Exp. Eye Res. (1990) 50, 62 l-624
Clinical
Trials
with
Aldose
Reductase
Inhibitors
D. STRIBLING ICI Pharmaceuticals,
Alderley
Park, Macclesfield,
U.K.
Following the identification of potent, orally active inhibitors of aldose reductase, the next challenge comes in demonstrating clinical utility in the treatment of diabetic complications. It is not feasible to conduct pivotal studies in all indications within the same trial and neuropathy evolved as the lead indication. There was little precedent for clinical trials in any diabetic complication and one of the challenges has been the derivation of standardized methodology for quantifying changes. In order to establish the value of any intervention in a chronic disease, it is necessary to demonstrate sustained benefit in long-term trials. This has been complicated by the absence of quantitative prospective data on the natural history of the disease process. In well-controlled clinical trials, patients are likely to improve their diabetic control which will have an additional effect over and above the usual placebo effect seen in any trial. In the absence of any standard therapy, the definition of clinical as opposed to statistical significance of any effect needs to be established. In the case of neuropathy this is further complicated by the heterogeneity of the disease and questions over the relevance of small changes in electrophysiology as an index of improvement in histopathology. The majority of the animal studies have concentrated on prevention of changes rather than reversal. However, in diabetics, complications develop slowly and significant tissue damage has occurred before clinical signs can be detected. To date, all clinical studies have concentrated on treatment of complications. The dose of AR1 chosen for such studies has been based on the surrogate of erythrocyte aldose reductase activity confirmed by short-term studies on disease reversal. However, recent studies in both neuropathy and retinopathy have shown that the degree of aldose reductase inhibition required for long-term protection is much greater than that required for a short-term effect. The clinical studies which have been reported to date have not shown convincing proof of efficacy but, against this background, it is doubtful whether the aldose reductase hypothesis has really been tested. Key words: aldose reductase inhibitors : diabetes ; neuropathy : retinopathy ; nephropathy : cataract.
1. Introduction From the earliest appreciation of the possible consequences of excess sorbitol production, Jin Kinoshita
began the search for potent selective inhibitors of aldose reductase to substantiate the role of the polyol pathway in the development of sugar cataract (HaymanandKinoshita, 1965:Kinoshitaet al., 1968: Kinoshita, 1974). As an intracellular enzyme, aldose reductase did not prove to be an easy target but since those early days a number of potent and structurally diverse inhibitors have been evaluated in a wide spectrum of animal models of diabetic complications. The results indicate that excessive flux through the polyol pathway may play a key role in the development of not only increased cataracts in diabetics but also neuropathy, retinopathy, nephropathy and other debilitating sequelae of diabetes (Kador, Robison and Kinoshita, 1985). The challenge then comes in demonstrating clinical utility in a diabetic population. The pursuit of this goal has raised some key issues-some of which might have been predicted on entering any new area of medicine with no established therapies to act as benchmarks but others are unique to the diabetes field. These are discussed in turn.
In theory one could study the effect of an aldose inhibitor
00144835/90/060621+04
Catarnct
The original
work establishing
the role of aldose
reductase in the development of sugar cataract in vitro
and in vivo (Kinoshita,
Merola and Dikmak,
Kinoshita et al., 1962) did not readily translate
1962; to the
major problem affecting diabetics which is the approximate fivefold increase in senile cataract compared to the norm. A further detraction from cataracts as a target indication was the absence of reproducible methods for quantification of early cataracts whilst interruption or reversal of cataracts sufficiently established to affect visual acuity is probably an unrealistic target. With the demonstration that aldose reductase may be responsible for the acceleration of senile changes in the diabetic lens, this indication is being reconsidered and will no doubt benefit from the introduction of new methods of quantification of early cataract such as nuclear magnetic resonance. Neuropathy
2. Target Indication reductase
taneously. However, the development of the various lesions is not sufficiently closely correlated and therefore in initial clinical studies it has been necessary to recruit patients with a specified range of severity of one particular complication. This in turn led to the diffficult decision of which indication to target.
