Does Salvage Radiation Therapy Change the Biology of Recurrent Prostate Cancer Based on PSA Doubling Times? Results from the SEARCH Database

Abstract

OBJECTIVE

To investigate whether salvage radiation therapy (SRT) may promote prostate cancer (PCa) transformation to more aggressive phenotypes. To accomplish that, we identified men who underwent SRT after radical prostatectomy for PCa and failed SRT. PSA doubling time (PSADT) was used as a surrogate endpoint for cancer aggressiveness. We compared PSADT calculated before start of SRT and after SRT failure.

METHODS

Of 287 men in the SEARCH database since 1988 who underwent SRT, we detected 78 with SRT failure defined as PSA ≥0.2 ng/mL above the post-SRT nadir. Of these, 39 had PSADT available before and after SRT, which was compared using Wilcoxon’s paired test with men serving as their own controls. We tested predictors of PSADT change using multivariable logistic regression.

RESULTS

There were no differences in PSADT before and after SRT (10.2 vs 12.6 months; P = .46). However, in some individual cases, large changes were observed. Only seminal vesicle invasion showed a trend towards an association with a shorter post-SRT PSADT relative to the pre-SRT PSADT (P = .13).

CONCLUSION

Overall, the PSADT after and before SRT were statistically identical, suggesting that after SRT failure, PCa does not emerge with more aggressive biological features. Further studies are needed to identify predictors and the clinical relevance of individual PSADT changes noted in our study.

Prostate cancer (PCa) is a costly disease^1,2 and is the second leading cause of cancer death. Radical prostatectomy (RP) is a common treatment for early-stage disease. In most surgical series, at least one in four patients will have a biochemical recurrence (BCR).^3–5 Although PSA recurrence can often be indolent, for others it is a harbinger of metastases and PCa death.^6–8

Salvage radiation therapy (SRT) after RP offers a second opportunity for disease control. Results from small series tend to show less than ideal responses, with only 15–40% of patients achieving and maintaining long-term PSA control.^9,10 Causes for SRT failure may include poor patient selection and timing for SRT, leading to treating patients with disease outside the radiation field or with concomitant subclinical metastasis. Retrospective studies have identified predictors of PSA control, and specific nomograms were developed and externally validated.^11,12 Their applications include clinical practice or clinical trial enrollment. Recent studies have even suggested that SRT improves overall survival, although prospectively randomized trial data remain lacking.^13,14 Furthermore, a recent study suggested that SRT would be preferred to adjuvant XRT because of relatively similar cancer control with better quality of life.¹⁵

Despite its therapeutic role in cancer control, tumor resistance to radiation is a known problem. SRT uses ionizing radiation and produces its effects through a radiochemical process that depends on oxygen. Some genes involved in hypoxia and angiogenesis have been implicated as potential candidates for causing radioresistance, such as hypoxia-inducible factor-1 (HIF-1) and vascular endothelial growth factor (VEGF).¹⁶ Because PCa is a heterogeneous tumor, it is likely that there exist elements within the tumor with distinct radiosensitivity. It is conceivable that radiation preferentially kills the more sensitive, less hypoxic tumor fraction, leaving the more hypoxic fracture free to proliferate. As such, there is the possibility that genetic mutations caused by ionizing radiation may actually accelerate the growth of these radioresistant tumors. If true, it is possible that for some men, SRT may actually be harmful, promoting the growth of more aggressive cancers.

One approach to gauging the biological behavior of PCa is to measure PSA kinetics. Specifically, for men with BCR after RP, PSA doubling time (PSADT) is a validated independent predictor of long-term survival.^8,17,18 In this study, we analyzed patients who underwent SRT for BCR after RP and who progressed after SRT. We hypothesized that SRT may alter the biological behavior of PCa, leading to a shorter PSADT after SRT failure compared with the PSADT before SRT. To test this, we compared pre- and post-SRT PSADT, allowing each patient to serve as their own control.

