- Email: [email protected]
Reliance on credibility to prioritise interventions can lead to sub-optimal management strategies
Medical Hypotheses 77 (2011) 541–543
Contents lists available at ScienceDirect
Medical Hypotheses journal homepage: www.elsevier.com/locate/mehy
Reliance on credibility to prioritise interventions can lead to sub-optimal management strategies P.J. Elton ⇑, R. Watkin NHS Bury, Public Health Department, 2nd Floor, Maple House, 8 Haymarket St., Bury BL9 0AR, United Kingdom
a r t i c l e
i n f o
Article history: Received 7 February 2011 Accepted 9 June 2011
a b s t r a c t When there is more than one possible intervention, clinicians have to decide which intervention to offer first. This paper hypothesises that where there is more than one intervention for which the evidence indicates there is similar effectiveness that selecting the intervention with the lowest credibility first may lead to optimal long-term results. Ó 2011 Elsevier Ltd. All rights reserved.
Introduction In chronic conditions, the lack of an imperative for the first line treatment to provide the best outcome initially gives an additional opportunity to maximise the optimal long-term outcome. For many chronic diseases, it is recommended that failure to respond to initial treatment may require switching to another drug in the same class or another class [1]. This raises the issue about whether the order of treatment affects the overall success rate to achieve the optimal dynamic treatment regime [2]. For interventions with similar costs and level of adverse events, the natural assumption is for clinicians to initiate treatment with the intervention with which they have most confidence even when there is no clear evidence of superiority over an alternative intervention. However if this initial choice does not result in the optimal dynamic treatment regime, as it ignores the effect of the initial choice on future options, it is a myopic choice [3]. This should effect pathway development when certain interventions are considered more credible by most clinicians rather than when there is a variety of clinician preferences. The majority opinion can be influenced by how well an intervention is promoted by advertising, sponsorship by the pharmaceutical industry and visits by drug representatives. This can, for example, lead to drugs that are out of patent often not being considered as credible, compared to more recently introduced drugs, even when there is no evidence to support this.
Hypothesis Where there is more than one intervention for which the evidence indicates there is similar effectiveness that selecting the ⇑ Corresponding author. Tel.: +44 0161 272 4062; fax: +44 0161 272 4098. E-mail address: [email protected] (P.J. Elton). 0306-9877/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.mehy.2011.06.026
intervention with the lowest credibility first may lead to optimal long-term results.
Evaluation of the hypothesis Novick et al. [4] found that patients switching from olanzapine to another antipsychotic were less likely to respond than patients switching from another antipsychotic to olanzapine (OR: 0.59; 95% CI: 0.40, 0.87). Similarly the CUtLASS study found that the patients starting on a typical antipsychotic had a non-significantly better outcome than patients starting on an atypical antipsychotic. As this was an open (non-blinded) study, the authors considered that one explanation may be that clinicians may have failed to change an atypical in the face of a poor response, whereas they were more willing to change conventionals [5]. Emslie et al. [6] have suggested that since eventual remission of depression is evident within the first 6 weeks in many, suggesting that earlier switching among non-responders, than common clinical practice, could be important. The credibility given to the initial treatment, especially if there is some response which is insufficient, may make switching less likely. There may also be a reluctance to switch nutritional supplements or herbal remedies which many clinicians do not use [7] after using a pharmacological product. In determining the optimal dynamic treatment regime, the factors that need to be taken into account are the effectiveness of each treatment, the likelihood of the clinician changing to the second intervention and the reduction in effectiveness when the second intervention is used. The intervention that is used second may still have considerable value but may be less effective than the first intervention used [8] possibly because these patients who are judged not to have responded sufficiently to the first intervention are more generally resistant to any intervention.
542
P.J. Elton, R. Watkin / Medical Hypotheses 77 (2011) 541–543
Drug A
Drug B
Start
2 months
Start
50%
30%
20%
change to
responded
no change
intervention B
5%
response
response
rate
4 months
33%
42%
change to
responded
no change
intervention A
26.4%
13.2% responded
25%
24% response rate
5.5% response rate
rate
30% responded
1% responded
6% responded
44.2% response
33% responded
2.31% responded
41.31% response Fig. 1. Illustration of model.
