Quantitative evaluation of bone metastases in patients with advanced prostate cancer during systemic treatment

Authors


M. Noguchi, MD, PhD, Department of Urology, Kurume University School of Medicine, 67 Asahi-machi, Kururume, Fukuoka, 830–0011, Japan.
e-mail: noguchi@med.kurume-u.ac.jp

Abstract

OBJECTIVE

To assess the clinical usefulness of quantifying the extent of disease on bone scans in monitoring treatment response in patients with advanced prostate cancer, using computer-assisted image analysis.

PATIENTS AND METHODS

The percentage of the positive area on the bone scan (%PABS) was quantified automatically using a personal computer with an image-analysis program. Serial measurements of %PABS in 42 patients with bone metastasis from prostate cancer followed for a mean (range) of 33 (4–72) months with hormonal therapy were compared with those of the extent of disease (EOD) grades in bone lesions and serum prostate specific antigen (PSA) levels according to treatment response.

RESULTS

Serial values of EOD grades and %PABS decreased during treatment in 11 patients with a partial response and in 12 with progressive disease who had no bone metastasis progression. However, EOD grades and %PABS increased in the remaining 19 patients with progressive disease who had bone metastatic progression. Estimated survival curves showed that %PABS was a useful prognostic indicator, in that patients with a > 25% decline in %PABS survived longer than those with a < 25% decline after treatment (P = 0.0207).

CONCLUSIONS

The %PABS is a simple and reproducible estimate of the proportion of the skeleton involving tumours in patients with advanced prostate cancer, and serial measurements of %PABS can assist in monitoring the treatment response in patients with bone metastatic prostate cancer.

INTRODUCTION

A well-recognized difficulty in assessing the response to therapy for advanced prostate cancer is the infrequency of measurable metastatic disease. The most common metastatic site is bone, and it is manifested by diffuse osteoblastic lesions that cannot be measured reliably to allow assessments of therapeutic benefit. Although serum PSA has been a very sensitive test for detecting treatment failure, it does not indicate the specific sites of failure and is influenced by concurrent hormone treatment [1,2]. It is important to recognize that some agents may modulate PSA independent of their effect on cell growth [3]. A phase II clinical trial in patients with androgen-independent prostate cancer was conducted using carboxyamido-triazole inhibitor, an agent that had been shown to down-regulate the expression of PSA [4]. In this trial, a large proportion of patients had a decrease in their PSA level but their soft-tissue lesions continued to grow, as shown by serial radiographic studies. As the development of new therapeutic regimens and agents must be based on evaluating therapeutic responses, the development of useful, reliable response criteria in prostate cancer remains a subject of intense investigation.

Bone scintigraphy remains the most effective radiological procedure for detecting bone metastasis, with a 1% false-negative rate [5], and for monitoring the progression of bone metastases. Visual analysis of bone scans is a common method to estimate the extent of skeletal disease. Several studies have proposed a grading classification for bone metastasis, based on the pattern of uptake on the bone scan [6,7] or on quantifying the number of sites [8]. However, to quantify all bone metastases is a time-consuming task because patients with metastatic involvement usually have more than one disease site. In addition, the assessment of response and progression varies widely, and is difficult to reproduce in a large population. To more accurately determine the amount of tumour present at baseline and to monitor the tumour's response to therapy, we developed a useful and simple technique to quantify the extent of bone metastasis, i.e. the percentage of the positive area on a bone scan (%PABS), evaluated automatically using computer-assisted image analysis [9].

In the present study %PABS and serum PSA levels were measured serially during hormonal therapy of patients with prostate cancer and bone metastasis. We also determined whether the %PABS was an objective marker for bony metastases compared with the clinical outcome and concurrent PSA tests.

PATIENTS AND METHODS

Forty-two consecutive patients with prostate cancer and bone metastasis were retrospectively assessed by quantitatively evaluating bone metastasis, between January 1995 and December 2001. All patients had newly diagnosed, histologically confirmed adenocarcinoma of the prostate and had not been previously treated. At presentation each patient was evaluated using a standard protocol, with a careful history and physical examination, serum PSA determination (Tandem-R, Hybritech Inc., San Diego, CA, USA; normal range 0–4.0 ng/mL), TRUS and radionuclide bone scintigraphy. All patients underwent CT or MRI of the pelvis.