MATERIAL AND METHODS

Study Population and Definitions for Cancer Progression

Since 1988, a total of 2829 men underwent primary RP at 5 Veterans Affairs Medical Centers (West Los Angeles and Palo Alto, CA; Augusta, GA; Asheville and Durham, NC) and have been followed for cancer outcomes in the SEARCH database.¹⁹ A total of 1012 (35.8%) of these men had cancer recurrence defined as a single PSA >0.2 ng/mL, two values at 0.2 ng/mL, or secondary treatment for a rising PSA. Radiotherapy after recurrence was performed in 535 cases. We excluded 104 patients because of receipt of hormonal therapy in combination with the radiotherapy, 126 patients who were treated with adjuvant radiation therapy within 6 months after surgery (ie, no preradiation PSADT was available), 6 patients with radiation therapy for bone metastasis, and 12 patients with incomplete data, achieving a final population of 287 men who underwent SRT without concomitant hormonal therapy. Of these 287 men, 78 (27.2%) developed a rising PSA after SRT defined as a PSA ≥0.2 ng/mL above the PSA nadir obtained after SRT, confirmed by a second PSA measurement (not necessarily consecutive) that was higher than the first by any amount or the initiation of hormonal therapy. Of these 78 men with SRT failure, we excluded 34 (43.6%) because of missing data to calculate PSADT before and after SRT. We also excluded other 5 patients (6.4%) whose PSA nadir after SRT was greater than or equal to the PSA pre-SRT. They likely represent men who had no response in PSA levels after SRT because of presence of disease outside the radiation field (ie, subclinical micrometastasis). Hence, the remaining 39 patients constitute our final population (Fig. 1).

Figure 1.

Flow chart of the patients considered for the study and the final population after exclusion criteria

PSADT Calculation Procedures and Statistical Analysis

PSADT was calculated by the natural log of 2 (0.693) divided by the slope of the linear regression line of log of PSA over time.⁸ Only samples collected within 2 years before SRT and within 2 years after SRT failure were used, to have a more recent and comparable estimate of PSADT times. The minimum requirement for the interval between first and last sample to calculate doubling times was 90 days.²⁰ Negative and extremely long PSADTs were arbitrarily set to 100 months as previously described.⁸

Because this was a retrospective analysis, the number of PSA values available to calculate PSADT during the before and after SRT time frames often varied. To ensure that any differences in results were not caused by differences in the time frames and number of PSA values used to calculate PSADT, we performed a sensitivity analysis wherein we performed a case-wise comparison of the number of PSA samples available for calculation before and after SRT. We restricted the number of samples used to the lowest number of samples available in each period. For example, if there were 3 samples before SRT and 4 samples after SRT failure, we used only the first 3 samples collected after SRT failure. A similar procedure was repeated in the case of more samples before SRT than after SRT, discarding the “excess” of PSA samples before SRT.

Descriptive statistics were used to characterize the final population of 39 men. We also compared characteristics between men who were included in our analysis to the men who failed SRT but were excluded of the final population (n = 39). Because of the non-normal distribution of PSADT, the before and after SRT PSADTs were compared for differences using the Wilcoxon sign-rank test for paired data. Correlation between PSADT after BCR and SRT was also measured using a non-parametric test (Spearman’s test). Categorical data were compared using the chi-square and Fisher’s exact tests. All tests were 2-sided with an alpha of 0.05. We also used multinomial logistic regression models to explore predictors of increased, stable (<10% change from baseline values) or decreased PSADT after SRT. We included in the model age PSA before surgery and PSA before SRT, Gleason score, nodal involvement, seminal vesicle invasion, extracapsular extension, and positive margins. For all analyses we used Stata 11.0 (StataCorp LP, College Station, TX).