Model The effectiveness of the differing orders in which interventions, such as drugs, are prescribed can be compared as follows: Effectiveness of starting with drug A and switching to drug B is:
SA þ C B SB R C þ U A SA R U ; whereas the effectiveness of starting with drug B and switching to drug A is:
SB þ C A SA R C þ U B SB R U ; where SA, proportion of patients on drug A reaching a defined level of effectiveness; SB, proportion of patients on drug B reaching a defined level of effectiveness; CB, proportion changed from drug A to drug B at a defined time; CA, proportion changed to drug B to drug A at a defined time; UA, the proportion on drug A unchanged at a defined time; UB, the proportion on drug A unchanged at a defined time; RC, reduced effectiveness due to being second drug used; RU, reduced effectiveness when maintained on same drug at a defined time. There are a number of assumptions made to illustrate the model although they may not hold true empirically. However these assumptions do not create bias in any one direction. These assumptions include that the proportionate reduced effectiveness for each drug when it is used second rather than first is constant and the reduced effectiveness when a drug is maintained after the review is the same for each drug. In giving an example of how optimal dynamic therapy would affect the long-term outcome, it is important to show the effect even if the clinician’s feelings, that one intervention is somewhat
superior to another, is correct despite the absence of evidence to support that feeling. So in order to illustrate how the order of treatments may be important, an example is given in which drug A is less effective than drug B but starting with drug A is more effective in the longer term. This is because clinicians are more willing to switch from drug A to drug B than vice versa (Fig. 1). In this example:
SA ¼ 0:3 SB ¼ 0:33 C B ¼ 0:5 at two months C A ¼ 0:25 at two months U A ¼ 0:2 U B ¼ 0:42 RC ¼ 0:8 RU ¼ 0:167
Testing the hypothesis The only testing of the hypothesis to date has been serendipitous [4,5]. As such the results cannot be considered reliable. However they do form the basis of the trail design. Studies need to be randomised trial of two interventions which, based on a critical review of the literature, appear to be of equal effectiveness but for which there is good reason to believe have unequal credibility amongst clinicians. The two arms need to start with one intervention in each arm but the clinician allowed to switch to the other intervention when they see fit. The trial must be an open trial as
P.J. Elton, R. Watkin / Medical Hypotheses 77 (2011) 541–543
the clinician needs to know which intervention is being used as their view of its credibility is likely to affect the timing of the switch.
543
smoking aids but it is rarely prescribed for this purpose by clinicians. Conclusion
Discussion This paper shows that it may be important to take into account the level of credibility of an intervention when determining the order of choice of intervention. The need to consider cost in deciding the order of treatment has been previously described, e.g. the use of intravenous iron therapy before erythropoietin in patients with anaemia accompanying chronic renal failure [9]. Similarly the need to consider adverse events has led to recommending a certain order of treatment, e.g. NICE has recommended paracetamol and/or topical NSAIDs should be considered ahead of oral NSAIDs [10]. However the need to consider the level of credibility of interventions has not previously been described. This issue may have an important effect when clinicians are much more likely to switch in one direction than another. For example, a systematic review indicated that serenoa repens was as effective for benign prostatic hyperplasia as finasteride [11]. It may well be that, because of clinicians’ attitude to herbal remedies that it is much more likely for patients to be switched from serenoa repens to finasteride as a result of an inadequate response than vice versa. The principle of selecting the intervention with the lower level of credibility has a superficial resemblance to the fail first [12] principle in computer programming that states that: ‘‘to succeed, try first where you are most likely to fail’’. In relation to credibility, the principle can be stated as to succeed, try first where you feel you are most likely to fail although there is no evidence to support that feeling. The implications of this model is that when the evidence indicates that two drugs are of comparable effectiveness, level of adverse events and cost is not a significantly different, that the optimal dynamic treatment regime may be to initiate treatment with the drug that has less clinical credibility. Designers of clinical pathways should bear