Critique of Hamdy FC et al. N Engl J Med 2016; 375:1415-1424
Hamdy FC et al. 10-year outcomes after monitoring, surgery or radiotherapy for localised prostate cancer. N Engl J Med 2016; 375:1415-1424.
An assignment by: Darren Wogman MSc. Completed as part of Pharmaceutical Medicine MSc at King's College London
1. Describe the paper, its context, and comment on the value of the research, focusing on the statistical and data aspects.
Prostate cancer accounts for the most diagnoses of cancer in men from developed countries (Siegel et al., 2014) and represents the 2nd most common cause of cancer deaths in men in the UK (Jemal et al., 2010). Despite this, clear uncertainties surround the available treatment options.
Hamdy et al (2016) completed the largest randomised study to compare the efficacy of prostate cancer treatments in patients with localised prostate cancer over a 10-year follow-up period. The primary efficacy endpoint was reduction in mortality and secondary endpoints were disease progression and rate of metastasis. There appears to be a further aim to assess the cost-effectiveness of these treatments, but this is not made clear in the final report despite appearing in a preliminary paper on the study design (Lane et al., 2014).
The significance of this paper is two-fold. Firstly, it was the first study of its kind to effectively randomise patients to several therapies or observation. Secondly, it was the first paper to compare between different treatment options instead of a direct comparison of one specific intervention against monitoring in prostate cancer. There appears to be a shortcoming in collection of patient-reported outcomes for treatments that this trial addresses and in doing so, provides a useful perspective into the patient-related effects of treatment and monitoring. This paper sits alongside a sea of similar studies such as, PIVOT trial, comparing surgery with observation (Wilt et al., 2012), SPCG-4 study comparing surgery with “watchful waiting” (Bill-Axelson et al., 2014), ERSPC looking at the effect of screening men with PSA and outcomes of mortality (Schröder et al., 2009) and PLCO the trial, to determine if screening by PSA levels and physical examination can reduce mortality (Andriole et al., 2009).
The study finds “no significant difference among treatments” regarding disease-related mortality (P=0.48) as statistically significant differences between treatments was not found, it is not possible to directly compare across pairs of interventions. Prostate cancer progression and metastases improvements in the active arms are shown (P=<0.001 and P=0.004 respectively). Sensitivity analysis are reported to be close to those in the primary analysis which provides a level of robustness to the calculations and findings.
The paper concludes that surgery and radiation therapies are equally efficacious over a 10-year period in the reduction of mortality. However, for metastatic disease progression, the study justifies active treatment based on the significant reduction observed compared to monitoring. While side-effects are mentioned, the study’s support of active intervention does not adequately balance the highly likely, adverse effects from these treatments which have considerable impacts on quality of life. Urinary incontinence, bowel problems and erectile dysfunction for patients in the active groups compared to the monitoring cohort were greater than the increase in metastases. Furthermore, active monitoring in patients with low-grade prostate cancer is generally supported by the literature covering similar studies (Albertsen, 2005). Having said that, the rate of metastases is likely to increase over time, as seen in Wilt et al., (2017). As such, metastases seen in the monitoring group could exceed the difference in the rate of side-effects and this is likely to provide a decision-point where clinical and patients may differ in their treatment preferences.
2. Give a critique of the methodological steps used by the researchers for the study..Discuss specifically any methodological limitations, and how these limitations impact on the conclusions of the study (40%). Indicate, where relevant, how these limitations could have been better managed (10%).
The trial tested for equivalence in the active treatments and found no statistical significance between them. As such, individual treatments cannot be compared. In any case, low-risk cancer will have a minimal effect on survival within the first 10 years following diagnosis (Albertsen, 2005) and is likely to make it difficult to determine superiority for any of the active treatments. However, it addressed a clearly focussed issue, patients were randomised (although not blinded). Groups at the start were all similar and accounted for at the end. It is possible that bias was introduced as a result of lack of blinding but, this would have been impossible to control for given the treatments being provided.
