Fax: (773) 702-4347
Hormone therapy adjuvant to external beam radiotherapy for locally advanced prostate carcinoma
A complication-adjusted number-needed-to-treat analysis
Article first published online: 20 OCT 2003
Copyright © 2003 American Cancer Society
Volume 98, Issue 11, pages 2351–2361, 1 December 2003
How to Cite
Jani, A. B., Kao, J. and Hellman, S. (2003), Hormone therapy adjuvant to external beam radiotherapy for locally advanced prostate carcinoma. Cancer, 98: 2351–2361. doi: 10.1002/cncr.11804
- Issue published online: 17 NOV 2003
- Article first published online: 20 OCT 2003
- Manuscript Accepted: 22 AUG 2003
- Manuscript Revised: 11 AUG 2003
- Manuscript Received: 10 JUN 2003
- prostate carcinoma;
- number needed to treat;
- hormone therapy
Hormone therapy commonly is used to treat metastatic, locally advanced, and localized prostate carcinoma. The objective of the current investigation was to determine, using the number-needed-to-treat (NNT) method, the effect of using hormone therapy to treat locally advanced disease, with consideration given to both the complications and the known advantages associated with hormone therapy.
A literature review was performed to determine 1) the absolute benefit, based on available clinical endpoints, associated with the addition of hormone therapy to external beam radiotherapy for locally advanced prostate carcinoma; 2) the incidence of side effects of short-term and long-term hormone therapy; and 3) the stepwise progression from biochemical failure to death. A model was constructed to estimate the complication/utility-adjusted survival detriment resulting from the side effects of short-term (≤ 6 months) and long-term (> 6 months) hormone therapy, and the absolute/unadjusted and complication-adjusted NNTs for the addition of short-term and long-term hormone therapy were computed. In all cases, the magnitudes and signs of the NNTs obtained were used to gauge the effect of hormone therapy.
The unadjusted NNTs were positive and in most cases had relatively small magnitudes (the greater the NNT, the smaller the benefit) for both short-term and long-term hormone therapy; these results were expected, and they suggested that there is a strong benefit associated with the use of hormones adjuvant to radiotherapy for locally advanced disease. Adjusted NNTs remained positive and had relatively small magnitudes even after the introduction into the analysis of complications of short-term and long-term hormone therapy. This finding, although weak with respect to the effect of short-term hormone therapy on cause-specific survival, remained robust over the range of values for utility impairment expected from short-term and long-term hormone therapy.
The benefits of short-term and long-term hormone therapy for locally advanced prostate carcinoma appear to be significant and to outweigh the associated side effects. Long-term therapy appears to be better than short-term therapy in terms of virtually all endpoints studied, even when the increased incidence of side effects is considered. The current investigation was successful in the use of the complication-adjusted NNT method for oncologic and radiotherapeutic scenarios in which the results of randomized trials could be summarized, adjusted for treatment toxicity, and individualized to a given patient. Cancer 2003. © 2003 American Cancer Society.
Prostate adenocarcinoma is one of the most common malignancies for which health care intervention is sought.1, 2 Prostate carcinoma is known to be hormonally sensitive,3 and this feature has prompted the use of hormonal manipulation (originally by orchiectomy and later by pharmaceutical agents). As hormone therapy has evolved from surgical castration to medical therapy, it has been used across a wide spectrum of prostate carcinoma presentations.
Early efforts involving the use of hormones were aimed at metastatic disease.4, 5 When results in the metastatic setting were found to be encouraging, attention was shifted to nondisseminated prostate carcinoma. Although hormonal manipulation is used to manage metastatic disease with palliative intent, management of clinically localized prostate carcinoma involves local therapy, with endocrine therapy used as an adjuvant to improve local control and to possibly eliminate occult distant metastases. Locally advanced prostate carcinoma poses its own set of challenges with respect to surgical resection, brachytherapy, cryotherapy, and other local control techniques.6 Consequently, external beam radiotherapy has occupied a central role in the management of locally advanced prostate carcinoma, and thus the following discussion will be restricted to considering external beam radiotherapy as the method of local control.
Many investigators have documented the limited benefits of radiotherapy alone for treatment of locally advanced prostate carcinoma. The use of hormone therapy in combination with external beam radiotherapy has been demonstrated in a number of randomized studies7–14 to improve survival outcome as measured by a variety of endpoints, as described in detail below. The increased efficacy of radiotherapy in the presence of hormones may be due to radiosensitization, to the reduction of local tumor burden, and/or to the killing of occult disseminated cells outside the irradiated volume.