this in mind when making their decisions. This may lead to more drugs out of patent and herbal remedies being used. This paper describes a theoretical situation which may rarely occur. One of the reasons for this is the ability to charge more for drugs with higher credibility. A recent example is the use of ranibizumab rather than bevacizumab for wet age-related macular degeneration [13] because only the former was licensed. Both drugs were made by the same manufacturer but they only sought a license for the much more expensive product. NICE has lent credibility to one drug despite the most recent review of the evidence of the use bevacizumab for wet age-related macular degeneration concluding that it is likely that a randomised controlled trial will show that bevacizumab is equivalent in effect to ranibizumab [14]. Even when drugs are licensed, there tends to be little promotion of drugs that are out of patent. The example of typical antipsychotic drugs being given preference over all typical antipsychotic drugs despite a lack of evidence of superiority has led to a major cost pressure in the NHS for no health gain [5,15]. Similarly Joint British Societies have endorsed the use of nortriptyline [16] to help people give up smoking based on a Cochrane review [17] showing that it was effective as other, more expensive,
Patients may benefit if pathways recommend clinicians initiate treatment with an intervention with the lowest credibility provided there is evidence to indicate that there is equivalence between interventions in effectiveness, adverse events and cost. As higher credibility can allow more to be charged for that intervention, starting pathways with the intervention of lower credibility will often increase cost-effectiveness. Conflicts of interest statement There are no conflicts of interest for either author. Acknowledgements We would like to thank Prof S Lewis, Prof G Dunn and Dr Dyfrig Hughes for their comments on drafts of this paper. References [1] British National Formulary. London (UK): BMJ Group and RPS Publishing; March 2008. p. 193, 202, 535. [2] Lavori PW, Dawson R. Dynamic treatment regimes: practical design considerations. Clinical Trials 2004;1:9–20. [3] Sutton RS, Baro AG. Reinforcement learning: an introduction. Cambridge, Massuchusetts: The MIT Press; 1998. [4] Novick D, Haro JM, Suarez D, et al. Clinical consequences of switching antipsychotic drugs in outpatients with schizophrenia: 36-month results from the European Schizophrenia outpatient health outcomes study. Int Clin Psychopharmacol 2008;23:203–8. [5] Lewis SW, Davies L, Jones PB, et al. Randomisedcontrolled trials of conventional antipsychotic versus new atypical drugs, and new atypical drugs versus clozapine, in people with schizophrenia responding poorly to, or intolerant of, current drug treatment. Health Technol Assess 2006;10(17):79. [6] Emslie GJ, Mayes T, Porta G, et al. Treatment of resistant depression in adolescents (TORDIA): week 24 outcomes. Am J Psychiatry 2010;167:782–91. [7] McGarry H, Pirotta M, Hegarty K, et al. General practitioners and St. John’s Wort: a question of regulation or knowledge? Complemt Ther Med 2007;14:142–8. [8] Brent D, Emslie G, Clarke G, et al. Switching to another SSRI or to venlafaxine with or without cognitive behavioral therapy for adolescents with SSRIresistant depression: the TORDIA randomized controlled trial. JAMA 2008;299:901–13. Feb. 27. [9] Bhandari S, Brownjohn A, Turney J. Effective utilization of erythropoietin with intravenous iron therapy. J Clin Pharm Therapeut 1998;23:73–8. [10] National Institute for Health and Clinical Excellence. Osteoarthritis: the care and management of osteoarthritis in adults. NICE clinical guideline 59; February 2008. [11] Wilt T, Ishani A, Mac Donald R. Serenoa repens for benign prostatic hyperplasia. Cochrane Database of Systematic Reviews 2002, Issue 3. Art. No.: CD001423. doi: 10.1002/14651858.CD001423. [12] Haralick RM, Elliott GL. Increasing tree search efficiency for constraint satisfaction problems. Artif Intell 1980;14:263–314. [13] NICE technology appraisal guidance 155 Ranibizumab and pegaptanib for the treatment of age-related macular degeneration; 2008. [14] Schouten JS, La Heij EC, Webers CA, et al. A systematic review on the effect of bevacizumab in exudative age-related macular degeneration. Grafes Arch Clin Exp Ophthalmol 2009;247:1–11. [15] Lieberman JA, Stroup TS, McEvoy, et al. Effectiveness of antipsychotic drugs in patients with chronic Schizophrenia. New England J Med 2005;353:1209–23. [16] JBS2: Joint British Societies’ guidelines on prevention of cardiovascular disease in clinical practice. Heart 2005;91(Suppl. V):v17. [17] Hughes J, Stead L, Lancaster T. Antidepressants for smoking cessation (cochrane review). The Cochrane Library, Issue 2. Oxford: Update Software; 2002.
