000 patients in 1 study | False positive rates were 66.5%, 66.0%, and 63.0% in first, second, and third round of screening respectively | High (high quality prospective cohort from 5 arms of the ERSPC trial) | Among men with PSA level ≥4 ng/mL at screening, about 67% will have a negative subsequent biopsy | ||
| False negative screening results | Based on data from 2950 patients in 1 study. Follow-up 7 years |
Among men with PSA ≤4 (age 62-91 years), 15.2% were diagnosed with prostate cancer during follow-up, 2.3% developed a cancer Gleason score ≥7 | Low (prospective observational cohort with possible verification bias (only 2950 of 3568 men with PSA ≤4 ng/mL had end-of-study biopsy)) | Among men with PSA ≤4 ng/mL at screening, about 15% could be false negative and will subsequently be diagnosed with prostate cancer, about 2% with high grade cancer | ||
| Anxiety about having cancer | Based on data from 2 large observational studies. Follow-up ≤1 year |
No evidence comparing PSA screening with non-screening. A large cohort study in Sweden (n=4.3 million) showed an increased risk of suicide (RR 2.6 (95% CI 2.1 to 3.0)) and cardiovascular events (RR 1.3 (1.3 to 1.3)) during first year after diagnosis. Another cohort study in the US (n=343 000) showed no increased risk of suicide during first year since the widespread use of PSA screening (after 1993) but an increased risk of cardiovascular death during first month after diagnosis (aRR=1.55 (1.3 to 1.8)) | Very low (risk of residual confounding in observational data) | It is uncertain whether screening results in changes in anxiety about having cancer, but a diagnosis of prostate cancer might increase immediate risks of suicide and cardiovascular death | ||
IR=incidence ratio. RR=relative risk. PSA=prostate-specific antigen.
The trials varied in their duration of follow-up from 10 to 18 years, and relative estimates of effect were pooled at the longest available follow-up time. However, we estimated the absolute effect on a 10-year time-horizon. We used as baseline risk in the non-screening arm of the CAP trial, because it provided the most contemporary estimates of risks from large sample of men representative of a general practice setting.
Risk of bias, serious: inadequate concealment of allocation during randomisation resulting in potential for selection bias; inadequate or lack of blinding of participants and personnel, resulting in potential for performance bias; some contamination. Imprecision, not serious: lower border of confidence interval compatible with a relevant yet small clinical risk reduction.
Risk of bias, serious: inadequate concealment of allocation during randomisation resulting in potential for selection bias; inadequate or lack of blinding of participants and personnel, resulting in potential for performance bias; some contamination. Inconsistency, serious: I2 53%; ERSPC trial shows significant reduction while all other trials show no significant difference.
Risk of bias, serious: inadequate concealment of allocation during randomisation resulting in potential for selection bias; inadequate or lack of blinding of participants and personnel, resulting in potential for performance bias; some contamination. Inconsistency, serious: confidence intervals of some of the studies do not overlap with those of most included studies or the point estimate of some of the included studies. Imprecision, not serious: we did not downgrade for imprecision because it resulted from inconsistency.
Risk of bias, serious: inadequate concealment of allocation during randomisation resulting in potential for selection bias; inadequate or lack of blinding of participants and personnel, resulting in potential for performance bias; inadequate or lack of blinding of outcome assessors, resulting in potential for detection bias; some contamination. Inconsistency, serious: confidence intervals of some of the studies do not overlap with those of most included studies or the point estimates of some of the included studies.
Risk of bias, serious: based on cross-sectional analysis of random sample, self-reported measure of health related quality of life. Indirectness, serious: patients without prostate cancer were excluded from the screening arm (health related quality of life of patients screened with no diagnosis of prostate cancer may differ).