Although hormone therapy has its demonstrated advantages, it also has potential side effects.15–23 In particular, hormone therapy is associated with hot flashes, diarrhea, weight gain, gynecomastia, liver inflammation, erectile dysfunction, loss of libido, and osteoporosis. Furthermore, the incidence and severity of side effects is dependent in large part on the duration of hormone therapy.
The toxicities associated with hormone therapy have not been systematically analyzed in the context of existing randomized trials in which hormone therapy was administered adjuvant to external beam radiotherapy. To accomplish this task, a method of outcome analysis in which complications (in addition to benefits) associated with an intervention can be quantitatively incorporated is needed. Such a method has been used successfully to compare two competing treatment options. The technique, known as number-needed-to-treat (NNT) analysis, involves computation of the number of treated patients required for one additional patient to receive benefit using one treatment method as compared with its alternative.24–27
NNT is defined as the reciprocal of the absolute risk reduction and can be modified and expanded to allow a systematic comparison of survival, morbidity, and associated loss of utility (impairment of quality of life) for two competing oncologic treatment options. The NNT method has been used with success in many areas of medicine, but it has not been used regularly in clinical oncology, except in recent studies performed by our group.28–30 The objective of the current study was to quantitatively evaluate, using the NNT method, the effects of hormone therapy—both its benefits and its side effects–when administered as an adjuvant to external beam radiotherapy for locally advanced prostate carcinoma.
MATERIALS AND METHODS
Because the ultimate objective of the current investigation was to quantify the benefits and detrimental effects of hormone therapy adjuvant to radiotherapy, it is necessary to discuss the model used in some technical detail. The following section describes the methods used to compute the NNT for the different clinical scenarios considered in the current investigation.
First, the raw survival rates achieved with the addition of hormones were considered. In this case, because the complications of treatment are not incorporated into the model, the NNT is computed as follows:
where NNTA = absolute/unadjusted NNT, B1 = benefit/survival in the group receiving hormone therapy, and B2 = benefit/survival in the group not receiving hormone therapy.
Note that in Equation 1, the endpoint being considered varied—the endpoints reported in most randomized studies were biochemical failure–free survival (BFFS), disease-free survival (DFS), local progression–free survival (LPFS), distant metastasis–free survival (DMFS), cause-specific survival (CSS), and overall survival (OS). The duration of hormone therapy has been established to have an impact on these survival endpoints. Thus, NNTA varies as a function of the length of hormone administration—hormone therapy was considered to be either short term (NNTAS) or long term (NNTAL). Short-term therapy was defined as ≤ 6 months of hormone administration, with long-term therapy defined as > 6 months. This cutoff was chosen because Radiation Therapy Oncology Group (RTOG) and Cancer and Leukemia Group B studies of ‘short-term’ therapy typically involve 4–6 months of hormone administration31, 32 and because the difference in incidence and severity of sequelae of hormone therapy appear to increase significantly after 6 months.
Next, the side effects of treatment were considered. Equation A.1 (Table 1) was the starting point for this analysis.33, 34 Equation A.1 is difficult to apply, however, because the precise incidence and severity of each individual side effect is not readily obtainable from the published literature. Thus, an alternative form of Equation A.1 was used. The steps taken to construct the following expression from Equation A.1 are detailed in Table 1.
where NNTC = complication-adjusted NNT, B1 and B2 are the same as in Equation 1, IB = importance of the benefit being measured relative to the harms, H1 = composite rate of the harms under consideration in the group receiving hormone therapy, H2 = composite rate of the harms in the group not receiving hormone therapy, and UH = complication-adjusted utility, incorporating the harms associated with hormone therapy.