Patients were clinically evaluated monthly for the first 3 months and every 3 months thereafter during treatment. Biochemical tests, PSA measurements and TRUS were repeated with the same frequency, while a bone scan, chest X-ray and CT were repeated every 6 months after the first three months.

For bone scintigraphy, 99mTc-hydroxy-methane-diphosphonate imaging was obtained after an intravenous injection with radiolabelled tracer (555 MBq, Nihon Mediphysics Co. Ltd, Japan). In all patients bone scintigrams were obtained ≈ 3 h after the injection. Whole body images (scan speed 15 cm/min, matrix 256 × 1024) were recorded with a low-energy, high-resolution collimator (E-Cam, Siemens Medical Systems, Inc., Germany). The whole-body field was used to record anterior and posterior views digitally (256 × 1024) on a dedicated computer system (Toshiba 5500 A/PI, Tokyo, Japan). Energy discrimination was provided by a 10% window centred on the 140 keV emission of 99mTc. The extent of metastatic bone disease was determined according to the extent of disease (EOD) grade [8] and the proportion of the positive area on the bone scan as the %PABS [9].

The total number of lesions was determined by visually counting each discrete lesion using the method described previously [8], and the metastatic findings on bone scans were classified into four groups of EOD: grade 1, < 6 bony metastases (a lesion occupying the entire vertebral body was counted as two lesions); grade 2, 6–20 metastases; grade 3, > 20 metastases but less than a ‘super scan’ (diffuse symmetrical uptake with no visualization of the kidneys); or grade 4, ‘super scan’ or its equivalent (involvement of > 75% of the ribs, vertebrae and pelvic bones).

To provide a simple quantitative measure of the extent of the disease, the bone scans were analysed to give the %PABS. As shown in Fig. 1, all outlines of positive regions on a bone scan (both anterior and posterior) were transferred by tracing to a comprehensive map of the entire bone metastases. Positive areas on a bone scan were converted to image files using a digitizing pad and a drawing program (MacPaint®, Claris Corp., Santa Clara, CA) with a Macintosh computer (Apple, Inc., Cupertino, CA). The sum of all positive areas was measured automatically using an image analysis program (NIH Image, developed and maintained by the National Institutes of Health, Bethesda, MD). The %PABS was calculated as the (positive area on bone scan/square area) × 100, in which the square area was multiplied by the width at the gluteal region to the height of the entire skeleton on the bone scan [9].

Figure 1.

The %PABS is calculated as (positive area on bone scan/square area) × 100, in which the square area was multiplied by the width at the gluteal region to the height of the entire skeleton on the bone scan. (A) An entire skeleton on a bone scan. (B) Direct tracings of the same ‘hot spot’ perimeters in B from the bone scan. (C) Computer tracings with numbering after measurement by image analysis.

The overall response for the study was based on the PSA response, that of measurable lesions (changes in size of lymph nodes or parenchymal masses on physical examination or X-rays), and the response on bone scans using consensus criteria [10,11]. For the PSA response, a complete response (CR) was defined as normalization to < 0.2 ng/mL on at least two PSA estimates and maintained for ≥ 4 weeks. A partial response (PR) was defined as a decrease by half maintained for ≥ 4 weeks, and progressive disease (PD) as an increase of 25% above the nadir. For a measurable disease response, a CR was defined as the disappearance of all target lesions for ≥ 4 weeks, a PR as a decrease of ≥ 30% in the sum of the longest diameters of target lesions maintained for ≥ 4 weeks, and PD as an increase of ≥ 20% in the sum of the longest diameters of target lesions or the appearance of new lesions. For a response of bone metastasis, a CR was defined as the disappearance of all positive areas on bone scans, a PR as a decrease in the extent of the EOD grade, and PD as an increased number of positive sites, increased intensity of the existing lesions, or the two findings concurrently. For the overall response, a CR required a CR of both measurable lesions and bone scans, and normalization of PSA levels. An overall PR required a CR or PR of measurable lesions, bone scans or PSA with no PD of each variable; PD of measurable lesions, bone scans or PSA constituted overall PD.