Only 1% of the patients are black compared to 14% in a similar American study (Kelly et al., 2017). This is important as Afro-Caribbean populations are understood to suffer from more aggressive types of prostate cancer compared to a Caucasian population. The ability of the study results to be extrapolated or applied to other patient populations is low. When these data are contrasted against data collected a fatal prostate cancer study from America (Kelly et al., 2017) we see that even though the PSA levels and clinical diagnoses are similar, ~50% of the American trial participants had a Gleason score of 7, contrasted with just 21% in UK trial (Hamdy et al., 2016).
The study shows that PSA testing is more likely to identify patients low-risk disease, compared to patients with higher grades of cancer and perhaps helps explain some of the patient population demographics. As patients must not have previously had a PSA screen, it is possible that those with higher disease grades will have indirectly been selected out as they are likely to have reported prostate cancer symptoms to their clinicians and be ineligible for the study.
Nearly a third of patients who received surgery were found to have pT3 disease highlighting clinical judgements made before commencement of the trial were not accurate. Despite similar findings in other prostate cancer studies (Coughlin et al., 2018, Pearce et al., 2016, Gardiner et al., 2012) and raises questions over the assumptions made in the statistical analysis plan. The low number of patients with an intermediate-risk in each active group (108-102), means the trial is underpowered to show a treatment difference outcome. The planned power calculation assumed a disease-mortality rate of 10% and a mortality difference of 4.6% (Lane et al., 2014), whilst the observed disease-mortality rate was only 1%. As a result, any inference that surgical or radiotherapy interventions are preferable over monitoring cannot be made (for mortality). The lack of relationship between disease severity and mortality is surprising, it seems reasonable that more severe cases of cancer would result in higher mortality (Kishan et al., 2018) however, this was not reflected here, possibly due to the small number of events.
The results were generated from an intention-to-treat (ITT) analysis and calculations were carried out on the randomised groups, not the treatments received. This provides a serious integrity issue as there was a considerably high proportion of patients that crossed-over into treatment groups they were not assigned to. Over half of the patients in the monitoring arm migrated over into active treatments. This is considerably higher than other studies such as PIVOT, where this proportion is 10% (Wilt et al., 2017). No explanation for treatment changes was provided and raises questions about what motivated these.
Treatment failure was observed in a surprisingly large proportion of those assigned to radiation therapy compared to the surgery. Design shortcomings in the monitoring group must also be highlighted as these patients should at the very least, have been receiving periodic biopsies as per guidelines (Morash et al., 2015).
The length of follow-up is a key feature of this trial, and also a drawback. At trial design, guidelines surrounding observation of patients with high-risk cancer were less clear and would not be best-practise today. The proportion of high-risk individuals in this group will have influenced a conclusion that therefore is less relevant today. Due to improvements in technology notably, multiparametric MRI imaging, improved baseline stratification could have been carried out compared to biopsy (Langer et al., 2009).
The proportion of patients in the monitoring group with metastatic disease was considerably higher than the treatment groups. As metastatic disease is associated with higher mortality, death rates between active treatment and monitoring will diverge over time and provides a compelling argument for further extended follow-up of patient outcomes. Notwithstanding, any active treatment in unlikely to improve outcomes in the patient population within 10 years, especially for those who are considered low-risk where disease history could be upwards of 20 years or more (Johansson, 2004).
3. What original clinical study would you conduct to further investigate the author’s findings, or some of the findings identified? Describe the statistical aspects of the study, its objectives, its design, the data collection and data management (25%)
While it might seem attractive to complete a similar study over a longer, 20-year follow-up period, the change in treatment guidelines means active monitoring finds itself with greater acceptance than at the start of the ProtecT trial (NIHR, 2018). As such, the proposed multi-centre, survival-analysis study will aim to determine if either lifestyle change (Ornish et al., 2005) or active monitoring show superiority with regards to disease progression or disease-related mortality in individuals with low-risk prostate cancer.
I feel that the side-effect profile from surgical and radiotherapy interventions are too detrimental to include these treatments in the study of low-risk patients and data from this study will therefore support conclusions from ProtecT, in determining if active monitoring is a viable regimen for men with low-level prostate cancer compared to surgical or radiotherapeutic intervention.
