Determining Good Practice
For BCG Therapy Of Urinary Tract Cancer By Using A Knowledge Model
Sachiko Iwai, Fukuya Ishino, Waseda University , Japan
ABSTRACT:
This study aimed to predict the recurrence of carcinoma following Bacillus Calmette-Guérin (BCG) therapy, an effective treatment for urothelial carcinoma. In everyday practice, cytodiagnosis of cells of the whole urinary tract is performed because it is not possible to identify and isolate the bladder or the cancerous cells of the upper urinary tract and the affected sites. We performed statistical analysis on the entire urinary tract using urine cytology specimens of patients with urinary tract cancer who had undergone BCG therapy. Subsequently, we developed a formula for predicting the recurrence of urothelial carcinoma by statistical analysis of 17 parameters in each patient. Validation based on this prediction formula yielded a hitting ratio of 80.4%. The number of BCG injections was minimized to avoid excessive side effects, and we determined the relationship between cancer recurrence, treatment and side effects. These results provide good practice guidelines for the treatment of individual patients.
Keywords: Good practice, Knowledge
model, BCG therapy, Recurrence, Urinary tract cancer
1. Background
Bacillus Calmette-Guérin (BCG) therapy is regarded as one of a number of
effective treatments for urothelial carcinoma and
has been used in many patients as a gold standard.（Japanese
Urological Association , The Japanese Society of Pathology , 2001） However, BCG causes severe side effects and
therefore many patients prematurely terminate their treatment. Several studies
have investigated the relationship between BCG dose and adverse events, and
reported that a high dose (80 mg) causes more severe side effects than a lower
dose (40 mg). However, many centers continue to use the highdose of BCG, based
on previously
established treatment guidelines, because a
lower dose of BCG may result in recurrence.
In this study, we used urinary cytology samples from patients with
urothelial carcinoma who received BCG therapy and statistically analyzed 17
parameters to establish a formula for recurrence prediction. The formula that
we established using “number of BCG dose”and “last positive
time of cytology” (Okubo, 1997; Rishmann, 2006) was considered to be simple and useful.
2. Materials And Methods
2.1. Data Collection For
Establishing A Formula For Recurrence Prediction
We selected those patients with urothelial carcinoma from all patients who
had been treated at the Department of Urology,
2.2. Collected Information And
Establishment Of Parameters As Data Preprocessing
The following 17 variables were used (Table 1): “recurrence”,
“gender”, “age”, “BCG dose”, “number
of BCG doses”, “last positive time of cytology”, “BCG
infusion into the bladder”, “BCG infusion into the upper urinary
tract”, “histologic type of cancer”, “cancer
grade”, “pain (side effect)”, “pyrexia (side
effect)”, “bladder irritation (side effect)”, “withdrawal
of BCG treatment due to side effect”, “status of urothelial
carcinoma 2 or more years after the
start of BCG therapy (prognosis)”, “death due to diseases other
than urothelial carcinoma 2 or more years after the start of BCG therapy
(prognosis)”, and “combination with other therapy”. The recurrence rate after BCG therapy was
investigated in order to assess the validity of the above variables for our
patients. According to the Japan BCG
Laboratory, among patients in
Table 1: Details
Of 102 Experimental Cases
|
|
|
Total |
Non-recurrence |
Recurrence |
|
Gender |
Male |
84 |
40 |
44 |
|
Female |
18 |
8 |
10 |
|
|
BCG dose |
40 mg |
55 |
27 |
28 |
|
80 mg |
47 |
20 |
27 |
|
|
Cancer grade |
Grade 2 |
32 |
18 |
14 |
|
Grade 3 |
70 |
29 |
41 |
|
|
Histologic type of
cancer |
UC |
88 |
10 |
78 |
|
UC+others |
14 |
4 |
10 |
|
|
Last positive time of
cytology |
40 mg |
47 |
16 |
31 |
|
80 mg |
55 |
29 |
26 |
UC: urothelial carcinoma
3. Results
3.1. Multiple Linear
Regression Analysis Using A Variety Of Parameters: Formula For Recurrence
Prediction
In a multiple linear regression analysis, the presence and absence of recurrence
was expressed as “0” and “1”, respectively, and a
discriminative point as “0.5”.