Correlations among EOD, %PABS and PSA were evaluated using Spearman's rank correlation test. Disease-specific survival curves were calculated using the Kaplan-Meier technique and groups compared using a log-rank test, with P < 0.05 considered to indicate statistical significance.

RESULTS

Table 1 shows the clinical characteristics of the 42 study patients (mean age 69 years, median 71, range 54–82). The mean (median, range) PSA levels and %PABS at diagnosis were 730 (125, 1.4–9400) ng/mL and 6.1  (3.1, 0.4–40)%, respectively. All 42 patients received hormonal therapy with various regimens for a mean duration of 33 (29, 4–72) months. The hormonal therapy included combined androgen blockade, as leuprolide plus flutamide or bicalutamide (20 patients), leuprolide plus estramustine phosphate (13), and bicalutamide (one), leuprolide (five) or bilateral orchidectomy alone (three). The overall clinical responses at the end of follow-up were PR in 11 patients and PD in 31 (12 with no and 19 with progression of bone metastasis).

Table 1.  The clinical characteristics of the 42 study patients
CharacteristicsN (%)
  • *

    Eastern Cooperative Oncology Group performance status;

  • as described previously [8].

Age at diagnosis (years):
50–59  5 (12)
60–6914 (33)
70–7919 (45)
≥  80  4 (10)
Performance status*:
015 (36)
119 (45)
2  5 (12)
3  3 (7)
Bone metastasis (EOD grade):
110 (24)
219(45)
3  8 (19)
4  5 (12)
Biopsy grade:
well differentiated  1 (2)
moderately differentiated20 (48)
poorly differentiated21 (50)
Overall response:
Partial response11 (26)
Stable disease  0
Progression 
with progression of bone metastasis19 (45)
no progression of bone metastasis12 (29)

Table 2 shows the Spearman's rank correlations among EOD grade, %PABS and PSA; the correlation between EOD grade and %PABS at diagnosis was strong, but neither EOD grade nor %PABS correlated strongly with PSA level.

Table 2.  Spearman rank correlations among EOD, %PABS and PSA in 42 patients
 EOD%PABSPSA
EOD1.000
%PABS0.8131.000
PSA0.5140.3761.000

EOD grades, %PABS and PSA levels were measured serially during hormonal therapy in all study patients and the changes evaluated according to the overall response. The EOD grades and %PABS tended to decrease in 11 patients with PR disease (Fig. 2A) and in 12 with PD (Fig. 2B) who had no progression of bone metastasis. Although EOD grades and %PABS in the remaining 19 patients with PD disease who had progression of bone metastasis tended to increase (Fig. 2C), it was difficult to distinguish the degrees of progression of bone metastasis using only changes in the EOD grades. Of the 19 patients who had progression of bone metastasis, 15 were assessed as having the same EOD grade after progression, but the %PABS in all 19 increased after progression of bone lesions. All 31 patients with overall PD were diagnosed by increasing PSA levels, in 19 followed by progression on bone scans.

Figure 2.

Serial measurements of EOD grade, %PABS and PSA in 42 patients with prostate cancer with bone metastasis during treatment: (A) in 11 with PR disease; (B) in 12 with PD and no progression of bone lesions; (C) in 19 with PD and progression of bone metastases. Final: at the end of follow-up; PD (PSA) at the PSA failure; PD (3M), at 3 months after PSA failure; PD (6M), at the 6 months after PSA failure; PD* (Bone) at the progression of bone metastases.

At 3 months after treatment, 16 patients had a ≥ 25% decline in %PABS (group A) and the remaining 26 had a < 25% decline (group B). Figure 3 shows the estimated survival curves in both groups; the %PABS was a useful prognostic indicator, with patients in group A surviving longer than those in group B (P = 0.0207).

Figure 3.

Kaplan-Meier estimates of survival for the two groups (A green circles and line; B, red squares and line) of patients with decreasing %PABS at 3 months after treatment. There were three deaths from prostate cancer in group A and 16 in group B (P = 0.0207). The mean time to death was 52.3 months in A and 37 in B.

