Prognosis research strategy (PROGRESS) 4: Stratified medicine researchBMJ 2013; 346 doi: https://doi.org/10.1136/bmj.e5793 (Published 05 February 2013) Cite this as: BMJ 2013;346:e5793
All rapid responses
I agree with the Hingorani et al.’s proposal which is applicable to treatment of drug resistant cancers especially in a subset of patients, starting with discovery of prognostic markers, and on that basis, rational drug design followed by their evaluation in pre-clinical and clinical trials.
Drug resistance is an undesirable occurrence in the treatment of cancer. At the cellular level, it is theorised that a number of properties of such cancer cells, besides their drug resistance pose a major problem - 1) they are capable of early dissemination to other parts of the body, and have invasive and metastatic properties, as evidenced by a study which detected early disseminated cells with stem cell phenotype in bone marrow of breast cancer patients  2) they are capable of self-renewal and cell division.  Therefore, they are known as “cancer stem cells”.
There are two sources from which such cells are believed arise from – 1) from existing stem/progenitor cells 2) acquisition of multiple genetic lesions causing the gain of oncogenic and loss of tumour suppressor functions that result in a terminally differentiated cell reverting to the primitive de-differentiated state through a process known as epithelial to mesenchymal transition (EMT).  In EMT, the cells lose their epithelial property such as cell-to-cell junction and apico-basolateral polarity, and are transformed into invasive mesenchymal cells. These “cancer stem cells” are also known to express ABC family of transporters, which are ATP-dependent drug efflux pumps, which mediate drug resistance. 
A pathway that was implicated in firstly, acquired drug resistance especially to receptor tyrosine kinase (RTK)inhibitors like gefitinib, and secondly, manifestation of stem cell-like properties, is the epidermal growth factor receptor (EGFR) signalling pathway.  Shien and colleagues found that resistant cell-lines of lung cancer demonstrated secondary EGFR-T790M mutation, MET oncogene amplification and EMT features. The resistant cells also possess stem cell-like properties such as aldehyde dehydrogenase isoform 1 overexpression, increase of side population and self-renewal capabilities.  Although RTK inhibitors have been used in the treatment of lung cancers with EGFR mutations, the inevitable outcome is the subsequent acquisition of resistance via different mechanisms among virtually all patients who responded initially. 
Hingorani and colleagues’ paper on prognosis research strategy provided a timely, yet essential discussion on the basis of stratified medicine, especially with regards to identification of factors that determine patient treatment response. This topic is especially relevant in the treatment of patients with lung cancers. Hingorani et al. used the example of how HER-2 status in breast cancer was initially discovered as a prognostic indicator, which translated into molecularly targeted therapy towards the subgroup of patients who are HER-2-positive to illustrate the approach of stratified medicine. A parallel approach occurred in determining suitable lung cancer patients for RTK inhibitors therapy. As activating mutations in the gene that encodes EGFR could confer sensitivity to the inhibitors, patients were thus tested for such mutations to identify those who are more likely to benefit from the agent. 
Two prominent mechanisms that lead to development of resistance to RTK inhibitors is the EGFR-T790M mutation and MET amplification, both of which are also associated with stem cell-like characteristics as discussed earlier.  The EGFR-T790M was thought to be a secondary mutation in acquired resistance, but the same mutation was also detected in 40% of untreated non-small cell lung cancers (NSCLC), which suggests it could also occur as a pre-existing phenomenon.  Majority of EGFR mutations harbour sensitivity to RTK inhibitors which decrease affinity of the tyrosine kinase in binding to ATP. The T790M mutation restores this affinity.
Hence, the question is what prognosis research and in the future, management approaches, can be adopted in the context of treating patients with lung cancers, specifically with regards to detecting and anticipating drug resistance? As Hingorani et al. highlighted in their paper, the use of genomic and transcriptomic (of diseased tissues) approaches to identify possible prognostic factors could uncover new disease mechanisms and others that impact response to treatment. Such approaches allow for the detection of EGFR-T790M and MET amplification.
Hingorani et al. made a good point about the need to identify factors that predict differential treatment response especially in the context of trials where drugs fail in the late stage. Their rationale for identifying such factors is to filter and capture patients who may benefit from treatment. Thus, patients can be checked for pre-existing T790M mutations; those with the mutation will have less favourable response to conventional treatment.