Contents lists available at ScienceDirect
Medical Hypotheses journal homepage: www.elsevier.com/locate/mehy
Reliance on credibility to prioritise interventions can lead to sub-optimal management strategies P.J. Elton ⇑, R. Watkin NHS Bury, Public Health Department, 2nd Floor, Maple House, 8 Haymarket St., Bury BL9 0AR, United Kingdom
a r t i c l e
i n f o
Article history: Received 7 February 2011 Accepted 9 June 2011
a b s t r a c t When there is more than one possible intervention, clinicians have to decide which intervention to offer first. This paper hypothesises that where there is more than one intervention for which the evidence indicates there is similar effectiveness that selecting the intervention with the lowest credibility first may lead to optimal long-term results. Ó 2011 Elsevier Ltd. All rights reserved.
Introduction In chronic conditions, the lack of an imperative for the first line treatment to provide the best outcome initially gives an additional opportunity to maximise the optimal long-term outcome. For many chronic diseases, it is recommended that failure to respond to initial treatment may require switching to another drug in the same class or another class [1]. This raises the issue about whether the order of treatment affects the overall success rate to achieve the optimal dynamic treatment regime [2]. For interventions with similar costs and level of adverse events, the natural assumption is for clinicians to initiate treatment with the intervention with which they have most confidence even when there is no clear evidence of superiority over an alternative intervention. However if this initial choice does not result in the optimal dynamic treatment regime, as it ignores the effect of the initial choice on future options, it is a myopic choice [3]. This should effect pathway development when certain interventions are considered more credible by most clinicians rather than when there is a variety of clinician preferences. The majority opinion can be influenced by how well an intervention is promoted by advertising, sponsorship by the pharmaceutical industry and visits by drug representatives. This can, for example, lead to drugs that are out of patent often not being considered as credible, compared to more recently introduced drugs, even when there is no evidence to support this.
Hypothesis Where there is more than one intervention for which the evidence indicates there is similar effectiveness that selecting the ⇑ Corresponding author. Tel.: +44 0161 272 4062; fax: +44 0161 272 4098. E-mail address: [email protected] (P.J. Elton). 0306-9877/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.mehy.2011.06.026
intervention with the lowest credibility first may lead to optimal long-term results.
Evaluation of the hypothesis Novick et al. [4] found that patients switching from olanzapine to another antipsychotic were less likely to respond than patients switching from another antipsychotic to olanzapine (OR: 0.59; 95% CI: 0.40, 0.87). Similarly the CUtLASS study found that the patients starting on a typical antipsychotic had a non-significantly better outcome than patients starting on an atypical antipsychotic. As this was an open (non-blinded) study, the authors considered that one explanation may be that clinicians may have failed to change an atypical in the face of a poor response, whereas they were more willing to change conventionals [5]. Emslie et al. [6] have suggested that since eventual remission of depression is evident within the first 6 weeks in many, suggesting that earlier switching among non-responders, than common clinical practice, could be important. The credibility given to the initial treatment, especially if there is some response which is insufficient, may make switching less likely. There may also be a reluctance to switch nutritional supplements or herbal remedies which many clinicians do not use [7] after using a pharmacological product. In determining the optimal dynamic treatment regime, the factors that need to be taken into account are the effectiveness of each treatment, the likelihood of the clinician changing to the second intervention and the reduction in effectiveness when the second intervention is used. The intervention that is used second may still have considerable value but may be less effective than the first intervention used [8] possibly because these patients who are judged not to have responded sufficiently to the first intervention are more generally resistant to any intervention.
542
P.J. Elton, R. Watkin / Medical Hypotheses 77 (2011) 541–543
Drug A
Drug B
Start
2 months
Start
50%
30%
20%
change to
responded
no change
intervention B
5%
response
response
rate
4 months
33%
42%
change to
responded
no change
intervention A
26.4%
13.2% responded
25%
24% response rate
5.5% response rate
rate
30% responded
1% responded
6% responded
44.2% response
33% responded
2.31% responded
41.31% response Fig. 1. Illustration of model.