|Equation no.||Equation (with comments)|
|where NNTC = complication-adjusted number needed to treat|
|B1i = rate of the ith benefit (i.e., desirable outcome) under consideration using treatment option 1|
|B2i = rate of the ith benefit (i.e., desirable outcome) under consideration using treatment option 2|
|i = index that varies from 1 to x (where x = the total number of benefits)|
|UBi = utility enhancement associated with the ith benefit under consideration|
|H1j = rate of the jth harm under consideration using treatment option 1|
|H2j = rate of the jth harm under consideration using treatment option 2|
|j = index that varies from 1 to y (where y = the total number of harms)|
|UHj = complication-adjusted utility of the jth harm under consideration|
|Although the clinical endpoint (BFFS, LFFS, DFS, or OS) varies, because only one benefit at a time is being considered, a single scale factor, IB, which represents the importance of the benefit being measured, can be used in place of the summation of the individual benefits weighted by their respective utilities. Making this replacement yields Equation A.2.|
|where IB = the relative importance of the benefit with respect to the total harms|
|B1 = the benefit (i.e., desirable outcome as measured by survival rate) using treatment option 1|
|B2 = the benefit (i.e., desirable outcome as measured by survival rate) using treatment option 2|
|H1j, H2j, j, y, and UHj are the same as in Equation A.1|
|Note that IB must appear in the numerator, so that when there are no harms, Equation A.2 reduces to Equation 1 (see Materials and Methods).|
|Because the treatment arms being compared in the current investigation are equal with the exception of one intervention (namely, the addition of hormone therapy), the harms can be condensed into a single factor. Doing so yields Equation A.3.|
|where UH = the composite utility compromise associated with the harms in relation to the benefit|
|H1 = the incidence of the harms using treatment option 1|
|H2 = the incidence of the harms using treatment option 2|
|IB, B1, and B2 are the same as in Equation A.2|
|Rearrangement of Equation A.3 yields a mathematically equivalent expression, Equation A.4.|
|When UH = 1.0 (i.e., when there are no consequences associated with the harms), Equation A.4 reduces to Equation 1 (see Materials and Methods). In all other cases, the denominator is reduced by 1) the harms scaled by the utilit y and 2) the relative importance of the benefit of the intervention with respect to the harm.|
NNTC (Eq. 2) is essentially NNTA (Eq. 1) adjusted for the difference in the toxicity of hormone therapy. B1, B2, H1, H2, and IB can be estimated from the published literature—the only principal parameter for which there is significant uncertainty is UH. UH = 1.0 indicates that there is no consequence related to the harms, in which case Equation 2 reduces to Equation 1, whereas UH = 0.0 indicates that a profound compromise in quality of life (i.e., death or near-death) has resulted from the additional treatment, in which case the denominator of Equation 1 is reduced by the quantity (H1 − H2)/IB, the difference in incidence of the harms scaled by the relative importance of the survival endpoint being measured to the harms. Because of the uncertainty in measuring UH, this parameter was varied over a range of values in the model.
In a manner similar to the one described above for Equation 1, Equation 2 was used to compute the NNT for each survival endpoint under consideration while accounting for the side effects of short-term hormone therapy—the resulting complication-adjusted NNT was referred to as NNTCS. Then, because the effect on survival endpoints and the incidence and utility impact of complications differ for long-term hormone therapy relative to short-term therapy, Equation 2 was applied for each survival endpoint to include the side effects of long-term hormone therapy; the resulting quantity was referred to as NNTCL.
The values entered to compute NNTAS, NNTCS, NNTAL, and NNTCL for each scenario are described in detail below. In all cases, the sign and magnitude of the NNT were used to gauge the effect of the intervention. A positive NNT indicates an overall benefit associated with hormone use, and a negative NNT indicates an overall harm. Furthermore, larger magnitudes (i.e., absolute values) of NNT correspond to weaker differences between the two treatment options. For example, an NNT of 4 indicates that 4 additional patients must be treated for 1 additional patient to survive, whereas an NNT of 40 indicates a much smaller advantage, as many more patients must be treated for 1 additional patient to benefit. The acceptable magnitude depends on the intent of the intervention—whereas an NNT in the hundreds would be considered adequate for a screening intervention, an NNT of approximately 20 or less is necessary to justify most routine clinical interventions.24–30
Survival Benefit Associated with the Addition of Hormone Therapy
A literature review was undertaken to examine the role of hormone therapy in different prostate carcinoma presentations/scenarios. To ensure that the most reliable values possible were being used in the model, only the results of randomized trials were considered. Specifically, randomized studies examining external beam radiotherapy alone versus external beam radiotherapy plus hormone therapy for patients with nonmetastatic T2c–4N0–1 disease (according to the 2003 American Joint Committee on Cancer staging system35) that were referenced in PubMed36 after 1990 (to ensure prostate-specific antigen follow-up), were written in English, and reported 5-year follow-up data were included.7–14 After this preliminary review, the requirement that hormone therapy be delivered with modern hormonal agents was added. Thus, studies involving orchiectomy or the use of estrogens were excluded, because the side effects (and corresponding detriments to quality of life) associated with these treatments are different from the side effects associated with modern agents, such as testosterone receptor antagonists and luteinizing hormone–releasing hormone agonists.
