- Email: [email protected]
Precision medicine: lessons learned from the SHIVA trial – Authors' reply
Correspondence
question raised by the SHIVA trial is how do oncologists interpret and prioritise mutations identified by next generation sequencing that are potentially targets of molecular therapies, since their strategy could substantially affect outcomes of histology-agnostic therapy? In-silico analysis of the missense mutations identified by next generation sequencing has the potential to aid the guidance of histology-agnostic treatment. Insilico analysis algorithms, such as Harvard’s Poly-Phen 2, can be done on personal computers for free and can help identify which missense mutations are likely to be pathogenic versus benign.2 In-silico algorithms look at several abnormalities in a mutation to predict pathogenicity, including physiochemical characteristics of wild-type and variant aminoacids, evolutionary conservation at the location of a mutation, the evolutionarily tolerated aminoacid range at the mutation, and the protein’s structural considerations.3 Poly-Phen 2 has been validated for predicting pathogenicity in missense mutations but has never been used to interpret next generation sequencing results. We recently described the use of in-silico analysis to guide interpretation of next generation sequencing results as a potential means to identify which missense mutation in an advanced malignant disease should be targeted.4 The SHIVA trial raises important questions on the validity of treatments that are based on the results of next generation sequencing and should temper the present role of next generation sequencing in the treatment of heavily pretreated metastatic disease. The SHIVA trial also raises an important financial question, since commercial next generation sequencing platforms can cost US$6000 and targeted therapies can cost in excess of $8000 per month.5,6 In future studies, insilico analysis of the mutations
identified as potential targets by next generation sequencing could result in more successful application of the histology-agnostic approach and potentially improve outcomes compared with standard of care without further increasing costs. We declare no competing interests.
*Andrew W Hahn, Mike G Martin [email protected] University of Utah, Salt Lake City, UT 84106, USA (AWH); and The West Cancer Center, Memphis, TN, USA (MGM) 1
2
3
4
5
6
Le Tourneau C, Delord JP, Goncalves A, et al. Molecularly targeted therapy based on tumour molecular profiling versus conventional therapy for advanced cancer (SHIVA): a multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. Lancet Oncol 2015; 16: 1324–34. Adzhubei IA, Schmidt S, Peshkin L, et al. A method and server for predicting damaging missense mutations. Nat Methods 2010; 7: 248–49. Chan PA, Duraisamy S, Miller PJ, et al. Interpreting missense variants: comparing computational methods in human disease genes CDKN2A, MLH1, MSH2, MECP2, and tyrosinase (TYR). Hum Mutat 2007; 28: 683–93. Hahn AW, Giri S, Patel D, Sluder H, Vanderwalde A, Martin MG. Next-generation sequencing and in silico analysis facilitate prolonged response to pazopanib in a patient with metastatic urothelial carcinoma of the renal pelvis. J Natl Compr Canc Netw 2015; 13: 1181–85. Gallego CJ, Shirts BH, Bennette CS, et al. Next-generation sequencing panels for the diagnosis of colorectal cancer and polyposis syndromes: a cost-effectiveness analysis. J Clin Oncol 2015; 33: 2084–91. Delea TE, Amdahl J, Diaz J, Nakhaipour HR, Hackshaw MD. Cost-effectiveness of pazopanib versus sunitinib for renal cancer in the United States. J Manag Care Spec Pharm 2015; 21: 46–54.
