Association Between Lead Time and Prostate Cancer Grade: Evidence of Grade Progression from Long-term Follow-up of Large Population-based Cohorts Not Subject to Prostate-specific Antigen Screening

Abstract

Background

Lead time (LT) is of key importance in early detection of cancer, but cannot be directly measured. We have previously provided LT estimates for prostate cancer (PCa) using archived blood samples from cohorts followed for many years without screening.

Objective

To determine the association between LT and PCa grade at diagnosis to provide an insight into whether grade progresses or is stable over time.

Design, setting, and participants

The setting was three long-term epidemiologic studies in Sweden including men not subject to prostate-specific antigen (PSA) screening. The cohort included 1041 men with PSA of 3–10 ng/ml at blood draw and subsequently diagnosed with PCa with grade data available.

Outcome measurements and statistical analysis

Multivariable logistic regression was used to predict high-grade (Gleason grade group ≥2 or World Health Organization grade 3) versus low-grade PCa at diagnosis in terms of LT, defined as the time between the date of elevated PSA and the date of PCa diagnosis with adjustment for cohort and age.

Results and limitations

The probability that PCa would be high grade at diagnosis increased with LT. Among all men combined, the risk of high-grade disease increased (odds ratio 1.13, 95% confidence interval [CI] 1.10–1.16; p < 0.0001) for every 1-yr increase in LT, with no evidence of differences in effect by age group or cohort. Higher PSA predicted shorter LT by 0.46 yr (95% CI 0.28–0.64; p < 0.0001) per 1 ng/ml increase in PSA. However, there was no interaction between PSA and grade, suggesting that the longer LT for high-grade tumors is not simply related to age. Limitations include the assumption that men with elevated PSA and subsequently diagnosed with PCa would have had biopsy-detectable PCa at the time of PSA elevation.

Conclusions

Our data support grade progression, whereby following a prostate over time would reveal transitions from benign to low-grade and then high-grade PCa.

Patient summary

Men with a longer lead time between elevated prostate-specific antigen and subsequent prostate cancer diagnosis were more likely to have high-grade cancers at diagnosis.

1. Introduction

Lead time is defined as the time between screen detection and clinical detection of a cancer, and remains a key concept for early detection of cancer. If lead time is short, it is difficult to ensure that men are screened in the short window between when a cancer is first amenable to screen detection and when it is detected clinically. Conversely, long lead times tend to lead to overdiagnosis. Despite its critical importance, lead time is not directly observable for an individual man, as this would require unethical withholding of a cancer diagnosis from a patient.

Various approaches for estimating lead time have been reported. Microsimulation models, which are calibrated against population trends and the results of large trials, can provide model-based estimates of lead time [1]. A second approach is to compare the time to a given incidence of prostate cancer between control and intervention arms of a randomized trial [2,3]. Our group has pioneered a novel approach based on measurement of prostate-specific antigen (PSA) in cryopreserved anticoagulated blood plasma samples from longitudinal epidemiological studies of large, population-based cohorts not subject to PSA screening [4]. We hypothesize that a man with elevated PSA at baseline and subsequently clinically diagnosed with prostate cancer would have had screen-detectable cancer at study entry. The time between blood draw and diagnosis is thus an estimate of lead time. Using this approach, we reported that the mean lead time is probably longer than had previously been assumed and that the distribution of lead times is normal, rather than exponential, as had often been assumed [5].

One key question in prostate cancer is whether the development of high-grade cancer is an early or late event in the disease process. One hypothesis is that prostate cancer is either low-grade or high-grade at initiation and that grade remains relatively stable over time. The alternative hypothesis is that grade progresses from low to high over time as mutations accumulate in a cancerous prostate. In this study, we explored the relationship between lead time and grade to evaluate these two competing hypotheses. If cancer grade at diagnosis is inversely associated with lead time, this would suggest that grade progression is an early event on the grounds that high-grade cancer is more aggressive and is likely to be diagnosed sooner than low-grade disease. Alternatively, if a longer lead time is associated with a higher risk of high-grade disease, this would suggest that grade changes over time. Our aim was to investigate the association between lead time (from detection of elevated PSA in blood and subsequent clinical diagnosis) and prostate cancer grade.

2. Patients and methods

We used data from the Malmö Preventive Project (MPP) [6], the Malmö Diet and Cancer study (MDC) [7], and the Västerbotten Intervention Project (VIP) [8] in Sweden. In brief, the cohorts are from long-term, epidemiological population-based projects including a representative sample of healthy men aged 33–74 yr who provided cryopreserved anticoagulated blood plasma samples at baseline. There was very low rate of opportunistic PSA testing for prostate cancer during study entry and many years of study follow-up, so the large majority of the diagnoses were the result of clinical work-up. Outcomes were ascertained from a cancer registry that is known to be highly accurate [9]. Blood samples were retrieved for cases and matched controls, and PSA was measured using techniques that provide estimates that have been shown to be equivalent to contemporaneous PSA measurement [10].