The pathway described by Hingorani et al for HER-2 could also be applied in identifying and anticipating patients who develop resistance to anti-cancer treatment through research on prognostic markers. It is preferable for such prognostic markers to be putative drug targets at the same time, like HER-2. The EGFR-T790M is itself a drug target with the second generation of irreversible EGFR inhibitors such as afatinib proving to have activity against it in in vitro and in vivo models. In phase 2b/3 trials, patients with NSCLC who met the criteria for resistance to RTK inhibitors were randomised to receive oral afatinib plus best supportive care. The result is longer progression free survival than the placebo group. 
Other drug targets include the histone deacetylase (HDAC) enzymes, in which the use of inhibitors against such enzymes can induce the differentiation of mesenchymal-like cancer stem cells which trigger apoptotic responses and make them sensitive to chemotherapy. HDAC inhibitors have entered clinical trials either as single agents or in combination with others. 
Hence, Hingorani et al.’s description of Prognosis Research leading to stratified medical approaches provides a pathway in dealing with the problem of drug resistance in cancers. The starting point is identifying prognostic markers indicative of resistance which should preferably be putative drug targets, with rationally designed drugs binding to the appropriate targets evaluated in pre-clinical and subsequently clinical trials.
1. Balic M, Lin H, Young L, Hawes D, Giuliano A, McNamara G, Datar RH, Cote RJ. Most early disseminated cancer cells detected in bone marrow of breast cancer patients have a putative breast cancer stem cell phenotype. Clin Cancer Res. 2006 Oct 1;12(19):5615-5621.
2. Singh A & Settleman J. EMT, cancer stem cells and drug resistance: an emerging axis of evil in the war on cancer. Oncogene (2010) 29, 4741–4751.
3. Shien K, Toyooka S, Yamamoto H, Soh J, Jida M, Thu KL, Hashida S, Maki Y, Ichihara E, Asano H, Tsukuda K, Takigawa N, Kiura K, Gazdar AF, Lam WL, Miyoshi S. Acquired resistance to EGFR inhibitors is associated with a manifestation of stem cell-like properties in cancer cells. Cancer Res. 2013 May 15;73(10):3051-3061.
4. Landi L & Cappuzo F. Irreversible EGFR-TKIs: dreaming perfection. Transl Lung Cancer Res 2013;2(1):40-49.
5. Miller VA, Hirsh V, Cadranel J, Chen YM, Park K, Kim SW, Zhou C, Su WC, Wang M, Sun Y, Heo DS, Crino L, Tan EH, Chao TY, Shahidi M, Cong XJ, Lorence RM, Yang JC. Afatinib versus placebo for patients with advanced, metastatic non-small-cell lung cancer after failure of erlotinib, gefitinib, or both, and one or two lines of chemotherapy (LUX-Lung 1): a phase 2b/3 randomised trial. Lancet Oncol. 2012 May;13(5):528-538
Competing interests: No competing interests
Stratified medicine is key to improving outcomes for patients. By targeting those patient groups that are expected to benefit most and/or incur the least harm from treatment, significant improvements in patient outcomes can be achieved, and a more efficient allocation of health care budgets can be accomplished. In the fourth paper of the PROGRESS series, Hingorani et al. (1) discuss stratified medicine research. They point out the lack of a systematic framework for guiding research on stratified medicine, and highlight four key areas of research that need to be improved. We were triggered by the opening sentence of the paper, concerning a woman with newly diagnosed breast cancer who is deciding on a course of therapy, guided by her physician. For this and other groups of patients stratified medicine is rapidly becoming an integral part of their care. We would therefore like to draw attention to a fifth key area for research on stratified medicine: the research on clinical communication of evidence from prognostic research. This area is thus far lagging behind.
