br Blood samples were analyzed in Dr Lilja s laboratory
Blood samples were analyzed in Dr. Lilja's laboratory at Lund University, Malmö, Sweden. Plasma aliquots from cases and controls were grouped in blinded case-control sets and handled together throughout processing and analysis along with blinded quality control samples. Total PSA measurements were performed on the AutoDelfia 1235 automatic immunoassay system using the dual-label DELFIA Prostatus total/free PSA-Assay (Perkin-Elmer, Turku, Finland) as described previously
. Intra- and inter-batch coefficients of variation were 4.5% and 6.5%, respectively. Concentrations of total PSA in λ-Carrageenan stored at–80 C for 20 yr are comparable to concentrations in samples measured soon after blood draw .
2.4. Statistical analysis
Exact conditional logistic regression was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between categories of baseline PSA and risk of total and aggressive PCa, overall and within two age groups: 40–54 and 55–64 yr. Baseline PSA values were categorized into quantiles within narrower age groups (40–49, 50–54, 55–59, and 60–64 yr) according to the distribution of levels among the controls, as age is a major determinant of PSA levels. For the analysis of aggressive PCa among men aged 40–54 yr, there were no cases in the reference group ( median); therefore, the median unbiased estimate was used .
In secondary analyses, we excluded men with PSA >4 ng/ml (81 cases, 24 controls) to test the predictive ability of baseline PSA among men with PSA in a normal range for whom baseline PSA could be used to determine an on-going risk-stratified screening strategy. We also conducted analyses excluding cases diagnosed within the first 2 yr (excluding 46 total, 41 aggressive) and 5 yr (excluding an additional 64 total, 47 aggressive) of follow-up to better assess the ability of baseline PSA to predict disease over a longer period. Finally, we conducted a sensitivity analysis among cases that were fatal or AJCC stage III/IV at diagnosis (n = 33), as these advanced cases are most likely to be clinically apparent and less likely to be diagnosed based only on PSA screening. For these sensitivity analyses, as matching was broken due to exclusions, we controlled for matching factors.
Self-reported information was recorded at enrollment on race/ ethnicity, height, weight, marital status, education, diabetes, smoking, family history of PCa, and ever having a PSA test, or digital rectal exam prior to study entry. We present only models adjusted for matching factors, as results were similar when adjusted for covariates. Data on PSA testing after study entry was available from a subset of men who completed one or two follow-up questionnaires an average of 4.5 and 7.5 yr after baseline.
Two-sided p values <0.05 indicated statistical significance. Analyses were performed using SAS version 9.4 (SAS Institute, Inc.; Cary, NC) and R version 3[2T$DIF].2.2 (R Foundation for Statistical Computing, Vienna, Austria, 2015).
Descriptive statistics of the study population are shown in Table 1. Median age at blood draw was 56 yr. Median time from blood draw to diagnosis was 4.4 yr among all cases and 5.0 yr among aggressive cases, with a maximum time of 10.25 yr (Table 1). Median PSA levels among controls ranged from 0.72 ng/ml for age 40–49 yr to 1.03 ng/ml for age 60– 64 years; full distributions by age group, which were used to create the categories discussed below, are shown in Table 3.
Compared to men with PSA at or below the age-specific median, men with levels above the median had significantly increased risk of aggressive PCa across age groups.
Baseline PSA was still highly associated with risk of total and aggressive disease when excluding men with PSA >4 ng/ml (Table 2). The elevated risks persisted after excluding cases diagnosed within 2 or 5 yr of blood draw, though results were attenuated. Finally, the elevated risks persisted when aggressive cases were limited to those that were fatal or AJCC stage III/IV at diagnosis, with an OR for >90th percentile versus median of 39.5 (95% CI, 7.37– infinity) for ages 40–54 yr and 28.3 (95% CI, 5.73–infinity) for age 55–64 yr.
Table 3 shows the proportion of cases captured in different PSA categories by age group. All 36 aggressive PCa cases in men aged 40–54 yr occurred among those with baseline PSA above the age-specific median. Among men aged 55–64, 52/55 aggressive cases (95%) were in men with baseline PSA above the age-specific median.
In this prospective study among US black men, we found that a single baseline PSA level measured during midlife strongly predicted subsequent diagnosis of total and aggressive PCa up to 12 yr after blood draw. Risk was significantly higher for men with PSA levels above the age-specific median than for those with PSA levels below the age-specific median. Men above the 90th percentile had the greatest risk relative to those below the median. As expected, PSA was somewhat less predictive of PCa among the older age group of 55–64 yr, likely because BPH becomes more common in older men and influences PSA levels apart from PCa .