on
all
complications
$03.00/O
simul-
Neuropathy offered a more attractive prospect for evaluating ARIs in that studies of the effects of 0 1990 Academic Press Limited
622
improved diabetic control had shown that increases in nerve conduction velocity could be achieved over 1-2 months. However, the variable extent and distribution of lesions under the broad heading of diabetic polyneuropathy has presented problems. Retinopathy
Diabetic retinopathy is perhaps the hardest complication to model in animals and data illustrating chronic effects of ARIs on a recognizable model of diabetic retinopathy have only recently become available (Kinoshita et al., 1979 ; Robison, Kador and Kinoshita, 1983). A major detractor was the perceived slow evolution of retinopathy. Coupled with the low likelihood of achieving reversal, this introduced the necessity for long trials. Nephropathy
This presents perhaps the least well substantiated target for selective inhibitors of aldose reductase because of the heterogeneity of the enzymes present in the kidney. Evidence is now available from animal models that ARIs can prevent increases in proteinuria and glomerular filtration rate but mixed results have been obtained from measurements of glomerulosclerosis. There is no agreement on the relevance of an effect on early markers of the disease such as microalbuminuria and therefore definitive trials may need to be large and very long to demonstrate an effect on progression of disease to classical endpoints. Reflecting these concerns, the focus of the majority of the clinical trials has been on neuropathy followed by retinopathy with nephropathy and cataract only recently attracting attention. 3. Methods of Measurement Neuropathy
Although neurologists have been assessing the extent of diabetic neuropathy using electrophysiological techniques for many years, problems arose with standardization of methods of measurement. Without previous experience of multicentre trials, neurologists were assessing different nerves for different aspects of function such as conduction velocity or amplitude by a variety of techniques. Although multiple measures were being made, no established method of producing combined variables had been established and validated as a measure of neuropathy or its improvement. Vibratory and thermal thresholds provide objective measures of sensory deficits but standardized methods and age-related normal ranges had not been widely established prior to trials starting with ARIs. No validated scoring scale existed for symptoms of neuropathy which are heterogeneous and range from
D. STRIBLING
hypersensitivity to loss of sensation. Although it is possible to apply other scoring scales to aspects of diabetic neuropathy, no systematic questionnaire or natural history data derived from its use are available even now. A compromise has been to use visual analogue scales to assess the patient’s individual rating vs. the severity at entry into the trial. Inevitably symptom scores are subjective and in a long trial depend on the patient recalling how they felt many months previously. In the absence of a definitive measure of polyneuropathy, recourse has been made to biopsies. However, whilst methods of quantification are established, there is no information on which specific aspect of nerve structure is the most important to affect nor how this correlates to a patient benefit. Objective methods for quantifying autonomic neuropathy affecting the vagal control of the heart have been more standardized together with the development of combined variables but age-related normal ranges were equally lacking. Even when best efforts have been made to address these issues, the variability in the individual measures is still large compared to the improvement anticipated. patients exhibit a spectrum of lesions with varying intensity, and in this situation composite variables run the risk of concealing an effect. Retinopathy
In the case of retinopathy sophisticated methods of scoring colour fundus photographs have been developed with the option of using a central reading facility to standardize quantification provided photographers are trained to a high enough standard. An alternative is to use fluorescein angiography which may be more sensitive and reveals vessel leakage, areas of non-perfusion and microaneurysms. The problems encountered have related more to gaining acceptance of surrogate endpoints of retinopathy such as microaneurysms. Nephropathy
Diabetic nephropathy is most frequently monitored by determining the albumin content in urine. Originally detected by calorimetric reactions, RIA has allowed detection at lower levels, and the phase of microalbuminuria preceding gross proteinuria has been recognized. Although there is a general association between the extent of proteinuria and decline in glomerular filtration (GFR), the real endpoint is GFR since patients require renal transplants or dialysis because of inadequate GFR not excess protein excretion. Unfortunately GFR is affected by both glycaemia and blood pressure and therefore interventions which may sustain GFR in the long term are likely to cause a reduction in GFR in the short term. As for