Model The effectiveness of the differing orders in which interventions, such as drugs, are prescribed can be compared as follows: Effectiveness of starting with drug A and switching to drug B is:
SA þ C B SB R C þ U A SA R U ; whereas the effectiveness of starting with drug B and switching to drug A is:
SB þ C A SA R C þ U B SB R U ; where SA, proportion of patients on drug A reaching a defined level of effectiveness; SB, proportion of patients on drug B reaching a defined level of effectiveness; CB, proportion changed from drug A to drug B at a defined time; CA, proportion changed to drug B to drug A at a defined time; UA, the proportion on drug A unchanged at a defined time; UB, the proportion on drug A unchanged at a defined time; RC, reduced effectiveness due to being second drug used; RU, reduced effectiveness when maintained on same drug at a defined time. There are a number of assumptions made to illustrate the model although they may not hold true empirically. However these assumptions do not create bias in any one direction. These assumptions include that the proportionate reduced effectiveness for each drug when it is used second rather than first is constant and the reduced effectiveness when a drug is maintained after the review is the same for each drug. In giving an example of how optimal dynamic therapy would affect the long-term outcome, it is important to show the effect even if the clinician’s feelings, that one intervention is somewhat
superior to another, is correct despite the absence of evidence to support that feeling. So in order to illustrate how the order of treatments may be important, an example is given in which drug A is less effective than drug B but starting with drug A is more effective in the longer term. This is because clinicians are more willing to switch from drug A to drug B than vice versa (Fig. 1). In this example:
SA ¼ 0:3 SB ¼ 0:33 C B ¼ 0:5 at two months C A ¼ 0:25 at two months U A ¼ 0:2 U B ¼ 0:42 RC ¼ 0:8 RU ¼ 0:167
Testing the hypothesis The only testing of the hypothesis to date has been serendipitous [4,5]. As such the results cannot be considered reliable. However they do form the basis of the trail design. Studies need to be randomised trial of two interventions which, based on a critical review of the literature, appear to be of equal effectiveness but for which there is good reason to believe have unequal credibility amongst clinicians. The two arms need to start with one intervention in each arm but the clinician allowed to switch to the other intervention when they see fit. The trial must be an open trial as
P.J. Elton, R. Watkin / Medical Hypotheses 77 (2011) 541–543
the clinician needs to know which intervention is being used as their view of its credibility is likely to affect the timing of the switch.
543
smoking aids but it is rarely prescribed for this purpose by clinicians. Conclusion
Discussion This paper shows that it may be important to take into account the level of credibility of an intervention when determining the order of choice of intervention. The need to consider cost in deciding the order of treatment has been previously described, e.g. the use of intravenous iron therapy before erythropoietin in patients with anaemia accompanying chronic renal failure [9]. Similarly the need to consider adverse events has led to recommending a certain order of treatment, e.g. NICE has recommended paracetamol and/or topical NSAIDs should be considered ahead of oral NSAIDs [10]. However the need to consider the level of credibility of interventions has not previously been described. This issue may have an important effect when clinicians are much more likely to switch in one direction than another. For example, a systematic review indicated that serenoa repens was as effective for benign prostatic hyperplasia as finasteride [11]. It may well be that, because of clinicians’ attitude to herbal remedies that it is much more likely for patients to be switched from serenoa repens to finasteride as a result of an inadequate response than vice versa. The principle of selecting the intervention with the lower level of credibility has a superficial resemblance to the fail first [12] principle in computer programming that states that: ‘‘to succeed, try first where you are most likely to fail’’. In relation to credibility, the principle can be stated as to succeed, try first where you feel you are most likely to fail although there is no evidence to support that feeling. The implications of this model is that when the evidence indicates that two drugs are of comparable effectiveness, level of adverse events and cost is not a significantly different, that the optimal dynamic treatment regime may be to initiate treatment with the drug that has less clinical credibility. Designers of clinical pathways should bear