Serum PSA levels will be measured at screening and every quarter in the first year and then every 6 months until the end of the study. If PSA levels increase by half during the assessment year, a disease management review will take place to determine if these patients should move on from the study and receive surgical, brachytherapy or external beam 3D conformal radiotherapeutic treatment.
Patients will be recruited and randomised (1:1) for either active monitoring or lifestyle changes. Allocations will be stratified by site with stochastic minimisation to ensure an even distribution of individuals age, race, family history, baseline PSA levels and Gleason score with the inclusion of multiparametric MRI imaging.
Kaplan-Meier survival curves, with 95% confidence intervals will be produced to compare outcome status for the two study groups regarding disease progression and metastases. Logrank tests and hazard ratios will be calculated in order to inform conclusions drawn from the data and to provide a basis to determine a difference between the arms.
The nominated data manager will be responsible for producing and maintaining the trial master file. Database setup and documentation testing will take place in advance of enrolment. Data entry will be done at each centre, verified and reviewed by the monitor and data manager. All adverse events will be coded and reconciled and data will be continually reviewed by the independent data committee to ensure patient safety is being maintained. The database will be locked at the end of the entire follow-up period. Data extracts will be made every 5 years, following CDISC standards, in order to provide some interim data analysis that may be helpful to the research area.
Andriole, G., Crawford, E., Grubb, R., Buys, S., Chia, D., Church, T., Fouad, M., Gelmann, E., Kvale, P., Reding, D., Weissfeld, J., Yokochi, L., O'Brien, B., Clapp, J., Rathmell, J., Riley, T., Hayes, R., Kramer, B., Izmirlian, G., Miller, A., Pinsky, P., Prorok, P., Gohagan, J. and Berg, C. (2009). Mortality Results from a Randomized Prostate-Cancer Screening Trial. New England Journal of Medicine, 360(13), pp.1310-1319.
Bill-Axelson, A., Holmberg, L., Garmo, H., Rider, J., Taari, K., Busch, C., Nordling, S., Häggman, M., Andersson, S., Spångberg, A., Andrén, O., Palmgren, J., Steineck, G., Adami, H. and Johansson, J. (2014). Radical Prostatectomy or Watchful Waiting in Early Prostate Cancer. New England Journal of Medicine, 370(10), pp.932-942.
Coughlin, G., Yaxley, J., Chambers, S., Occhipinti, S., Samaratunga, H., Zajdlewicz, L., Teloken, P., Dunglison, N., Williams, S., Lavin, M. and Gardiner, R. (2018). Robot-assisted laparoscopic prostatectomy versus open radical retropubic prostatectomy: 24-month outcomes from a randomised controlled study. The Lancet Oncology, 19(8), pp.1051-1060.
Gardiner, R., Yaxley, J., Coughlin, G., Dunglison, N., Occhipinti, S., Younie, S., Carter, R., Williams, S., Medcraft, R., Bennett, N., Lavin, M. and Chambers, S. (2012). A randomised trial of robotic and open prostatectomy in men with localised prostate cancer. BMC Cancer, 12(1).
Hamdy, F., Donovan, J., Lane, J., Mason, M., Metcalfe, C., Holding, P., Davis, M., Peters, T., Turner, E., Martin, R., Oxley, J., Robinson, M., Staffurth, J., Walsh, E., Bollina, P., Catto, J., Doble, A., Doherty, A., Gillatt, D., Kockelbergh, R., Kynaston, H., Paul, A., Powell, P., Prescott, S., Rosario, D., Rowe, E. and Neal, D. (2016). 10-Year Outcomes after Monitoring, Surgery, or Radiotherapy for Localized Prostate Cancer. New England Journal of Medicine, 375(15), pp.1415-1424.
Jemal, A., Center, M., DeSantis, C. and Ward, E. (2010). Global Patterns of Cancer Incidence and Mortality Rates and Trends. Cancer Epidemiology Biomarkers & Prevention, 19(8), pp.1893-1907.