Using the definitions of short-term and long-term described above, each trial then was placed into one of two general categories: 1) radiotherapy alone versus radiotherapy + short-term hormone therapy or 2) radiotherapy alone versus radiotherapy + long-term hormone therapy. In each case, the inclusion criteria (particularly with regard to clinical stage), the survival endpoints used and reported, and the reported benefit based on these stated endpoints were reviewed and tabulated. The main endpoints in the analysis were BFFS, DFS, LPFS, DMFS, CSS, and OS. In using these studies to calculate the composite outcome, the difference in survival outcome was weighted by the power (i.e., number of patients) of each study. It should be noted that this calculation of composite outcome was not intended to represent a formal metaanalysis of the current literature but rather to serve as a means of providing reliable estimates of the parameters to be entered into the quantitative model discussed above. In this manner, NNTAS and NNTAL (Eq. 1) were computed for each survival endpoint.
Adjustment for Side Effects and Relative Importance of Survival Endpoints
Adjustment for side effects
The analysis described in “Survival Benefit Associated with the Addition of Hormone Therapy” was performed without the inclusion of complications. Complications of hormone therapy are well documented in both the short-term and long-term settings; however, modeling the exact incidences and magnitudes of the complications and, more notably, the utility adjustments for such complications is somewhat challenging. The objective of this portion of the investigation was to provide general estimates that could be entered into the quantitative model.
The general complications of short-term hormone therapy include hot flashes, weight gain, erectile dysfunction, loss of libido, fatigue, diarrhea, gynecomastia, liver inflammation, and osteoporosis.15–23 Using these references,15–23 the composite value for the incidence of side effects due to short-term therapy was estimated to be 0.20. Because precise estimates for individual toxicities are not provided in any particular reference, this value was obtained by doubling the approximate hormone withdrawal rate, as hormones usually were withdrawn when side effects were intolerable. Prolonging the duration of hormone therapy increases the risk of each of the side effects listed above, particularly fatigue, osteoporosis, gynecomastia, and sexual side effects.15–23 Based on these references, the composite value for the incidence of side effects of long-term hormone therapy was estimated to be 0.60, a value that also was calculated by doubling the approximate withdrawal rate. Although the authors recognize that these incidence values are approximate and somewhat arbitrary, the model that has been developed is far less dependent on the incidence of side effects than on the detriment to quality of life that results from these side effects. The model included complications due to hormones but not complications due to radiotherapy,37–39 because the goal was to determine the added benefit/detriment associated with hormone therapy, and not with the local modality; it is presumed that the complications associated with the local modality would affect both groups similarly. Thus, in Equation 2 above, for short-term hormone use, H1 = 0.20 and H2 = 0.0. The complications due to hormone therapy were evaluated to be 0.0 in the arm not receiving hormones. This does not imply that there are no side effects due to radiotherapy; it only implies that, as described above, these radiotherapeutic side effects are assumed to be equal in both arms and that any slight differences would affect the calculation inconsequentially. Similarly, for long-term hormone use, in Equation 2, H1 = 0.60 and H2 = 0.0.
The most difficult parameter to obtain from the published literature is UH, the utility compromise resulting from hormone therapy. Therefore, in applying the model, UH was varied over the entire range of possible, clinically relevant values.
Adjustment for the relative importance of the survival benefit
Short-term hormone therapy.
The parameter IB (Eq. 2) reflects the importance of the endpoint in relation to the side effects. A central postulate is that at the lowest value possible for the utility (UH = 0.0, which implies death or near-death due to complications from hormone use), the detriment of hormone therapy exactly negates the CSS benefit. In other words, at the smallest value of UH, the harm associated with hormone therapy (in causing death or near-death) exactly negates the benefit associated with hormone therapy (in preventing prostate carcinoma–related death). Thus, in Equation 2, when UH = 0.0, the left side of the denominator was set equal to the right side of the denominator, and the values of B1 and B2 for CSS were entered to compute IB.