Figures 1 and 2 show flow diagrams for study inclusions. Of the 26 656 participants in the MDC and MPP cohorts, 3005 men were diagnosed with prostate cancer. After excluding autopsy cancer and men with missing PSA or grade, 1946 Malmö participants were eligible for analysis, of whom 382 (20%) had high-grade cancer according to the World Health Organization (WHO) classification. Of 40 379 participants in the VIP study, 1218 men were diagnosed with cancer, with 1159 eligible for analysis based on a known biopsy grade and PSA measured in blood obtained at baseline.

Fig. 1.

Flow diagram of the inclusion process for the Malmö Preventive Project (MPP) and the Malmö Diet and Cancer (MDC) cohorts. PSA, prostate-specific antigen.

Fig. 2.

Flow diagram of the inclusion process for the Västerbotten Intervention Project (VIP) cohort. PSA, prostate-specific antigen.

Our aim was to describe empirically the distribution of the estimated time by which a PSA screen would theoretically advance the date of diagnosis—that is, the lead time—by age, PSA, and grade at subsequent prostate cancer diagnosis. High grade was defined as Gleason grade group (ng/ml) 2 or higher or WHO grade 3. We assumed that those with elevated PSA who were later clinically diagnosed with prostate cancer already had screen-detectable prostate cancer at the time of blood sampling. For our main analysis, we defined elevated PSA as ≥3.0 ng/ml. Given that few men undergoing regular screening present with PSA >10 ng/ml, we only included men with PSA of 3.0–10 ng/ml.

We plotted the risk of high-grade cancer at diagnosis against the time from baseline collection of blood to diagnosis for various age and PSA level combinations. Some participants provided blood samples at multiple ages. In this case, the earliest PSA measurement available for each age group was used. We tested the association between lead time and the risk of a high-grade cancer diagnosis among those diagnosed with cancer using univariable logistic regression for each age and study cohort separately

As a sensitivity analysis, we changed the cutoff point for elevated PSA from 3.0 ng/ml to PSA values ≥2–4 ng/ml. We also repeated our analyses changing the definition of high-grade cancer to ≥3 ng/ml. Lastly, to account for the fact that high-grade cancers have a greater propensity for metastasis, that metastasis almost inevitably leads to clinical detection, and that metastasis is a late event—something that may lead to an apparently longer lead time for high-grade cancers—we repeated all the analyses, excluding patients with metastasis at diagnosis. All analyses were performed using Stata version 13.0 (StataCorp, College Station, TX, USA).

3. Results

Table 1 lists details for the study cohorts. Of the 1945 Malmö participants subsequently diagnosed with prostate cancer, 838 (43%) had high-grade cancer at diagnosis. For the VIP cohort, there were 1159 diagnoses and 500 (43%) high-grade cancers. Further details of the study cohort and lead times are given in the Supplementary material.

Table 1.

Description of the study cohorts

Study	Age at any screening (yr)	Men (n)		Median PSA at first PSA > CO, ng/ml (IQR)	Total cancers diagnosed b		LGC at diagnosis b		HGC at diagnosis b
		Total	PSA > CO		Number	Median LT, yr (IQR)	Number	Median LT, yr (IQR)	Number	Median LT, yr (IQR)
Malmö	35–45	8187	113	4.4 (3.7–5.7)	16	18.7 (14.0–21.7)	10	19.1 (14.1–21.6)	5	18.3 (17.8–21.9)
Malmö	45–55	14 274	826	4.5 (3.5–6.5)	156	13.4 (9.2–19.8)	88	12.1 (8.4–16.7)	55	16.2 (11.3–23.0)
Malmö	55–65	8376	793	4.6 (3.6–7.0)	313	8.0 (5.0–11.8)	177	6.6 (4.2–10.3)	121	9.7 (6.1–15.0)
VIP	35–45	17 086	7	3.8 (3.3–7.3)	6	9.9 (6.7–13.2)	2	Range 13.2–15.7	4	7.5 (6.3–9.9)
VIP	45–55	17 837	151	4.3 (3.5–5.5)	121	7.6 (5.2–11.4)	81	7.0 (4.4–10.6)	39	8.5 (6.1–12.0)
VIP	55–65	15 634	864	4.7 (3.7–7.0)	501	5.6 (2.9–9.5)	277	4.9 (2.8–8.1)	215	6.5 (3.2–10.3)

VIP = Västerbotten Intervention Project; PSA = prostate-specific antigen; CO = cutoff (PSA ≥3.0 ng/ml); LT = lead time; IQR = interquartile range; LGC = low-grade cancer; HGC = high-grade cancer.