Authors’ reply We thank all correspondents for their constructive overview on the SHIVA trial results.1 Most of the limitations they mention were discussed in the Article. For the major histological cancer subtypes, the median numbers of prior treatments were five for breast cancer (range one to 15), five for ovarian cancer (range one to nine), three for lung cancer (range one to seven), two for cervical cancer (range one to five), and 3·5 for colorectal cancer (range one to 12). Median turnaround time from taking of biopsy
www.thelancet.com/oncology Vol 16 December 2015
to release of molecular profile was 26 days (range 14–42).2 The treatment algorithm, the performance of which is being assessed, is the core of SHIVA. The algorithm is composed of several features clearly stated in the protocol, including the technology and thresholds used to assess molecular alterations of interest, drugs used, the molecular alterations or drugs matching, and the prioritisation of molecular alterations or drugs.3 The overall result of SHIVA is negative and emphasises the need to refine the treatment algorithm, taking into account levels of evidence of predictive biomarkers. Use of insilico analysis algorithms, as suggested by Andrew Hahn and Mike Martin, to help prioritise identified mutations would need to be specified as part of the whole treatment algorithm. Apostolia Tsimberidou and Razelle Kurzrock advocate the use of drug combinations when several molecular alterations are identified. The presence of coexisting molecular alterations might indeed account for the lack of activity of targeted drugs. In view of the high incidence of grade 3 and 4 adverse events recorded with single targeted drugs in the SHIVA trial, we agree with Glen Weiss that this strategy might face safety issues. We fully agree with Weiss’ comment underlining the necessity to incorporate predictive biomarkers of efficacy for immunotherapy and chemotherapy in future trials, and to eventually consider stratification based on oligo-mutant versus polymutant tumours. Of the 194 patients randomly assigned in SHIVA with targeted sequencing data, 63 patients had no mutation identified, 66 had one mutation, and 65 had two or more mutations. Treatment effect was not modified by the number of mutations (pinteraction=0·50). Future ancillary studies using exome sequencing on remaining frozen samples are planned to assess the effect of all exonic molecular alterations, including mutationassociated neoantigens, on treatment response. e581
Correspondence
Tsimberidou and Kurzrock mention several studies that lend support to the use of precision medicine, which we believe have methodological issues. First, although the progressionfree survival ratio is an appealing endpoint in trials mixing tumour types, progression on last treatment needs to be assessed according to RECIST criteria in the same way as the experimental treatment.4 Progression-free survival on last prior treatment was not recorded in SHIVA. Results of the progression-free survival ratio—between progressionfree survival of patients on treatment selected on the basis of their molecular profile versus on treatment based on physician’s choice—assessed by RECIST in patients who crossed-over will be reported soon. Second, retrospective studies are hypothesis-generating, but are not able to conclude whether putative biomarkers are predictive of efficacy or only prognostic.5 Third, non-randomised basket studies are parallel phase 2 studies that do not have comparators. Beside objective responses, disease stabilisations can be difficult to interpret.6 Fourth, results from meta-analyses assessing drugs approved by the US Food and Drug Association obviously cannot be extrapolated to studies assessing drugs outside their indications. Finally, none of these studies reported comparative safety data, which are of primary importance. A detailed scrutiny of the tumour molecular profile of the patients in SHIVA is planned and will hopefully improve understanding of patients’ response to targeted drugs. We declare no competing interests.
*Christophe Le Tourneau, Lisa Belin, Xavier Paoletti, Ivan Bièche, Maud Kamal [email protected] Institut Curie, 92210 Saint-Cloud, Paris, France 1
e582
Le Tourneau C, Delord JP, Goncalves A, et al. Molecularly targeted therapy based on tumour molecular profiling versus conventional therapy for advanced cancer (SHIVA): a multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. Lancet Oncol 2015; 16: 1324–34.
2
3
4
5
6
Le Tourneau C, Paoletti X, Servant N, et al. Randomised proof-of-concept phase II trial comparing targeted therapy based on tumour molecular profiling vs conventional therapy in patients with refractory cancer: results of the feasibility part of the SHIVA trial. Br J Cancer 2014; 111: 17–24. Le Tourneau C, Kamal M, Tsimberidou AM, et al. Treatment algorithms based on tumor molecular profiling: the essence of precision medicine trials. J Natl Cancer Inst (in press). Doroshow JH. Selecting systemic cancer therapy one patient at a time: is there a role for molecular profiling of individual patients with advanced solid tumors? J Clin Oncol 2010; 28: 4869–71. Ballman KV. Biomarker: predictive or prognostic? J Clin Oncol 2015; published online Sept 21. DOI:10.1200/JCO.2015.63.3651. Le Tourneau C, Paoletti X, Coquan E, et al. Critical evaluation of disease stabilization as a measure of activity of systemic therapy: lessons from trials with arms in which patients do not receive active treatment. J Clin Oncol 2014; 32: 260–63.