The values of IB for the other endpoints under investigation could be estimated using knowledge of the progression from the earlier stages of disease to the later stages. There is an abundance of literature documenting the stepwise progression of prostate carcinoma,40–46 and based on our review of this literature, Figure 1 was constructed. As indicated in Figure 1A, a stepwise progression is observed from (i) no evidence of disease to (ii) prostate-specific antigen (PSA) failure to (iii) local progression to (iv) development of distant metastasis, and finally to (v) death due to prostate carcinoma. In this figure, the endpoints being analyzed at each stage—BFFS, LPFS, DMFS, and CSS—also are shown. Note that OS is not included in Figure 1A, because OS is an endpoint that carries significant uncertainty, even for patients with advanced-stage prostate carcinoma, as many patients die of comorbidities other than prostate carcinoma. Also note that DFS is not an endpoint that can be extracted cleanly from the stepwise progression, because the definition of DFS involves local failure, distant failure, and (in some studies) biochemical failure. The DFS endpoint would be expected, based on its definition, to lie between BFFS and LPFS, as is shown in Figure 1A and B.
The expected order of the survival endpoint values at the 5-year mark, based on the stepwise progression in Figure 1A, is shown in Figure 1B. Although this panel demonstrates the idealized order, it should be noted that Figure 1A, on which Figure 1B is based, displays the progression of untreated prostate carcinoma. Treatment with a local modality, such as radiotherapy, alters the observed failure pattern, so that the LPFS curve may not necessarily lie below the DMFS curve as shown in Figure 1B. In fact, the observed values for LPFS and DMFS at 5 years (Table 2), for both short-term and long-term therapy, are similar for this reason.
|Study||Arm||n||Parameter||5 yr endpoint|
|RT vs. RT + short-term HT|
|Pilepich et al., 200110||RT + HT||226||B1||0.28||0.49||0.78||0.71||0.85||0.72|
|B1 − B2||0.18||0.15||0.13||0.10||0.05||0.04|
|Laverdiere et al., 19978||RT + HT||45||B1||n/a||n/a||0.82||n/a||n/a||n/a|
|B1 − B2||n/a||n/a||0.47||n/a||n/a||n/a|
|Composite results||B1 − B2||0.18||0.15||0.17||0.10||0.05||0.04|
|RT vs. RT + long-term HT|
|Lawton et al., 200111||RT + HT||477||B1||0.54||0.62||0.85||0.85||0.97||0.75|
|B1 − B2||0.33||0.18||0.16||0.14||0.16||0.04|
|Bolla et al., 200212||RT + HT||203||B1||0.76||0.74||0.98||0.90||0.94||0.78|
|B1 − B2||0.31||0.34||0.14||0.19||0.15||0.16|
|Composite results||B1 − B2||0.32||0.23||0.15||0.15||0.16||0.08|
Using this stepwise progression model, the calibration factor IB computed for CSS will be close in magnitude to IB for the other endpoints being studied (i.e., DMFS, DFS, LPFS, and BFFS). This statement can be made because, based on the model shown in Figure 1, progression occurs along a stepwise pathway, so that patients with biochemical failure ultimately will experience prostate carcinoma–specific death. Thus, once the constant IB has been calibrated for CSS, its value can serve as a general estimate for the preceding endpoints, from BFFS through DMFS. Of course, a small percentage of patients will not experience the stepwise progression (e.g., a patient may die of prostate carcinoma before PSA failure is detected), but the small errors introduced by these outliers are not likely to affect the model significantly.
Due to the comorbidities encountered in the prostate carcinoma population, OS could not be analyzed in the same manner as the other endpoints. The construction of a model that includes, in detail, the magnitudes of these comorbidities as a function of PSA, disease stage, tumor grade, and patient age is beyond the scope of the current investigation, and OS therefore was excluded from the formal utility-adjusted NNT analyses described below.
Long-term hormone therapy.
A similar analysis to the one described above for short-term hormone therapy was performed for long-term hormone therapy. Because the same arguments concerning the natural history of prostate carcinoma apply whether the disease is treated with long-term therapy or short-term therapy, the resulting calculations regarding the importance of benefit relative to side effects are similar to those described above for short-term hormone therapy. Based on the same sequence of computations as used above for short-term therapy, IB was computed for long-term hormone use, this time using the values of B1, B2, H1, and H2 for long-term hormone use with respect to the CSS endpoint.
As was the case for short-term hormone therapy, the value of IB that was calculated for CSS was used for the BFFS, DFS, LPFS, and DMFS endpoints, and for the reasons described above, OS was excluded from the analyses.
The values of the parameters described in the Materials and Methods section were entered into Equations 1 and 2 to achieve the primary goal of the current investigation—the ascertainment of the impact of hormone therapy, both short-term and long-term, in patients with locally advanced prostate carcinoma. Thus, NNTs were computed for a variety of clinical scenarios and endpoints.