this in mind when making their decisions. This may lead to more drugs out of patent and herbal remedies being used. This paper describes a theoretical situation which may rarely occur. One of the reasons for this is the ability to charge more for drugs with higher credibility. A recent example is the use of ranibizumab rather than bevacizumab for wet age-related macular degeneration [13] because only the former was licensed. Both drugs were made by the same manufacturer but they only sought a license for the much more expensive product. NICE has lent credibility to one drug despite the most recent review of the evidence of the use bevacizumab for wet age-related macular degeneration concluding that it is likely that a randomised controlled trial will show that bevacizumab is equivalent in effect to ranibizumab [14]. Even when drugs are licensed, there tends to be little promotion of drugs that are out of patent. The example of typical antipsychotic drugs being given preference over all typical antipsychotic drugs despite a lack of evidence of superiority has led to a major cost pressure in the NHS for no health gain [5,15]. Similarly Joint British Societies have endorsed the use of nortriptyline [16] to help people give up smoking based on a Cochrane review [17] showing that it was effective as other, more expensive,
Patients may benefit if pathways recommend clinicians initiate treatment with an intervention with the lowest credibility provided there is evidence to indicate that there is equivalence between interventions in effectiveness, adverse events and cost. As higher credibility can allow more to be charged for that intervention, starting pathways with the intervention of lower credibility will often increase cost-effectiveness. Conflicts of interest statement There are no conflicts of interest for either author. Acknowledgements We would like to thank Prof S Lewis, Prof G Dunn and Dr Dyfrig Hughes for their comments on drafts of this paper. References [1] British National Formulary. London (UK): BMJ Group and RPS Publishing; March 2008. p. 193, 202, 535. [2] Lavori PW, Dawson R. Dynamic treatment regimes: practical design considerations. Clinical Trials 2004;1:9–20. [3] Sutton RS, Baro AG. Reinforcement learning: an introduction. Cambridge, Massuchusetts: The MIT Press; 1998. [4] Novick D, Haro JM, Suarez D, et al. Clinical consequences of switching antipsychotic drugs in outpatients with schizophrenia: 36-month results from the European Schizophrenia outpatient health outcomes study. Int Clin Psychopharmacol 2008;23:203–8. [5] Lewis SW, Davies L, Jones PB, et al. Randomisedcontrolled trials of conventional antipsychotic versus new atypical drugs, and new atypical drugs versus clozapine, in people with schizophrenia responding poorly to, or intolerant of, current drug treatment. Health Technol Assess 2006;10(17):79. [6] Emslie GJ, Mayes T, Porta G, et al. Treatment of resistant depression in adolescents (TORDIA): week 24 outcomes. Am J Psychiatry 2010;167:782–91. [7] McGarry H, Pirotta M, Hegarty K, et al. General practitioners and St. John’s Wort: a question of regulation or knowledge? Complemt Ther Med 2007;14:142–8. [8] Brent D, Emslie G, Clarke G, et al. Switching to another SSRI or to venlafaxine with or without cognitive behavioral therapy for adolescents with SSRIresistant depression: the TORDIA randomized controlled trial. JAMA 2008;299:901–13. Feb. 27. [9] Bhandari S, Brownjohn A, Turney J. Effective utilization of erythropoietin with intravenous iron therapy. J Clin Pharm Therapeut 1998;23:73–8. [10] National Institute for Health and Clinical Excellence. Osteoarthritis: the care and management of osteoarthritis in adults. NICE clinical guideline 59; February 2008. [11] Wilt T, Ishani A, Mac Donald R. Serenoa repens for benign prostatic hyperplasia. Cochrane Database of Systematic Reviews 2002, Issue 3. Art. No.: CD001423. doi: 10.1002/14651858.CD001423. [12] Haralick RM, Elliott GL. Increasing tree search efficiency for constraint satisfaction problems. Artif Intell 1980;14:263–314. [13] NICE technology appraisal guidance 155 Ranibizumab and pegaptanib for the treatment of age-related macular degeneration; 2008. [14] Schouten JS, La Heij EC, Webers CA, et al. A systematic review on the effect of bevacizumab in exudative age-related macular degeneration. Grafes Arch Clin Exp Ophthalmol 2009;247:1–11. [15] Lieberman JA, Stroup TS, McEvoy, et al. Effectiveness of antipsychotic drugs in patients with chronic Schizophrenia. New England J Med 2005;353:1209–23. [16] JBS2: Joint British Societies’ guidelines on prevention of cardiovascular disease in clinical practice. Heart 2005;91(Suppl. V):v17. [17] Hughes J, Stead L, Lancaster T. Antidepressants for smoking cessation (cochrane review). The Cochrane Library, Issue 2. Oxford: Update Software; 2002.