^aNote that the table includes patients with PSA >10 ng/ml or missing grade, whereas the main results are for men with PSA of 3.0–10 ng/ml and data on grade.

^bDoes not include cancers discovered at autopsy with no prior evidence of metastasis.

As shown in Table 2, there was a significant increase in the risk of high-grade diagnosis with lead time after a PSA measurement of 3.0–10 ng/ml among Malmö participants aged 45–65 yr and VIP participants aged 60 yr (all p ≤ 0.002). Results for cutoff points of ≥2.0, ≥2.5, ≥3.5, and ≥4.0 ng/ml were similar (Supplementary Table 2). Although we did not find an association for 50-yr-old VIP men (p = 0.13) the odds ratio is close to that reported for the 60-yr-old VIP men (1.08 vs 1.12). For Malmö participants aged 35–45 yr, we did not find evidence of an association between lead time and the probability of a high-grade diagnosis, although there were only five events and the confidence intervals (CIs) were wide. There were not enough data to assess the relationship between lead time and the probability of a high-grade diagnosis for VIP participants aged 40 yr with baseline PSA of 3.0–10 ng/ml (only 4 high-grade cases). Estimates from the additional sensitivity analyses were very similar if high-grade disease was defined as ≥3 ng/ml (Supplementary Table 3) and if patients with metastatic disease at the time of diagnosis were excluded (Supplementary Table 4). Figure 3 shows the probability of a high-grade and low-grade diagnosis by lead time over the lead-time density following elevated PSA of 3.0–10 ng/ml. The probability of a high-grade diagnosis increased with the lead time. When we combined all cohorts and ages into one data set, the odds ratio for a high-grade cancer and diagnosis was 1.13 (95% CI 1.10–1.16) after adjustment for cohort and age. There was no evidence of heterogeneity between the cohorts and ages (p > 0.3 for all), suggesting that the lack of association seen in the younger age groups was because of limited event numbers rather than true differences by age.

Table 2.

Univariable logistic regression testing the association between lead time and probability of a high-grade versus low-grade cancer diagnosis among men diagnosed with cancer by study and age. Participants with prostate-specific antigen of 3.0–10 ng/ml at screening were included in the analysis

Study	Age (yr)	Sample size (n)	High-grade cancers (n)	OR (95% CI)	p value
Malmö	35–45	14	5	0.99 (0.77–1.27)	0.9
Malmö	45–55	133	51	1.09 (1.03–1.16)	0.002
Malmö	55–65	242	100	1.13 (1.07–1.20)	<0.0001
VIP	40	6	4	–	–
VIP	50	107	31	1.08 (0.98–1.19)	0.13
VIP	60	399	155	1.12 (1.06–1.17)	<0.0001

VIP = Västerbotten Intervention Project; OR = odds ratio; CI = confidence interval.

Fig. 3.

Probability of high-grade (solid line) and low-grade (dashed line) prostate cancer by lead time among participants aged 45–70 yr with prostate-specific antigen of 3.0–10 ng/ml at baseline blood draw. The shaded area denotes the shape of the distribution of lead times rescaled for presentation. PCa, prostate cancer; PSA, prostate-specific antigen.

As exploratory analyses, we examined the association between PSA and lead time and whether PSA modified the association between lead time and grade. Higher PSA predicted a shorter lead time by 0.46 yr (95% CI 0.28–0.64; p < 0.0001) per 1 ng/ml increase in PSA, after adjusting for age and cohort. Estimates were similar after adjusting for grade at diagnosis, and the interaction term between PSA and grade at diagnosis was not significant (p = 0.8). PSA levels at the time of blood draw corresponding to the first PSA elevation above 3.0 ng/ml were very similar for men subsequently diagnosed with low-grade cancer (median 4.5 ng/ml, interquartile range 3.6–5.8) versus high-grade cancer (median 4.6 ng/ml, interquartile range 3.7–6.1).

4. Discussion

We found that the risk of high-grade prostate cancer increased with the lead time. This supports the hypothesis that development of high-grade cancer is a late event, such that serial sampling of a prostate in man destined to have a high-grade tumor would first reveal low-grade disease.