RESULTS

Of the 287 patients submitted to SRT, 78 (27.2%) failed therapy after a median post-SRT follow-up of 66.7 months (IQR 33.4–96.4). In 39 eligible men of this population (50.0% of men who failed SRT), we could calculate the PSADT both before and after SRT failure and had a decline in PSA in response to SRT, suggesting the presence of local recurrence. Median age at surgery was 63.0 years (Table 1). Median PSA before surgery and SRT were 10.6 and 1.1 ng/mL, respectively. Approximately 50% of the patients had a Gleason score of 7 and one-third had seminal vesicle invasion on the final pathology report after surgery. One patient had positive lymph nodes. Data on radiotherapy dosage could not be collected in approximately half the patients, but based on available data, mean dosage was 6650 cGy with a narrow range between minimum and maximum values. Compared with the final population of 39 men, men who were excluded had similar age (median 62.0 vs 63.0 years, P = .84). They were also statistically similar regarding PSA values before RP (median 9.6 vs 10.6 ng/mL; P = .27), PSA nadir after SRT (median 0.2 vs 0.1 ng/mL; P = .25), extracapsular extension (30.8 vs 46.2%, P = .16), positive surgical margins (47.4 vs 56.4%, P = .42), seminal vesicle invasion (17.9 vs 33.3%, P = .19), lymph node involvement (2.6 vs 2.6%, P = .23), and Gleason scores (Gleason 2–6: 15.4 vs 28.2%; Gleason 7: 64.1 vs 51.3%; Gleason 8–10: 20.5 vs 20.5%; P = .37). However, excluded men had lower PSA values before starting SRT (median 0.3 vs 1.1 ng/mL; P <.0001).

Table 1.

Baseline data of patients who failed SRT and have PSADT available after BCR and SRT failure

Characteristic	Final population (n = 39)
Age at surgery (y)
Median (IQR)	63.0 (58.0–65.0)
Range	50.0–75.0
PSA before surgery (ng/mL)
Median (IQR)	10.6 (7.0–23.1)
Range	1.5–70.9
PSA before SRT (ng/mL)
Median (IQR)	1.1 (0.8–2.2)
Range	0.3–9.3
PSA nadir after SRT (ng/mL)
Median (IQR)	0.1 (0.0–0.4)
Range	0.0–2.5
Pathologic Gleason score (%)
2–6	11 (28.2)
7	20 (51.3)
8–10	8 (20.5)
Extracapsular extension (%)
Yes	18 (46.1)
No	21 (53.9)
Positive surgical margins (%)
Yes	22 (56.4)
No	17 (43.6)
Seminal vesicle invasion (%)
Yes	13 (33.3)
No	26 (66.7)
Positive lymph nodes (%)
Yes	1 (2.6)
No/not done	36 (97.4)
Persistent detectable PSA nadir after surgery,* n (%)
Yes	23 (59.0)
No	16 (41.0)
Radiation dosage (cGy)†
Median (IQR)	6600 (6600–6660)
Range	6300–7200

^*Defined as PSA nadir within 6 months of surgery ≥0.03 ng/mL.³⁰

^†Dosage available only in 21 patients.

In Table 2, we found no differences in PSADT after BCR and after SRT failure (median PSADT 10.2 vs 12.6 months; P = .46). Time intervals between each consecutive PSA values were also similar (median 129.5 vs 125.8 days; P = .93), but the median number of available PSA values for PSADT calculation after BCR were less than after SRT failure (4 vs 5 values used in PSADT calculation; P = .0167). As a result, the median interval between the first and the last sample selected for calculation was significantly shorter after BCR (11.7 vs 18.3 months; P = .0111).

Table 2.

Comparison between PSADT after BCR and SRT failure

	After BCR (n = 39)	After SRT failure (n = 39)	P value*
PSADT (months)			.46
Median (IQR)	10.2 (5.5–21.9)	12.6 (7.8–23.6)
Range	2.4–100	1.5–100
Interval between consecutive samples (days)			.93
Median (IQR)	129.5 (71.0–178.5)	125.8 (90.0–175.3)
Range	44.0–403.0	46.4–340.0
Number of samples			.0167
Median (IQR)	4 (3–5)	5 (4–7)
Range	2–11	2–15
Time between first and last PSA value (months)			.0111
Median (IQR)	11.7 (5.3–19.2)	18.3 (14.5–22.3)
Range	3.0–23.9	3.2–23.9

^*Wilcoxon’s paired test.