Table 2 summarizes the findings of the unadjusted NNT analyses (i.e., the NNTs computed using Equation 1) for short-term hormone therapy (NNTAS) and for long-term hormone therapy (NNTAL). The NNTs obtained for all endpoints are positive and have relatively small magnitudes for both short-term and long-term hormone therapy. This finding reaffirms the established benefit of hormones when used along with external beam radiotherapy for treatment of locally advanced prostate carcinoma.
Next, Equation 2 was used to compute the utility adjustments. The quantity entered into the model for UH was varied over a range of values. This was done in part because UH was difficult to obtain from the published literature and in part to demonstrate the robustness of the model over the extremes with respect to reductions in quality of life due to hormone therapy use. For each value of UH, the NNTC was computed using Equation 2. This computation was performed for both short-term and long-term complications (resulting in the computed quantities NNTCS and NNTCL, respectively). The values of B1 and B2 were obtained from Table 2 in each case, and the values of H1, H2, and IB were determined as described in the Materials and Methods section.
Table 3 shows the results obtained from the analyses of short-term and long-term hormone use for all of the survival endpoints examined. In this table, when UH = 1.0, Equation 2 reduces to Equation 1, and the computed NNTC values reduce to the corresponding unadjusted (NNTA) values shown in Table 2. This occurs because UH = 1.0 implies that the side effects have no consequences, regardless of their incidence. The NNT calculations are most relevant to clinical practice when they use the other values of UH (0.4–0.8) listed in Table 3. That is, the NNTs computed using values of UH that range from 0.4 to 0.8 are likely to represent the actual clinical range of NNTs, because hormone use typically does not lack consequences (i.e., UH is not likely to equal 1.0), but in the vast majority of cases, the hormones do not have severe side effects (i.e., UH is not likely to be less than 0.4). In fact, in one of the largest series examining the detriment to quality of life associated with long-term hormone therapy, the lowest overall quality-of-life functioning score reported was 56.716 (which corresponds to a UH of approximately 0.6), and this value was obtained for patients with metastatic disease that was not in remission after therapy. Although other studies15, 17–22 do not specifically quantify utility compromise due to hormone therapy, they do report incidences and severities of hormone side effects that are similar to the ones described in the current report. Thus, even UH = 0.4 is a generous estimate of the utility compromise due to long-term hormone therapy.
|UH = 1.0b||5.6||3.1|
|UH = 0.8||5.9||3.5|
|UH = 0.6||6.3||3.9|
|UH = 0.4||6.7||4.5|
|UH = 1.0b||6.7||4.3|
|UH = 0.8||7.1||5.1|
|UH = 0.6||7.7||6.0|
|UH = 0.4||8.3||7.5|
|UH = 1.0b||5.9||6.7|
|UH = 0.8||6.3||8.5|
|UH = 0.6||6.7||11.6|
|UH = 0.4||7.1||18.5|
|UH = 1.0b||10.0||6.7|
|UH = 0.8||11.1||8.5|
|UH = 0.6||12.5||11.6|
|UH = 0.4||14.3||18.5|
|UH = 1.0b||12.5||6.3|
|UH = 0.8||25.0||7.8|
|UH = 0.6||33.3||10.4|
|UH = 0.4||50.0||15.6|
Considered individually, nearly all of the NNTs for BFFS, DFS, LPFS, DMFS, and CSS are positive and relatively small in magnitude (< 20), which indicates a true, strong benefit associated with hormone therapy, even when adjustments are made for complications. The only exceptions are the NNTs computed for CSS after short-term hormone therapy—these NNTs are positive but relatively large in magnitude (25.0–50.0), which indicates a relatively weak benefit associated with short-term hormone use; this benefit is strengthened significantly (NNT range, 7.8–15.6) when hormone use is continued for longer periods. In fact, for virtually all endpoints, although higher incidences of side effects were observed when hormones were administered for longer durations, in the range of UH values likely to represent the majority of clinical situations, using hormones for longer periods resulted in decreased (i.e., improved) NNTs.
As demonstrated by the application of the NNT method, the benefits of short-term and long-term hormone therapy for locally advanced prostate carcinoma appear to be significant and to outweigh the side effects over the range of UH values that represent the bulk of clinical situations for the patient population in question. These results indicate the high efficacy of hormone therapy, which already had been demonstrated in previous randomized studies; however, the current study takes the additional step of incorporating complications due to hormone therapy into the analysis.