With the advent of stratified medicine, clinicians and patients increasingly have access to prediction models that provide individualized probabilities of outcomes, such as survival or recurrence risk, given patient and disease characteristics. These models bridge the gap between evidence and clinical practice, informing clinicians about relevant probabilities so that they can discuss the tradeoffs involved in treatment choice with their patients. In an era of patient-centered care and shared decision making, such discussions have become increasingly important. For example, oncologists widely use Adjuvant! Online (2) (www.adjuvantonline.com) or Oncotype Dx (3) in the breast cancer setting to aid their own decision making, but also to communicate survival probabilities and potential benefits of treatment to patients during clinical encounters. (4)
Models like Adjuvant! Online can be valuable tools to help clinicians inform patients about the treatment benefits and involve them in decision making. However, the information that risk prediction models provide is often complex, and transmitting it to patients in ways they can understand it and apply it to their own situation, is not straightforward. The literature on risk communication shows that a significant proportion of patients have difficulty understanding probabilities. (5, 6) Yet, there is little research about the impact of using risk prediction models and different risk formats during clinical consultations with patients. Do patients understand the information or does it confuse them? Does it cause increased anxiety? How can the probabilities from these models best be conveyed? Should the uncertainties surrounding the numbers be communicated?
If we look at Adjuvant! Online as an example, studies found that patients generally could not correctly reproduce the risk estimates provided. Moreover, some patients became confused after viewing a printout of the model. Even highly numerate patients struggled, even though Adjuvant! Online does not only provide numbers, but also depicts the survival and treatment benefit estimates with bar charts. Thus, graphical support does not guarantee success. (4, 5, 7, 8) Some models, such as Adjuvant!, use a continuous scale, others, such as Oncotype Dx, categorize patients into low or high risk. Hingorani et al. advocate the use of continuous test in the evaluation (research) phase, adding that “categorization may then be done after analysis to aid clinical strategies”. Yet, it is unknown whether such categorization actually helps doctors and patients.
Researchers are building increasingly sophisticated models. However, if we want to use these models in the clinical encounter, we need evidence about effective communication along with better prognostic estimates. Therefore, whilst we fully concur with Hingorani and others’ (1) recommendations, we want to underscore the importance of concomitant research on the communication of the results of stratified medicine to patients. Such research will promote communication of the risks from prognostic models to patients and truly help patients decide on a course of therapy.
1. Hingorani A, van der Windt D, Riley R, et al.: Prognosis
research strategy (PROGRESS) 4: Stratified medicine
research. BMJ 346, 2013
2. Ravdin P, Siminoff L, Davis G, et al.: Computer program to
assist in making decisions about adjuvant therapy for
women with early breast cancer. J Clin Oncol 19:980-991,
3. Paik S, Shak S, Tang G, et al.: A multigene assay to
predict recurrence of tamoxifen-treated, node-negative
breast cancer. N Engl J Med 351:2817-2826, 2004
4. Stiggelbout A: Adjuvant! and Other Prediction Models in
the Clinical Encounter with Cancer Patients. Medical
Decision Making 30:422-423, 2010
5. Lipkus I, Peters E, Kimmick G, et al.: Breast cancer
patients' treatment expectations after exposure to the
decision aid program adjuvant online: the influence of
numeracy. Med Decis Making 30:464-473, 2010
6. Gigerenzer G, Galesic M: Why do single event
probabilities confuse patients? BMJ 344, 2012
7. Belkora J, Hutton D, Moore D, et al.: Does Use of the
Adjuvant! Model Influence Use of Adjuvant Therapy
Through Better Risk Communication? J Natl Compr Canc
Netw 9:707-712, 2011
8. Hutton D, Belkora J, Shachter R, et al.: Are patients
getting the "gist" in risk communication? Patient
understanding of prognosis in breast cancer treatment. J
Cancer Educ 24:194-199, 2009
Competing interests: No competing interests
The four-part PROGRESS series, of which the article by Hingorani et al. is the final, represents an ambitious and sensible effort at refining prognosis research strategy. Such an undertaking, however, should carefully evaluate whether all basic assumptions of prognosis research and clinical practice and sound.
The stratified approach to prognosis research described by the authors would seem to direct researchers away from reliance on an assumption that, absent evidence to the contrary, one should assume that a relative risk reduction observed in a clinical trial will apply to all baseline rates. But until there exist far more stratified research than there is now, much decision-making will rely on that assumption. And, like other efforts to provide guidance on research and decision-making, including many of its references, the Hingorani article fails to recognize that the assumption is fundamentally unsound.