Figure 2 shows the scatterplot of PSADT with the calculated values of each patient after BCR plotted on the x-axis and after SRT failure on the y-axis in the same scale. A 45° angle line was plotted to divide the scatter-plot in two areas. Cases below this line mean the PSADT after SRT failure was shorter than before SRT. Cases above this line represent the opposite (ie, after SRT, PSADT is longer than before SRT). PSADT after BCR and after SRT were only weakly correlated but did not reach statistical significance (Spearman’s ρ = 0.20, P = .23). Of the 39 patients, 15 (38.5%) had a decreased PSADT after SRT failure when compared with PSADT after BCR, 22 (56.4%) had an increased PSADT, and in 2 (5.1%) PSADT after SRT failure varied <10% from BCR levels and were thus considered stable.

Figure 2.

Scatterplot of the PSADT values after BCR (pre-SRT) and after SRT failure (post-SRT). The line divides the plot in 2 areas, the upper wherein the post-SRT PSADT was increased (recurrence hypothetically better) and the lower wherein it was decreased (worse).

In secondary analysis, we restricted PSADT calculations to use an equal number of PSA values before and after SRT, which resulted in two more cases being excluded because of only 2 samples with an interval of <90 days between them. The differences between pre-SRT and post-SRT PSADT were largely unchanged and remained nonsignificant (P = .08).

In a further secondary analysis, we explored predictors of increased or decreased PSADT after SRT. On multivariate analysis, no factors were significantly associated with increased or decreased PSADT. However, there was a suggestion that men with seminal vesicle invasion had shorter PSADT values after SRT (P = .13).

COMMENT

Prostate cancer is a very heterogeneous disease often requiring a multimodal approach. Previous retrospective studies have suggested that SRT may offer a survival benefit to patients who failed surgery.^13,14 However, given that many men have recurrence despite SRT, which suggests either micrometastases or radiation resistance, coupled with the mutagenic potential of ionizing radiation,²¹ we evaluated whether SRT could actually make recurrent PCa more aggressive among those men who failed SRT. To test this, we compared PSADT after BCR but before SRT with the PSADT after SRT failure.

Some studies have suggested that SRT can improve overall survival¹⁴ and has similar outcomes to adjuvant radiotherapy in selected intermediate- and high-risk cases.²² Despite these suggestions, the overall PSA recurrence rate of SRT remains high and often >50%.^9,23 Although on the whole, SRT may be beneficial, the question this raises is whether SRT could be deleterious for some men. For example, by promoting nonlethal transformation caused by ionizing radiation, SRT could lead to more genetic mutations in some radioresistant tumors, resulting in a faster-growing phenotype. To examine this, we compared PSADT, as a surrogate of tumor growth, before and after SRT to assess whether PSADT changed after SRT among patients who failed SRT. In this small pilot study, we found that PSADT after BCR and after SRT failure were statistically similar. However, when individual pre- and post-SRT PSADTs were plotted, we did note that in a substantial number of men, the post-SRT failure PSADT was shorter. This was balanced by the fact that in 50% of men, the post-SRT failure PSADT was longer. Indeed, the overall correlation between the two PSADT measurements was weak. Thus, we found no evidence that on the whole, SRT changed the biology of PCa among men who failed this therapy. Nonetheless, we cannot rule out whether a small percentage of men with shorter post-SRT PSADT may have been harmed.