Several points regarding the performance of error analysis warrant mentioning. Strictly speaking, the estimation of each individual parameter (B1, B2, H1, H2, IB, and UH) bears an intrinsic error; however, the largest uncertainty is associated with UH, and because this parameter was varied over the entire range of possible clinically relevant values (0.4–1.0), a formal error analysis of the other individual parameters was not believed to be necessary to answer the central question posed by the current investigation. In other words, because the parameter with the largest associated estimation error was varied over the full range of relevant values (and because the results were positive and significant over this range), the results of the current study are interpretable and valid in the absence of a formal error analysis of the other parameters. One additional, small error introduced by the current model is, as described above, the assumption that the effects of radiotherapy and hormone therapy on radiation-related complications are nonsynergistic (i.e., H2 = 0.0; see “Adjustment for Side Effects”). The authors are aware of reports documenting that hormone therapy may surrogate radiation dose47 and that downsizing of the prostate may reduce the complications of external beam therapy.48, 49 Nonetheless, these effects are likely to be very small in absolute magnitude and were not considered in the current investigation.
In addition to the incidence of complications, the severity also requires consideration. Specifically, although the incidence of side effects depends on the duration of hormone use (see Materials and Methods), the severity also often increases with increasing duration. The model is robust enough to account for this fact, as the utility compromise with respect to quality of life is a measure of the severity of the side effects. Thus, for long-term hormone use, UH is likely to be closer to 0.6 (or, possibly, in more severe cases, 0.4), whereas for short-term hormone use, UH is likely to be closer to 0.8. The results demonstrate that over this range of utilities, the observed NNTs are positive and small in magnitude for most of the survival endpoints that were analyzed.
One key feature of the current investigation is that although the randomized studies from which data were taken for input into our model were designed to elucidate the effect of hormones on survival for the overall population of patients with locally advanced prostate carcinoma, the methods used in the current investigation allow for the individualization of management decisions. After a detailed discussion with a given patient about the incidence and severity of the toxicities (or perceived toxicities) associated with short-term and long-term hormone therapy, the physician can estimate UH for that individual patient, enter the value into the developed model for computation of NNT, and use the result of the calculation to assist in deciding whether hormone therapy should be administered, and if so, whether short-term or long-term therapy should be used.
In a similar vein, although it was not explored in detail in the current report, one possible extension of the proposed work would be to determine the optimum duration of hormone treatment. If a constant rate of utility loss as a function of treatment duration is assumed, one can rearrange Equation 2 to determine the incidence of side effects, and thus the duration of hormone therapy that is maximally efficacious while preserving tolerability—i.e., the duration at which the NNT is no longer a small, positive value. To perform this analysis, however, it is necessary to await mature hormone therapy follow-up data for various durations of hormone administration to verify that the rate of utility loss as a function of treatment duration is linear.
The current investigation was possible in large part because of the foundation built by the randomized trials conducted in this treatment population. The model that has been developed will continue to be strengthened by ongoing randomized studies, which will provide increasingly accurate estimates of B1 and B2 (Eq. 1, 2). It is noteworthy that randomized studies comparing long-term and short-term hormone use (e.g., RTOG 92-029) and those comparing different sequences of radiotherapy and hormone use (e.g., RTOG 94-1350) were not included in the current study, because our model only allowed the use of studies that compared radiotherapy alone with radiotherapy plus hormone therapy. Although the model could be expanded to include individual arms of these other studies, errors and biases would be introduced by doing so. In addition, the specialized procedures required for inclusion of these studies were outside of the scope of the current investigation.
The results of the current investigation can serve as the foundation on which to develop a model for early-stage disease (i.e., T1–2b disease35). Retrospective evidence suggests a benefit associated with hormone use in patients with early-stage disease,51, 52 and because hormones commonly are used in this population, a formal NNT analysis should be undertaken; however, because no randomized data currently are available for this population (the results of RTOG 94-0853 and other such studies are eagerly awaited), specialized techniques that incorporate situations in which the input data are subject to error will be necessary to calculate NNT values.54–56 A similar situation would arise if an analysis of hormone therapy in combination with prostatectomy57–59 were undertaken. Again, because the relevant randomized data currently do not exist (except for lymph node–positive patients60), advanced NNT concepts and techniques would be necessary to adequately explore the effect of hormone therapy in this setting. It should be reinforced that the findings of the current study, which analyzed the use of hormones adjuvant to radiotherapy in patients with locally advanced prostate carcinoma, should not be applied to other populations; similar analyses using data that were relevant to the specific circumstances would be required.
