A prospective study of nomogram-based adaptation of prostate radiotherapy target volumes

We have demonstrated the feasibility of a risk-adjusted radiotherapy treatment protocol
that adapts target volume delineation based on nomogram estimates of risk of LRS.
This treatment was shown to be technically feasible, clinically practicable, and resulted
in acceptable levels of acute toxicity in line with standard of care.

It is important to appreciate the natural patterns of spread of disease when determining
target volumes to be treated. A rich surgical pathological literature is available
to inform this approach, demonstrating the frequency, and often extent of disease
involvement. For example, the risk of SV involvement in patients with T2 disease has
been described, as has the fact that in 90 % of such cases disease is confined to
the proximal 20 mm of the SV measured along the axis of the structure 4]. It is perhaps noteworthy that in the HNSCC setting, such data regarding pathological
risk of loco-regional involvement is deemed appropriate to allow clinical implementation
without prospective clinical trial validation 26]. Yet in the prostate radiotherapy setting, clinical trials attempting to quantify
the benefit of WPRT continue to be performed (e.g. RTOG 0534 and RTOG 0924). In the
era of improved imaging, integration of new systemic agents, and highly conformal
radiotherapy, it will be challenging for such studies to definitively answer such
questions for all patients, which is part of the reason that most modern protocols
simply mandate the extent of elective volume treatment 27].

The twenty-eight treatment hypofractionated radiotherapy regimen used in this study
was first described by the Cleveland Clinic 28]. This original protocol has been adapted to form the experimental arm in the RTOG
0415 study, a multi-centre phase III randomized controlled trial examining modest
hypofractionation for treatment of favourable risk prostate cancer. Neither of these
treatment regimens included elective WPRT. Two separate groups in the US have published
their experiences administering conventionally fractionated WPRT concurrently or sequentially
with hypofractionated prostate irradiation 16], 29]. Early data regarding biochemical control and toxicity from these four groups have
demonstrated encouraging results with the modestly hypofractionated treatment.

The frequency of grade ?2 acute GU toxicity (74.1 %) observed in this trial was slightly
higher than that recorded by the aforementioned studies of McDonald et al. (52 %)
and Pollack et al. (approximately 56 %) 16], 29]. This difference may be accounted for by the increased dose to the seminal vesicles
(61.6 Gy vs. 56 Gy or 50 Gy respectively) or more likely, a lower threshold for the
use of interventions. The increase in toxicity was limited in severity to RTOG grade
2, and it remains to be seen whether this will translate into more meaningful differences
in late toxicity. The absence of grade 3 acute toxicity in this study is reassuring
and consistent with the published data using similar treatments. The incidence of
grade ?2 acute GI toxicity (22.2 %) was in the same range as the levels seen in the
University of Alabama at Birmingham series (37 %) 16]. Their series treated all HRPC men with the same radiotherapy doses to the primary
disease and pelvic lymph nodes as in our cohort, and have reported efficacy and late
toxicity rates similar to conventional treatment. Our data adds to the literature
that supports the feasibility of moderately hypofractionated radiotherapy treating
the prostate and pelvis concurrently for men with HRPC.

The question remains as to how best to select patients for radiotherapy volume adaptation.
Some guidelines such as from the EORTC recommend using the D’Amico risk stratification.
This would probably lead to overtreatment, as some patients designated as high risk
actually have very favourable outcomes, illustrating the heterogeneity of such risk
groupings 30]. Clinical tools such as the ‘Partin tables’ 31] have analysed historical data from large cohorts of patients undergoing radical prostatectomy
to demonstrate the correlation between LRS and prognostic factors such as PSA, GS,
and clinical staging. This data could provide an individualized estimate for risk
and degree of LRS, which may then be used to adapt the extent of treatment. The use
of a web-based nomogram (such as the MSKCC nomogram) allows further refinements to
this approach. The clinical tool is widely accessible, simple to use, considers the
additional variable of tumour volume, and considers relevant prognostic factors as
continuous rather than discrete variables. Furthermore, as the calculations are not
completed manually, the underlying algorithm can be sufficiently complex to achieve
maximal accuracy. For these reasons, a computer-based nomogram is a powerful tool
that facilitates risk-adapted treatment individualization.