Maintenance treatment in metastatic colorectal cancer Susanna Hegewisch-Becker and colleagu es reported the results of a randomised trial (AIO 0207)1 comparing maintenance strategies in patients without disease progression after first-line induction treatment with oxaliplatin, a fluoropyrimidine, and bevacizumab in patients with metastatic colorectal cancer. The authors concluded that maintenance with a fluoropyrimidine plus bevacizumab might be the preferable option compared with bevacizumab alone or no treatment. As mentioned by the authors, their findings support the results of our phase 3 trial (CAIRO3),2 which showed a significant benefit in the primary endpoint of progression free survival for maintenance treatment with capecitabine plus bevacizumab compared with observation after induction treatment with capecitabine, oxaliplatin, and bevacizumab in patients with metastatic colorectal cancer. However, some important differences exist between these two trials: CAIRO3 was a superiority trial whereas AIO 0207
was a non-inferiority trial; the length of induction treatment in CAIRO3 was 18 weeks compared with 24 weeks in AIO 0207; and CAIRO3 excluded patients intolerant to oxaliplatin during induction treatment whereas AIO 0207 did not. We believe a superiority design is more appropriate for a trial that compares active treatment with observation than a non-inferiority design. The other two differences between the trials are probably the reason for the higher rate of oxaliplatin reintroduction in CAIRO3 compared with AIO 0207 in both the maintenance group with fluoropyrimidine plus bevacizumab (CAIRO3: 47% vs AIO 0207: 19%) and the observation group (CAIRO3: 60% vs AIO 0207: 46%). In CAIRO3, we did not think that using a strategy that included the reintroduction of oxaliplatin in patients who clearly were not able to tolerate it was appropriate. In their Discussion, HegewischBecker and colleagues state that median overall survival was not improved in CAIRO3. The primary endpoint of CAIRO3 was progressionfree survival after reintroduction of induction treatment. We would like to emphasise that trials should not be judged on endpoints that were not part of the primary analysis. Despite this, the median overall survival in CAIRO3, although not significantly longer in patients in the maintenance treatment group compared with those in observation, showed a clinically meaningful absolute benefit for maintenance treatment compared with observation of 3·5 months. Furthermore, the authors of the AIO 0207 study suggest that the slightly longer median overall survival in their study compared with the CAIRO3 study could be attributed to a difference in the use of anti-EGFR antibodies as salvage treatment. We believe that this suggestion is too simplistic, and caution against such cross-study comparison. In CAIRO3, the poor prognostic group of patients with disease progression before
www.thelancet.com/oncology Vol 16 December 2015
question raised by the SHIVA trial is how do oncologists interpret and prioritise mutations identified by next generation sequencing that are potentially targets of molecular therapies, since their strategy could substantially affect outcomes of histology-agnostic therapy? In-silico analysis of the missense mutations identified by next generation sequencing has the potential to aid the guidance of histology-agnostic treatment. Insilico analysis algorithms, such as Harvard’s Poly-Phen 2, can be done on personal computers for free and can help identify which missense mutations are likely to be pathogenic versus benign.2 In-silico algorithms look at several abnormalities in a mutation to predict pathogenicity, including physiochemical characteristics of wild-type and variant aminoacids, evolutionary conservation at the location of a mutation, the evolutionarily tolerated aminoacid range at the mutation, and the protein’s structural considerations.3 Poly-Phen 2 has been validated for predicting pathogenicity in missense mutations but has never been used to interpret next generation sequencing results. We recently described the use of in-silico analysis to guide interpretation of next generation sequencing results as a potential means to identify which missense mutation in an advanced malignant disease should be targeted.4 The SHIVA trial raises important questions on the validity of treatments that are based on the results of next generation sequencing and should temper the present role of next generation sequencing in the treatment of heavily pretreated metastatic disease. The SHIVA trial also raises an important financial question, since commercial next generation sequencing platforms can cost US$6000 and targeted therapies can cost in excess of $8000 per month.5,6 In future studies, insilico analysis of the mutations
identified as potential targets by next generation sequencing could result in more successful application of the histology-agnostic approach and potentially improve outcomes compared with standard of care without further increasing costs. We declare no competing interests.