An important finding of the current investigation for the practicing clinician is that compared with short-term hormone use, a longer duration of hormone use leads to improvements in virtually all established survival outcomes, even when adjustments are made for toxicities associated with hormone therapy. Although the preliminary results of RTOG 92-029 suggest that long-term hormone use may provide survival benefits relative to short-term hormone use, it is noteworthy that the current study is the first to suggest that these benefits are likely to be present even when the complications of long-term hormone therapy are considered. This result is most evident in Table 3, which shows that the NNTCL values were uniformly less than the corresponding NNTCS values. The only exception to this finding arose from analysis of the LPFS endpoint; with respect to LPFS, the NNTs for long-term hormone therapy were similar to or slightly greater than the NNTs for short-term hormone therapy. The biologic implication of this finding is that the benefit generated by extending the duration of hormone use probably is due in large part to the reduction of occult systemic disease. The benefits observed at the primary site—reduction in the size of the prostate and possible synergistic effects involving radiotherapy—are likely to be present for short-term therapy as well as for long-term therapy, because hormones and radiotherapy are administered concomitantly in both cases. Extending hormone use for a longer duration improves survival outcomes that are related to the eradication of distant metastases, which is not measured by LPFS.
The authors recognize that although the current results speak strongly for the benefit of hormone therapy based on the model that was constructed using data from randomized trials, radiotherapy has evolved over the past decade to the point where patients treated with radiotherapy today (particularly those treated with intensity-modulated radiotherapy61) receive higher doses with less observed toxicity compared with patients treated in the randomized trials included in the current investigation. The model developed and described in the current report will evolve, as will the corresponding conclusions, as randomized trials involving modern radiotherapy techniques and doses are reported and incorporated into the analysis.
The current analysis is not intended to provide a replacement for Quality-adjusted Time Without Symptoms and Toxicity (Q-TWiST)62 or for other methods that attempt to quantify the quality-of-life compromise associated with treatment—the NNT approach provides additional information that is pertinent to the analysis of the benefits and disadvantages of a given clinical intervention. The current report documents the successful application of the NNT method to a common clinical scenario, and we hope that this method will become more commonly used in the oncology community at large.
- 4Prostate Cancer Trialists' Collaborative Group. Maximum androgen blockade in advanced prostate cancer: an overview of 22 randomised trials with 3283 deaths in 5710 patients. Lancet. 1995; 346: 265–269.
- 9RTOG Protocol 92-02: a Phase III trial of the use of long term total androgen suppression following neoadjuvant hormonal cytoreduction and radiotherapy in locally advanced carcinoma of the prostate. Int J Radiat Oncol Biol Phys. 2000; 48(3 Suppl ): 112., , , et al.
- 11Updated results of the Phase III Radiation Therapy Oncology Group (RTOG) Trial 85-31 evaluating the potential benefit of androgen suppression following standard radiation therapy for unfavorable prognosis carcinoma of the prostate. Int J Radiat Oncol Biol Phys. 2001; 49: 937–946., , , et al.
- 30The number needed to treat (NNT) and its applications in radiation therapy. Cancer Invest. In press., , .
- 31Effect of a short course of neoadjuvant hormonal therapy on the response to subsequent androgen suppression in prostate cancer patients with relapse after radiotherapy: a secondary analysis of the randomized protocol RTOG 86-10. Int J Radiat Oncol Biol Phys. 2002; 54: 1302–1310., , , et al.
- 32Cancer and Leukemia Group B. Phase II trial of androgen suppression for 6 months combined with external beam radiotherapy (EBRT) with a brachytherapy (BT) boost for patients with intermediate risk prostate cancer. (CALGB 99809). Available from URL: http://www.calgb.org [accessed 8 August 2003].
- 35GreeneFL, PageDL, FlemingID, et al., editors. Prostate. In: GreeneFL, PageDL, FlemingID, et al., editors. AJCC cancer staging manual, 6th edition. New York: Springer-Verlag, 2002: 309–316.
- 36National Center for Biotechnology Information. PubMed [database online]. Available from URL: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?db=PubMed [accessed 8 August 2003].
- 53Radiation Therapy Oncology Group. A Phase III trial of the study of endocrine therapy used as a cytoreductive and cytostatic agent prior to radiation therapy in good prognosis locally confined adenocarcinoma of the prostate. (RTOG 94-08 protocol). Available from URL: http://www.rtog.org [accessed 8 August 2003].
- 61Intensity modulated radiotherapy (IMRT) for prostate cancer. Clin Prostate Cancer. In press., , .