DISCUSSION

The usual method of monitoring the response of bone metastases to therapy is by combining qualitative assessments of sequential bone scans and bony films with serum PSA level. However, to quantify all bone metastases in patients is a time-consuming task, as patients with metastatic involvement usually have more than one disease site. Furthermore, when visually counting the total number of bone lesions on a bone scan which shows many bone metastases, it is easy to make an error. Several studies have evaluated reproducible ways to quantify changes in serial bone scans in relation to other clinical outcomes [12–16]. Dann et al.[12] evaluated a simple and reproducible technique to measure the 24 h whole-body retention (skeletal uptake) as an objective marker for bony metastases compared with the clinical outcome. This method could be easily automated, but much useful information was lost in this simplified approach, e.g. the anatomical information about which bones are involved or progressing. As a result, it has not been adopted into routine clinical practice. Another method developed a bone scan index to more accurately quantify the extent of the skeletal involvement by the tumour. This method is based on a subjective interpretation of the bone scan, in which the fraction of each bone involved is estimated visually [13]. The technique is complex for calculating the sum of the involved area in the 158 bones, by summing the fraction multiplied by the percentage of the skeleton for each bone, and has shown a variation of 0–50% among individual studies in the estimates [14,15]. Techniques that use computer-generated ‘regions of interest’ for regional bone uptake have been attempted in the bone scan index method [16]. Expense, long processing times and confusion about the exact region to choose made this technique less than ideal. In the present study, we used the NIH Image program to measure all positive areas on bone scans, transferred by manually tracing to a comprehensive map, and the method gave good accuracy for area measurements, with a 1.54% coefficient of variation [15]. Noguchi et al.[9] reported that the mean (range) coefficient of variation in the %PABS was 1.54 (1.0–2.3)%, while that for numeric counting (EOD grade) was 8.94 (4.8–11.5)% when comparing results of measuring bone scans by five different recorders. The chief advantage of the %PABS method is its simplicity and accuracy when measuring regions of interest on bone scans.

However, the %PABS method has limitations; during the preliminary tracing to a comprehensive map of all bone metastases it is possible to make errors in delineating the lesion, which would generally result in difficulty in assessing lesions in the pelvis, complicated by the three-dimensional nature of the pelvic bone. It is also difficult to clearly trace outlines of positive areas on bone scans, because of the limited sharpness of the bone scan films.

The serial measurements of %PABS and serum PSA levels during treatment, although complementary, showed some qualitative differences. The %PABS was more sensitive overall for detecting the presence of progression in the bone metastasis than was EOD grade or PSA level. In the present study, 15 of 19 patients who had progression of bone metastasis were assessed as having the same EOD grade after progression but the %PABS in all 19 reflected the progression of bone lesions. It was also difficult to predict the progression of bone lesions in patients with PSA failure.

PSA is currently the most widely used tumour marker for diagnosing and following prostate cancer. Increasing values correlate with more advanced stages of disease, but no absolute value can predict tumour stage, or the presence or absence of bone metastases. Lorente et al.[2] detected a PSA level of > 100 ng/mL in eight of 52 patients with no bone metastasis, while in 10 of 48 with bone metastasis the PSA level was < 100 ng/mL. However, an analysis of 110 patients with hormone-refractory prostate cancer showed a significant improvement in the median survival for those with a decline by half or more in PSA level than in those with a decline of less than half after treatment [18,19]. A PSA Working Group recently recommended a standard method for reporting the PSA response in phase II trials [11]. Similarly, the present study showed that the %PABS was a useful prognostic indicator, in that patients who had a ≥ 25% decline in %PABS survived longer than those with < 25% decline after treatment.

The present findings suggest that the %PABS, quantified automatically using a computer program, is a simple and reproducible estimate of the proportion of the skeleton involving tumour in patients with advanced prostate cancer. In addition, serial measurements of %PABS can assist in monitoring the treatment response in patients with bony metastatic prostate cancer. However, no analysis is possible about the recommended frequency; this would be based on how the outcome of the test might provide useful clinical information. Therefore, further studies are needed of serial changes in %PABS during the progression to bone metastasis with more patients followed prospectively.

Abbreviations
%PABS

percentage positive area on a bone scan

EOD

extent of disease

CR

complete response

PR

partial response

PD

progressive disease.

Ancillary