*Andrew W Hahn, Mike G Martin [email protected] University of Utah, Salt Lake City, UT 84106, USA (AWH); and The West Cancer Center, Memphis, TN, USA (MGM) 1
2
3
4
5
6
Le Tourneau C, Delord JP, Goncalves A, et al. Molecularly targeted therapy based on tumour molecular profiling versus conventional therapy for advanced cancer (SHIVA): a multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. Lancet Oncol 2015; 16: 1324–34. Adzhubei IA, Schmidt S, Peshkin L, et al. A method and server for predicting damaging missense mutations. Nat Methods 2010; 7: 248–49. Chan PA, Duraisamy S, Miller PJ, et al. Interpreting missense variants: comparing computational methods in human disease genes CDKN2A, MLH1, MSH2, MECP2, and tyrosinase (TYR). Hum Mutat 2007; 28: 683–93. Hahn AW, Giri S, Patel D, Sluder H, Vanderwalde A, Martin MG. Next-generation sequencing and in silico analysis facilitate prolonged response to pazopanib in a patient with metastatic urothelial carcinoma of the renal pelvis. J Natl Compr Canc Netw 2015; 13: 1181–85. Gallego CJ, Shirts BH, Bennette CS, et al. Next-generation sequencing panels for the diagnosis of colorectal cancer and polyposis syndromes: a cost-effectiveness analysis. J Clin Oncol 2015; 33: 2084–91. Delea TE, Amdahl J, Diaz J, Nakhaipour HR, Hackshaw MD. Cost-effectiveness of pazopanib versus sunitinib for renal cancer in the United States. J Manag Care Spec Pharm 2015; 21: 46–54.
Authors’ reply We thank all correspondents for their constructive overview on the SHIVA trial results.1 Most of the limitations they mention were discussed in the Article. For the major histological cancer subtypes, the median numbers of prior treatments were five for breast cancer (range one to 15), five for ovarian cancer (range one to nine), three for lung cancer (range one to seven), two for cervical cancer (range one to five), and 3·5 for colorectal cancer (range one to 12). Median turnaround time from taking of biopsy
www.thelancet.com/oncology Vol 16 December 2015
to release of molecular profile was 26 days (range 14–42).2 The treatment algorithm, the performance of which is being assessed, is the core of SHIVA. The algorithm is composed of several features clearly stated in the protocol, including the technology and thresholds used to assess molecular alterations of interest, drugs used, the molecular alterations or drugs matching, and the prioritisation of molecular alterations or drugs.3 The overall result of SHIVA is negative and emphasises the need to refine the treatment algorithm, taking into account levels of evidence of predictive biomarkers. Use of insilico analysis algorithms, as suggested by Andrew Hahn and Mike Martin, to help prioritise identified mutations would need to be specified as part of the whole treatment algorithm. Apostolia Tsimberidou and Razelle Kurzrock advocate the use of drug combinations when several molecular alterations are identified. The presence of coexisting molecular alterations might indeed account for the lack of activity of targeted drugs. In view of the high incidence of grade 3 and 4 adverse events recorded with single targeted drugs in the SHIVA trial, we agree with Glen Weiss that this strategy might face safety issues. We fully agree with Weiss’ comment underlining the necessity to incorporate predictive biomarkers of efficacy for immunotherapy and chemotherapy in future trials, and to eventually consider stratification based on oligo-mutant versus polymutant tumours. Of the 194 patients randomly assigned in SHIVA with targeted sequencing data, 63 patients had no mutation identified, 66 had one mutation, and 65 had two or more mutations. Treatment effect was not modified by the number of mutations (pinteraction=0·50). Future ancillary studies using exome sequencing on remaining frozen samples are planned to assess the effect of all exonic molecular alterations, including mutationassociated neoantigens, on treatment response. e581
Correspondence
Tsimberidou and Kurzrock mention several studies that lend support to the use of precision medicine, which we believe have methodological issues. First, although the progressionfree survival ratio is an appealing endpoint in trials mixing tumour types, progression on last treatment needs to be assessed according to RECIST criteria in the same way as the experimental treatment.4 Progression-free survival on last prior treatment was not recorded in SHIVA. Results of the progression-free survival ratio—between progressionfree survival of patients on treatment selected on the basis of their molecular profile versus on treatment based on physician’s choice—assessed by RECIST in patients who crossed-over will be reported soon. Second, retrospective studies are hypothesis-generating, but are not able to conclude whether putative biomarkers are predictive of efficacy or only prognostic.5 Third, non-randomised basket studies are parallel phase 2 studies that do not have comparators. Beside objective responses, disease stabilisations can be difficult to interpret.6 Fourth, results from meta-analyses assessing drugs approved by the US Food and Drug Association obviously cannot be extrapolated to studies assessing drugs outside their indications. Finally, none of these studies reported comparative safety data, which are of primary importance. A detailed scrutiny of the tumour molecular profile of the patients in SHIVA is planned and will hopefully improve understanding of patients’ response to targeted drugs. We declare no competing interests.