The independent variable that showed the highest coefficient of multiple determination in a
multiple linear regression analysis with
The standard partial regression coefficient of “last positive time of cytology” had a positive influence (0.672), and its p value was very high (7.26E-05). The standard partial regression coefficient of “number of BCG doses” had a significant negative influence (-0.184), and its p value was high (0.013). The following formula for recurrence prediction was obtained through multiple linear regression analysis using SPSS:
Y = 0.661+0.133＊“last positive time of cytology” – 0.092＊“number of BCG doses”
Table 2: Established Parameters
|
Variables |
Actual number |
Dummy variable: 0 |
Dummy variable: 1 |
|
Presence of
recurrence |
|
Absent |
Present |
|
Gender |
|
Female |
Male |
|
Age |
Actual number |
|
|
|
BCG dose |
Actual number |
|
|
|
Number of
BCG doses |
Actual number |
|
|
|
Last
positive time of cytology |
Actual number |
|
|
|
BCG
infusion into the bladder |
|
No |
Yes |
|
BCG
infusion into the upper urinary tract |
|
No |
Yes |
|
Histologic
type of cancer |
|
Single |
Double |
|
Cancer
grade |
Actual number |
|
|
|
Pain (side
effect) |
|
Absent |
Present |
|
Pyrexia
(side effect) |
|
Absent |
Present |
|
Bladder
irritation (side effect) |
|
Absent |
Present |
|
Withdrawal
due to side effect |
|
Absent |
Present |
|
Conditions
of urothelial carcinoma 2 or more years after the start of BCG therapy
(prognosis) |
|
Favorable |
Death due to urothelial carcinoma |
|
Death due
to diseases other than urothelial carcinoma 2 or more years after the start
of BCG therapy (prognosis) |
|
Favorable |
Death due to other diseases |
|
Combination
with other therapies |
|
TUR or biopsy alone |
TUR + other therapies |
TUR: Transurethral resection of the bladder tumor
3.2. Evaluation
Of The Validity Of The Formula For Recurrence Prediction
The validity of the formula for recurrence prediction was evaluated by
statistical analysis of the data from the 102 cases. The formula was established based on the data
from these 102 cases; however, we could not perform a true validation using
these data alone. Therefore, to prove
the robustness of our formula, the evaluation was performed using stored data
from another 37 cases receiving BCG therapy at our center or other hospitals
(Munakata Suikokai General Hospital, n=1; Kushu Rosai Hospital, n=10; Nippon
Steel Yawata Memorial Hospital, n=4; Moji Rosai Hospital, n=8; and University
Hospital of Occupational and Environmental Health, n=14). The hitting rate in these 37 cases was 86.5%
(32/37), and in the 102 cases used for statistical analysis it was 80.4%
(82/102). The hitting rate in all cases
was 82.0% (114/139).However, we feared complications when using predictive
discriminate in actual clinical settings and therefore developed a simplified
chart (Table 3). The hitting ratio of recurrence prediction using this
simplified chart for verification, statistical analysis, and all examples were
34/37 (91.9%), 91/102 (89.2%) and 125/139 (89.9%), respectively, and higher
hitting ratio results were obtained than when using the formula for predicting
the recurrence of cancer.
3.3. Recurrence And Side Effects Of BCG Therapy
Data about the three side effects of pain, fever, and bladder irritation symptom come from a previous investigation (Table 4). The number of BCG injections and the appearance of side effects is given in each case, and the presence or absence of recurrence is noted. If no side effects were reported, the side effect level was considered to be level 0, whereas it was considered to be level 3 if all three side effects.