There are, however, a number of limitations in using a nomogram in this fashion. First
of all, the nomogram is dependent upon historical data that may not be suitable for
extrapolation to the current population. Changes in disease epidemiology, staging,
and screening practices mean that the effect of prognostic factors may differ between
contemporary and historical populations, and the estimates may therefore be inaccurate.
A key example of this was the upward migration of Gleason scoring in recent years,
partly due to the altered definitions of the core biopsy grading system introduced
in 2005 32]. There is therefore a need to regularly review the applicability of historical results
to current populations and update the nomogram algorithms accordingly (which then
also prompts the need for external re-validation). The degree to which this affects
results is illustrated in the difference in nomogram outputs between the time of planning
and analysis (Table 4).

Secondly, most clinical tools used to estimate the risk of LRS in prostate cancer
are based upon radical prostatectomy series that employed limited lymph node dissection.
It has been demonstrated that standard/limited pelvic lymph node dissections may result
in false negative rates for pathological involved nodes of over 50 % compared to extended
dissections 33]. Nomogram algorithms that have been derived from this data may therefore generate
estimates of LNI that are deceptively low.

Thirdly, there is a danger that data entry errors may result in grossly inaccurate
estimates of LRS and incorrect clinical decision-making. For example, a misplaced
decimal point, or inputting the post-ADT PSA rather than PSA at diagnosis may alter
the nomogram estimates considerably. The latter error occurred twice in our study
and resulted in artificially low estimates of LRS and incorrect non-treatment of seminal
vesicles and pelvic nodes in these patients. Simple safeguards would prevent such
errors from occurring, for example, an observer to verify correct data entry and nomogram
use.

Fourthly, there are small sub-groups of prostate cancer patients who are not suitable
for nomogram-directed adaptation of treatment. Outcomes for PSA-negative tumours for
example are not correctly predicted with current nomograms. This group however represents
only a very select subset of patients (1 % or less of total prostate cancer cases)
who very often present late with metastatic disease that is not suitable for curative
treatment 34]. Neuroendocrine carcinoma of the prostate is another group for which standard prognostic
tools are similarly unsuitable.

A further lesson from our experience was appreciating the danger in over-complicating
treatment. The novel treatment regimen used in this study involves a number of features
that increase its complexity compared to standard practice. These include the use
of a nomogram to define risk of LRS and adapt target delineation, protocolised generation
of multiple target structures to be treated using up to three different dose levels,
and a hypofractionated regimen with many unfamiliar dose constraints. Added complexity
is liable to increase the likelihood of errors and protocol non-compliance and must
be justified with a benefit to clinical outcomes. We identified a number of examples
of this, including rotation of the simulation CT images to contour the proximal seminal
vesicles along their axes, or the use of multiple, redundant dose constraints for
rectum and bladder. Here, excessive and unfamiliar processes are unlikely to improve
outcomes and should be simplified. If additional complexity is value-adding, it may
be necessary to implement further safeguards such as peer review of contouring and
the use of checklists to maximise protocol compliance.

We have demonstrated feasibility and deliverability of a complex risk-adapted treatment
for patients with HRPC. Many future directions are being pursued along similar lines.
The use of more extreme hypofractionation coupled with pelvic radiotherapy is increasing,
for example in the ‘SATURN’ trial, in which stereotactic radiotherapy treatment is
administered over 5 fractions to both the prostate and pelvic lymph nodes. A similar
protocol used in the earlier ‘FASTR’ trial however resulted in unacceptable levels
of late toxicity, suggesting caution in using such an approach 35].

In contrast, emerging imaging modalities such as PSMA PET are likely to detect early
metastatic spread with increased sensitivity, which may reduce the number of at-risk
patients with negative staging investigations who are therefore candidates for elective
loco-regional treatment. If this does eventuate, however, we would then face the question
of how to treat this growing group of patients with early loco-regional or oligometastatic
disease, an area where there is again a paucity of evidence to guide management. It
is likely that the management of prostate cancer will shift further towards a risk-adapted
approach as the results of current trials and integration of new imaging into clinical
practice continues.