Two well-known observations about prostate cancer provide further support for this hypothesis. The first is grade progression on active surveillance. Many patients with low-grade cancer are managed conservatively, with regular biopsy and clinical evaluation to determine whether there is an indication for active treatment. There is considerable evidence in the literature that men initially diagnosed with low-grade cancer are subsequently found to have higher-grade cancer at a subsequent active surveillance biopsy. For instance, in a multi-institutional study of active surveillance, Eggener et al [11] found that 25% of patients were treated by 5 yr, with 35% of those because of Gleason upgrading, suggesting close to a 10% risk of upgrading at 5 yr. Although this might suggest grade progression, at least some of this effect may represent initial undersampling, with high-grade cancer that was always present in the prostate being missed on one or more diagnostic biopsies and subsequently picked up on follow-up biopsies. Hence, grade progression on active surveillance provides necessary but not sufficient evidence of grade progression.

The second major line of evidence that grade increases over time is the association between age and grade at diagnosis. Older men are more likely to present with high-grade cancer, to the extent that age is included as a predictor in many prediction models, such as the Prostate Cancer Prevention Trial risk calculator [12]. For a typical patient, the prevalence of high-grade disease among men with positive biopsy rises from 25% to 29% to 35% for ages 55, 65, and 75 yr, respectively. That said, the increase in the risk of high-grade cancer with age does not unequivocally demonstrate grade progression. It could, for instance, be that the cancers emerging in an older prostate have a greater propensity to be of high grade.

In contrast to the data on age and active surveillance, it is very difficult to explain our finding that lead time is positively associated with grade in any terms other than grade progression. If grade is stable over time, high-grade tumors would only be clinically detected later if it were the case that they are less likely to cause signs or symptoms. Given that high-grade tumors are more aggressive, this seems extremely unlikely. Alternatively, an obvious limitation of our study is our assumption that a man with elevated PSA at time T0 who is found to have cancer at some subsequent time T1 did indeed have biopsy-detectable cancer at T0. It would therefore be possible to argue that grade is stable if it is also assumed that the probability that the cancer would be biopsy-detectable at T0 depends on whether it is high-grade at T1. However, there are no particular data or biological rationales for this position and one clear line of disproving evidence. It is known that high-grade cancer is more likely to be found at biopsy in men with higher PSA, but we found that lead time is shorter with higher PSA.

A second limitation of our study is that PSA was obtained only once or at a long interval, such as every 10 yr in the VIP cohort. This may underestimate lead time. For instance, a man in the VIP cohort with PSA of 2 ng/ml at age 50 yr, 8 ng/ml at age 60 yr, and clinical detection of prostate cancer at age 68 yr would be given a lead time of 8 yr. However, if this man had been undergoing regular screening in his 50s, his PSA would likely have become sufficiently elevated to trigger biopsy before the age of 60 yr. That said, at least some men with elevated PSA who were subsequently diagnosed with cancer would not have had biopsy-detectable cancer at the time of blood draw. This would lead to overestimation of lead time. For instance, a man aged 60 yr with PSA of 4 ng/ml and diagnosis of prostate cancer at age 70 yr would be given a lead time of 10 yr. However, such a man might have had a negative biopsy at age 60 yr, but then PSA of 8 ng/ml at age 65 yr would lead to a repeat biopsy with cancer found. The true lead time would be 5 yr, rather than the 10 yr we would estimate. As these two influences on lead time are countervailing, we anticipate that our study methods would not incur a large bias in estimation on lead time distributions. More importantly, any such biases are independent of the key hypothesis addressed in this paper, which concerns the relationship between lead time and grade. If we were to underestimate lead time because of infrequent PSA data, there is no reason to believe that this would affect lead times differentially according to grade, because PSA levels at baseline blood draw were very similar irrespective of grade at diagnosis. Moreover, median PSA at baseline blood draw was close to 5 ng/ml, which is not markedly higher than the common cutoff point of ≥3.0 ng/ml used in the European ERSPC trial and in the UK ProtecT trial. This suggests that there was no undue difference between the time of diagnosis in the current study cohorts and when cancer would have been detected during regular screening.

An additional limitation is that the prostate cancer grading at diagnosis was not independently reviewed in Malmö, whereas data on tumor grade at diagnosis for VIP participants subsequently diagnosed with prostate cancer were based on a review supervised by a senior genitourinary pathologist. We conducted several analyses to determine how this might have affected our results, but found no evidence suggesting bias with respect to our key findings (Supplementary material).

5. Conclusions

In conclusion, we found that men with a longer lead time between elevated PSA and subsequent prostate cancer diagnosis were more likely to have high-grade cancer at diagnosis. This supports the grade progression hypothesis, whereby prostate cancer followed over time exhibits a transition from benign to low-grade and then to high-grade cancer. However, we cannot know whether a given low-grade lesion becomes high-grade cancer, or whether a new high-grade focus arises in a prostate that already contains a low-grade cancer.

The probability that a cancer will be of high grade at diagnosis increases with the lead time. Our findings provide evidence of grade progression, whereby a prostate followed over time would exhibit transitions from benign to low-grade to high-grade prostate cancer.