Consider a situation where in a clinical trial an intervention is observed to reduce the risk of an adverse outcome from 20% to 10%. The trial itself reveals the 10 percentage point absolute risk reduction that is the crucial consideration for most clinical decision-making. But in situations involving different baseline rates, the standard approach for applying information from the trial to estimate the absolute risk reduction (and corresponding number-needed to treat) would be – again, absent sound evidence of a differential effect as the concept generally is understood – to apply the observed 50% relative risk reduction to other baseline rates. Thus, where the baseline rate is 40%, the estimated absolute risk reduction would be 20 percentage points, reflecting the pattern noted by the authors whereby the constant rate ratio yields a larger absolute reduction for the higher baseline rate.
But in applying the information from the clinical trial to estimate absolute risk reductions involving baseline rates other that in the trial, there is no rational basis for applying the observed 50% reduction from 20% to 10% in the adverse outcome rather than the observed 12.5% increase from 80% to 90% in the opposite, favorable outcome. Applying the 12.5% figure to the favorable outcome in the situation where the baseline adverse outcome rate is 40% would increase the 60% favorable outcome rate (reduce the 40% adverse outcome rate) by 7.5 percentage points. This is a smaller absolute risk reduction than observed with the 20% baseline adverse outcome rate.
How then might one employ the information from the trial to estimate the absolute risk reduction in the case of the 40% baseline adverse outcome rate. The most defensible course would be, while ignoring any ratio relationship, to derive from the 20% control and 10% treated adverse outcome rates in the trial, or the corresponding 80% and 90% favorable outcome rates, the difference between the means of hypothesized normal underlying risk distributions. Either approach yields a figure of .44 standard deviations. Based on the assumption that the intervention shifts underlying distributions by .44 standard deviations, one can estimate that in the case of a baseline adverse outcome rate of 40%, the absolute risk reduction would be about 15.6 percentage points. Tables 3 and 4 of reference 2 provide illustrations of the results of such approach compared with those yielded by the assumption of a constant relative risk reduction for the adverse outcome, constant relative risk increase for the favorable, or constant odds ratio for either outcome.
This approach reflects assumptions that the underlying risk distributions are normal and that an intervention shifts these distributions by the same distance on the x-axis. Both assumptions will typically be unproven. But, unlike assumptions as to the constancy of either rate ratio, these assumptions are not illogical. And, so far as I have been able to determine, there exists no sounder basis for applying risk changes observed in a trial to estimate absolute risk changes involving baseline rates other than that in the trial.
These same considerations undermine the standard approach to subgroup analysis, which regards a departure from a constant risk ratio as a differential effect/subgroup effect/interaction. As I explained here most recently in a comment on Hingorani’s reference 49, and as is implicit in the above discussion, since a factor cannot cause equal proportionate changes to different baseline rates of experiencing an outcome while causing equal proportionate changes to the corresponding rates of experiencing the opposite outcome, it is illogical to regard it as somehow normal that a factor should cause equal proportionate changes in either outcome. Conversely, anytime one observes that a factor causes equal changes in different baseline rates for some outcome (and hence finds no evidence of interaction as to that outcome as the concept is generally understood), one will necessarily find evidence of interaction as to the opposite outcome. More broadly, the rate ratio is an unsound measure of association and prognosis research would do well to cease to employ it at all.
As noted at the outset, the Hingorani paper directs prognosis away from reliance on the assumption of a constant relative risk reduction across different baseline rates. But one can make better progress in that undertaking, as well as better realize its urgency, with a recognition of how unsound that assumption is.
1. Hingorani AD, van der Windt DA, Riley RD, et al. Prognosis research strategy (PROGRESS) 4: Stratified medicine research. BMJ 2013;346:e5793
2. Scanlan JP. Subgroup effects [Internet]. 2012. [updated 2012 Feb 24; cited 2013 Feb 24] Available from: http://www.jpscanlan.com/scanlansrule/subgroupeffects.html
3. Scanlan JP. The inevitability of interaction. BMJ Dec. 19, 2011 ((responding to Altman DG, Bland JM. Interaction revisited: the difference between two estimates. BMJ 2003;326:219): http://www.bmj.com/content/326/7382/219?tab=responses
Competing interests: No competing interests