Johansson, J. (2004). Natural History of Early, Localized Prostate Cancer. JAMA, 291(22), p.2713.
Kelly, S., Rosenberg, P., Anderson, W., Andreotti, G., Younes, N., Cleary, S. and Cook, M. (2017). Trends in the Incidence of Fatal Prostate Cancer in the United States by Race. European Urology, 71(2), pp.195-201.
Lane, J., Donovan, J., Davis, M., Walsh, E., Dedman, D., Down, L., Turner, E., Mason, M., Metcalfe, C., Peters, T., Martin, R., Neal, D. and Hamdy, F. (2014). Active monitoring, radical prostatectomy, or radiotherapy for localised prostate cancer: study design and diagnostic and baseline results of the ProtecT randomised phase 3 trial. The Lancet Oncology, 15(10), pp.1109-1118.
Langer, D., van der Kwast, T., Evans, A., Trachtenberg, J., Wilson, B. and Haider, M. (2009). Prostate cancer detection with multi-parametric MRI: Logistic regression analysis of quantitative T2, diffusion-weighted imaging, and dynamic contrast-enhanced MRI. Journal of Magnetic Resonance Imaging, 30(2), pp.327-334.
Morash, C., Tey, R., Agbassi, C., Klotz, L., McGowan, T., Srigley, J. and Evans, A. (2015). Active surveillance for the management of localized prostate cancer: Guideline recommendations. Canadian Urological Association Journal, 9(5-6), p.171.
Ornish, D., Weidner, G., Fair, W., Marlin, R., Pettengill, E., Raisin, C., Dunn-Emke, S., Crutchfield, L., Jacobs, F., Barnard, R., Aronson, W., Mccormac, P., Mcknight, D., Fein, J., Dnistrian, A., Weinstein, J., Ngo, T., Mendell, N. And Carroll, P. (2005). Intensive Lifestyle Changes May Affect the Progression of Prostate Cancer. Journal of Urology, 174(3), pp.1065-1070.
Pearce, S., Pariser, J., Karrison, T., Patel, S. and Eggener, S. (2016). Comparison of Perioperative and Early Oncologic Outcomes between Open and Robotic Assisted Laparoscopic Prostatectomy in a Contemporary Population Based Cohort. Journal of Urology, 196(1), pp.76-81.
Schröder, F., Hugosson, J., Roobol, M., Tammela, T., Ciatto, S., Nelen, V., Kwiatkowski, M., Lujan, M., Lilja, H., Zappa, M., Denis, L., Recker, F., Berenguer, A., Määttänen, L., Bangma, C., Aus, G., Villers, A., Rebillard, X., van der Kwast, T., Blijenberg, B., Moss, S., de Koning, H. and Auvinen, A. (2009). Screening and Prostate-Cancer Mortality in a Randomized European Study. New England Journal of Medicine, 360(13), pp.1320-1328.
Siegel, R., Ma, J., Zou, Z. and Jemal, A. (2014). Cancer statistics, 2014. CA: A Cancer Journal for Clinicians, 64(1), pp.9-29.
Wilt, T., Brawer, M., Jones, K., Barry, M., Aronson, W., Fox, S., Gingrich, J., Wei, J., Gilhooly, P., Grob, B., Nsouli, I., Iyer, P., Cartagena, R., Snider, G., Roehrborn, C., Sharifi, R., Blank, W., Pandya, P., Andriole, G., Culkin, D. and Wheeler, T. (2012). Radical Prostatectomy versus Observation for Localized Prostate Cancer. New England Journal of Medicine, 367(3), pp.203-213.
Wilt, T., Jones, K., Barry, M., Andriole, G., Culkin, D., Wheeler, T., Aronson, W. and Brawer, M. (2017). Prostatectomy versus Observation for Early Prostate Cancer. New England Journal of Medicine, 377(13), pp.1301-1303.
An assignment by: Darren Wogman MSc. Completed as part of Pharmaceutical Medicine MSc at King's College London