*Christophe Le Tourneau, Lisa Belin, Xavier Paoletti, Ivan Bièche, Maud Kamal [email protected] Institut Curie, 92210 Saint-Cloud, Paris, France 1
e582
Le Tourneau C, Delord JP, Goncalves A, et al. Molecularly targeted therapy based on tumour molecular profiling versus conventional therapy for advanced cancer (SHIVA): a multicentre, open-label, proof-of-concept, randomised, controlled phase 2 trial. Lancet Oncol 2015; 16: 1324–34.
2
3
4
5
6
Le Tourneau C, Paoletti X, Servant N, et al. Randomised proof-of-concept phase II trial comparing targeted therapy based on tumour molecular profiling vs conventional therapy in patients with refractory cancer: results of the feasibility part of the SHIVA trial. Br J Cancer 2014; 111: 17–24. Le Tourneau C, Kamal M, Tsimberidou AM, et al. Treatment algorithms based on tumor molecular profiling: the essence of precision medicine trials. J Natl Cancer Inst (in press). Doroshow JH. Selecting systemic cancer therapy one patient at a time: is there a role for molecular profiling of individual patients with advanced solid tumors? J Clin Oncol 2010; 28: 4869–71. Ballman KV. Biomarker: predictive or prognostic? J Clin Oncol 2015; published online Sept 21. DOI:10.1200/JCO.2015.63.3651. Le Tourneau C, Paoletti X, Coquan E, et al. Critical evaluation of disease stabilization as a measure of activity of systemic therapy: lessons from trials with arms in which patients do not receive active treatment. J Clin Oncol 2014; 32: 260–63.
Maintenance treatment in metastatic colorectal cancer Susanna Hegewisch-Becker and colleagu es reported the results of a randomised trial (AIO 0207)1 comparing maintenance strategies in patients without disease progression after first-line induction treatment with oxaliplatin, a fluoropyrimidine, and bevacizumab in patients with metastatic colorectal cancer. The authors concluded that maintenance with a fluoropyrimidine plus bevacizumab might be the preferable option compared with bevacizumab alone or no treatment. As mentioned by the authors, their findings support the results of our phase 3 trial (CAIRO3),2 which showed a significant benefit in the primary endpoint of progression free survival for maintenance treatment with capecitabine plus bevacizumab compared with observation after induction treatment with capecitabine, oxaliplatin, and bevacizumab in patients with metastatic colorectal cancer. However, some important differences exist between these two trials: CAIRO3 was a superiority trial whereas AIO 0207
was a non-inferiority trial; the length of induction treatment in CAIRO3 was 18 weeks compared with 24 weeks in AIO 0207; and CAIRO3 excluded patients intolerant to oxaliplatin during induction treatment whereas AIO 0207 did not. We believe a superiority design is more appropriate for a trial that compares active treatment with observation than a non-inferiority design. The other two differences between the trials are probably the reason for the higher rate of oxaliplatin reintroduction in CAIRO3 compared with AIO 0207 in both the maintenance group with fluoropyrimidine plus bevacizumab (CAIRO3: 47% vs AIO 0207: 19%) and the observation group (CAIRO3: 60% vs AIO 0207: 46%). In CAIRO3, we did not think that using a strategy that included the reintroduction of oxaliplatin in patients who clearly were not able to tolerate it was appropriate. In their Discussion, HegewischBecker and colleagues state that median overall survival was not improved in CAIRO3. The primary endpoint of CAIRO3 was progressionfree survival after reintroduction of induction treatment. We would like to emphasise that trials should not be judged on endpoints that were not part of the primary analysis. Despite this, the median overall survival in CAIRO3, although not significantly longer in patients in the maintenance treatment group compared with those in observation, showed a clinically meaningful absolute benefit for maintenance treatment compared with observation of 3·5 months. Furthermore, the authors of the AIO 0207 study suggest that the slightly longer median overall survival in their study compared with the CAIRO3 study could be attributed to a difference in the use of anti-EGFR antibodies as salvage treatment. We believe that this suggestion is too simplistic, and caution against such cross-study comparison. In CAIRO3, the poor prognostic group of patients with disease progression before
www.thelancet.com/oncology Vol 16 December 2015