The primary measure we used to assess tumor growth was PSADT. Although PSA is not a cancer-specific but a prostate-specific marker, it is correlated with tumor burden.²⁴ Moreover, for men after RP wherein in all (or nearly all) sources of benign PSA production are removed, PSA is more likely to reflect “pure” cancer kinetics. Furthermore, PSADT is a strong predictor of cancer-specific survival after BCR.⁸

There is great variation in the literature about how to calculate PSADT. In our study, we only used PSA values ≥0.2 ng/mL and 2 years before SRT and 2 years after SRT failure to calculate PSADT based on prior work showing that PSADT measured in a similar fashion is a strong predictor of metastasis,⁶ PCa-specific survival,⁸ and overall survival.^18,25 In addition, factors, such as number of PSA tests and time over which they are measured²⁰ may influence the results. In fact, in a cohort of men on active surveillance, PSADT calculated using short-term PSA measurements correlated weakly with the overall PSADT using the totality of available samples, suggesting that obtaining more PSA values results in a more “reliable” PSADT.²⁶ Thus, we were also concerned that differing time points over which PSADT could have been measured in the pre- and post-SRT setting may have influenced our results. Indeed, we found that the post-SRT failure PSADT was based on one additional PSA sample and 6 months longer between the first to the last sample compared with the pre-SRT PSADT. To address this, in a subanalysis we restricted the pool of PSA samples used for PSADT calculation to the same number of samples before and after SRT. After this was done, the results were largely unchanged, supporting our initial findings. Thus, we concluded that our statistical approach to calculating PSADT did not influence our conclusions.

Understanding which factors predict a shorter post- SRT PSADT would be useful to identify high-risk men and target these selected men for early additional treatments. Among the variables studied, we found that none were significantly related to PSADT change. However, we did note a nonsignificant trend for men with seminal vesicle invasion to have a shorter PSADT after SRT failure compared with the pre-SRT PSADT (P = .067). Previous studies have shown that seminal vesicle invasion is an aggressive pathologic feature associated with a short PSADT after RP.^27,28 Moreover, seminal vesicle invasion has been shown to predict poorer responses to SRT.²⁹ As such, the trend toward shorter PSADT after SRT among men with seminal vesicle invasion may merely represent clinical progression in these men. Alternatively, this may indeed represent a more aggressive phenotype induced by the ionizing radiation, and these men may be more susceptible to this effect. However, it must be kept in mind that the current study was small and the results were not statistically significant. Further studies are warranted to explore predictors of a shorter post-SRT PSADT.

Our study has some limitations. First, it is a retrospective analysis, so men were not in a standard protocol. In addition, we only analyzed men treated with surgery and SRT without adjuvant hormonal therapy, limiting the results to this population. Also, of 78 men with SRT failure, we were able to calculate PSADT after both BCR and SRT failure in only 44 patients, 5 of whom never had declines in PSA, resulting in a final cohort of only 39 men for analyses. Samples of this size usually are limited to detect subtle differences between groups, although this was in part compensated by the more sensitive paired statistical test we used. Still, we fully acknowledge the limited statistical power of our study to detect small differences in PSADT before and after SRT. Moreover, although an ideal study would include only low-risk men who had delayed PSA failures after surgery enriching for those who are most likely to have had local-only recurrence, our small numbers prohibited us from taking this approach. Finally, some men who failed surgery and receive SRT directed to the surgical bed may actually have distant subclinical metastasis instead of a local recurrence before starting SRT. As a result, PSA-producing tissue outside the irradiated field would not be affected by ionizing radiation and may confound the interpretation of tumor kinetics changes after SRT based on PSADT calculation. We attempted to address this by excluding men in whom the PSA never dropped after SRT, suggesting either no or minimal local disease.

Although we cannot completely rule out the possibility that SRT may change the biology of PCa for some men, we surmise that it is unlikely to harm a significant number of patients. In our data, 73% of men did not recur after SRT, although with longer follow-up this percentage may go down. Of the remaining 27% who failed SRT, among those with available PSADT data, fewer than half had a shorter post-SRT PSADT, suggesting a recurrence with a faster growth rate. Finally, because small decreases in PSADT may not be clinically relevant, it is also possible that not all of these men with a “shorter” PSADT were “harmed.”

CONCLUSIONS

We found no evidence that on the whole PCa emerges with a more aggressive biological growth as measured by PSADT after SRT. Although for some men the PSADT after SRT was shorter, the clinical relevance of this is unclear but requires further study.