Table 3: Simplified Chart
|
|
0 |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
|
|
|||||||||
|
Injection time |
|||||||||
|
1 |
0.753 |
0.702 |
0.835 |
0.968 |
1.101 |
1.234 |
1.367 |
1.500 |
1.633 |
|
2 |
0.577 |
0.610 |
0.743 |
0.876 |
1.009 |
1.142 |
1.275 |
1.408 |
1.541 |
|
3 |
0.462 |
0.518 |
0.651 |
0.784 |
0.917 |
1.050 |
1.183 |
1.316 |
1.449 |
|
4 |
0.354 |
0.426 |
0.559 |
0.692 |
0.825 |
0.958 |
1.091 |
1.224 |
1.357 |
|
5 |
0.248 |
0.334 |
0.467 |
0.600 |
0.733 |
0.866 |
0.999 |
1.132 |
1.265 |
|
6 |
0.142 |
0.242 |
0.375 |
0.508 |
0.641 |
0.774 |
0.907 |
1.040 |
1.173 |
|
7 |
0.036 |
0.150 |
0.283 |
0.416 |
0.549 |
0.682 |
0.815 |
0.948 |
1.081 |
|
8 |
-0.070 |
0.058 |
0.191 |
0.324 |
0.457 |
0.590 |
0.723 |
0.856 |
0.989 |
|
9 |
-0.176 |
-0.034 |
0.099 |
0.232 |
0.365 |
0.498 |
0.631 |
0.764 |
0.897 |
|
10 |
-0.282 |
-0.126 |
0.007 |
0.140 |
0.273 |
0.406 |
0.539 |
0.672 |
0.805 |
|
11 |
-0.389 |
-0.218 |
-0.085 |
0.048 |
0.181 |
0.314 |
0.447 |
0.580 |
0.713 |
|
12 |
-0.495 |
-0.310 |
-0.177 |
-0.044 |
0.089 |
0.222 |
0.355 |
0.488 |
0.621 |
Table 4 shows the relationship between cancer recurrence and side effects of BCG treatment. This table may be used for determining the next injecton in treatment with BCG.
4. Discussion
An inappropriate decision was made in five out of 37 investigated cases. Among them, three cases received radical cystectomy after BCG treatment (within 2 years after the start of BCG treatment), and it was highly probable that recurrence would have occurred in these cases if they had not received surgery.
Table 4: Relationship Between BCG Side Effect And Number Of The Appearances Of
Positive Time
|
Side effect level |
Recurrence |
Number of the appearances
of positive time |
||||||||
|
Last positive time of cytology |
||||||||||
|
0 |
1 |
2 |
3 |
4 |
5 |
6 |
7 |
8 |
||
|
0 |
No |
2 |
1 |
4 |
1 |
1 |
2 |
2 |
1 |
1 |
|
Yes |
0 |
0 |
0 |
1 |
0 |
0 |
0 |
0 |
7 |
|
|
1 |
No |
3 |
2 |
5 |
5 |
1 |
3 |
0 |
1 |
0 |
|
Yes |
0 |
0 |
1 |
2 |
2 |
4 |
1 |
1 |
5 |
|
|
2 |
No |
2 |
1 |
1 |
7 |
1 |
0 |
0 |
0 |
0 |
|
Yes |
0 |
0 |
0 |
3 |
1 |
1 |
4 |
1 |
2 |
|
|
3 |
No |
0 |
0 |
3 |
2 |
0 |
0 |
0 |
0 |
0 |
|
Yes |
0 |
0 |
0 |
1 |
2 |
3 |
3 |
1 |
4 |
|
Urinary cytology was performed once a week during BCG therapy, and once every 6 months after the treatment. The inappropriate decision was made in the remaining two cases because the cytology finding at the last positive time did not meet the diagnostic criteria. In particular, a difference in classification between class III and IIIb was thought to result from the observers’ subjective criteria of diagnosis based on the cytology results. It was very important to have a consensus of diagnostic criteria among observers making the diagnosis (e.g., which cellular findings should be taken into account for diagnosis).
As the recurrence prediction model established in this study was expressed as a linear function, its outcome differed from that of the sigmoid logistic curve that is generally used clinically. (Zlotta ,1994) reported that six or more BCG infusions are unnecessary because immunoreactions are maximized after the fourth infusion. However, immunoreactions differed among individuals in general. Table 5 shows the probability of side effects after each BCG injection. Urinary cytology represents the individual reaction of each patient to BCG. Therefore, it might be appropriate to individually determine the optimal number of BCG doses based on the results of cytology.
Table 5 Relationship Between
Cancer Recurrence And BCG Side Effects
|
BCG injection time |
Judgment of cytology |
Probability of side effect appearing |
Actual probability of recurrence |
Prediction probability of recurrence |
Recommended injection time and predictive probability |
|
|
1 |
0 |
0.064 |
0.000 |
0.000 |
(4) 0.293 |
|
|
1 |
0.073 |
0.000 |
0.000 |
(4) 0.426 |
||
|
2 |
0 |
0.276 |
0.000 |
0.058 |
(4) 0.426 |
|
|
1 |
0.016 |
0.000 |
0.191 |
(5) 0.467 |
||
|
3 |
0 |
0.319 |
0.333 |
0.191 |
(5) 0.467 |
|
|
1 |
0.127 |
0.333 |
0.324 |
(7) 0.416 |
||
|
4 |
0 |
0.064 |
0.833 |
0.324 |
(7) 0.416 |
|
|
1 |
0.091 |
0.833 |
0.457 |
(8) 0.457 |
||
|
5 |
0 |
0.106 |
0.769 |
0.457 |
(8) 0.457 |
|
|
1 |
0.145 |
0.769 |
0.590 |
(9) 0.498 |
||
|
6 |
0 |
0.043 |
0.900 |
0.590 |
(9) 0.498 |
|
|
1 |
0.145 |
0.900 |
0.723 |
(11) 0.447 |
||
|
7 |
0 |
0.043 |
1.000 |
0.723 |
(11) 0.447 |
|
|
1 |
0.546 |
1.000 |
0.764 |
(12) 0.488 |
||
|
8 |
0 |
0.021 |
1.000 |
0.764 |
(12) 0.488 |
|
|
1 |
0.327 |
0.000 |
0.989 |
(14) 0.427 |
5. Conclusions
The results of this study could be used to determine the optimal number of BCG doses for prevention of cancer recurrence, while minimizing the side effects of treatment. As a method of resolving the drawbacks of using the same therapy for all patients, we recommend tailor-made medication based on our model for recurrence prediction. With regard to BCG dose, generally 40 or 80 mg is selected, and 80 mg has traditionally been preferred even though more severe side effects are reported; the issue of optimal BCG dose is still controversial. In our study, BCG dose was not a significant variable and there was no significant difference in the results of multiple linear regression analysis between the two doses (40 and 80 mg). We performed a chi-squared test between 40 and 80 mg, and the probability of cancer recurrence was slightly lower with 40 than with 80 mg. Therefore, we highly recommend the use of 40 mg BCG (see Table1). This suggested the multiple low-dose infusion of BCG as an optimal BCG therapy protocol.
The relationship between cancer recurrence and treatment side effects has been investigated in this study.
6. References
Japanese Urological Association , The Japanese Society of Pathology .(2001),General Rule for Clinical and Pathological Studies on Bladder Cancer,3,33-42
Okubo K. (1997), Intra renal Baccillus Calmette-Guerin therapy for Ca in situ of the upper urinary tract, Japanese Journal of Urology, :88(3), 386–390.
Rishmann P. (2006), Improving compliance of BCG immunotherapy: practical approaches to managing side effects, European Urology Supplements, 5, 660–662.
Zlotta A. (1994), Evolution of cellular and humoral response against Tuberculin and antigen 85 complex during intravesical treatment with BCG of superficial bladder cancer, Acta Urologica Belgica, 62(3), 63–68.
About the Authors:
Sachiko Iwai is a
graduate school student in Waseda University Graduate School of Information
Production and Systems, and has been working under the supervision of Professor
Fukuya Ishino, Chairman of the Information Management Networks Laboratory. She
has been working in the University Hospital of Occupational and Environmental Health as a cytotechnologist in
the Department of Surgical Pathology.
Contact information: 2-7
Hibikino, Wakamatsu-ku, Kitakyushu-shi,