ALDOSE
REOUCTASE
INHIBITOR
CLINICAL
TRIALS
neuropathy, biopsies are the absolute diagnostic but no data are available to indicate on which structural changes are critical and repeat biopsies on a wide scale are not a practical proposition in an efficacy study. 4. Natural
History
Up-to-date, accurate and reliable natural history data are essential to define the size, duration and endpoints for prevention or intervention studies. In the case of all the diabetic complications, knowledge of the natural history of the disease using the sensitive techniques discussed above is lacking. Information is required from longitudinal studies covering progression from the recognition of the initial lesion of the eventual outcome. Information on progression rates and opportunity for regression are important for identifying a dormant phase or one likely to respond to therapy. The historical data currently available are not representative of present clinical practice, e.g. where BP is controlled tightly, lower protein diets are recommended or clinical practice improved through introduction of HbAlc as an integral monitor of diabetic control. The progression of the disease may also be changed when patients are entered into a controlled clinical trial. Paradoxically, some of the best natural history data are emerging from controlled trials with ARIs. 5. Placebo Effects
It is well established that diabetics improve their diabetic control and observance of diet when closely monitored. Improvement in diabetic control is likely to impact on progression rates of complications producing an apparent placebo effect. In addition, where patients are required to report on their symptoms or perception of vibration and temperature, their expectation in an efficacy trial is that the condition should improve--encouraging a placebo response. It is important therefore to run double-blind, placebocontrolled studies to be able to exlude these effects. It is also important to obtain accurate data on the progression of disease when the diabetic is closely monitored in efficacy studies in order to make realistic size predictions in the design of new studies. 6. Clinical Significance
vs. Statistical Significance
IJsing large trials or sophisticated methods of measurements it is frequently possible to demonstrate a statistically significant effect on a parameter such as nerve conduction velocity which is small compared to the level of deficit existing, e.g. 1 m set-’ vs. a 10 m see-’ deficit compared to a non-diabetic control. This challenges everyone to consider what is a clinically significant change since, in assessing a large number of variables, at least S/100 are likely to be ’ significantly’ changed at P -C 0.05 level. An alternate
approach to dealing with multiple endpoints has been to propose composite variables. Unfortunately composite variables which reduce multiple endpoints to manageable proportions can obscure real effects when the disease is heterogeneous. An alternative situation is to score responders or count numbers of patients passing a clinical milestone. Whilst a statistically significant effect on the numbers of responders is generally regarded as clinically meaningful, the definitions must be well founded on clinical endpoints, not arbitrary changes in one or more parameters, for this to be a valid criterion. 7. Dose Having selected the preferred complication, the size and duration of the trial, the issue is then centred on dose selection. There is no general agreement on the extent of inhibition of sorbitol which would be required for effect. Sorbitol is a component of the target tissues and therefore the concept of returning levels to normal was considered appropriate. Some have believed sorbitol plays a protective role during hyperglycaemia and were in favour of a partial reduction. More customarily drugs aimed at enzyme systems aim for a 9 5 % effect. The aldose reductase hypothesis has focused on sorbitol but other biochemical changes may cause the damage. The dose response for a change in sorbitol may not be the same as that for the key factor thus undermining the selection of the dose. The target tissues are not readily accessible and erythrocyte sorbitol levels have been used as surrogate to predict the state of inhibition in other tissues. There are problems, however. in that the enzyme in the red cell is different from that in the tissues and the cellular levels of drug are not necessarily parallel either in time or concentration profile. An alternative is to use short-term trials in the target indication to predict the dose for long-term effects. Using this approach, however, there is a danger that the dose may be underestimated. For example, partial inhibition of AR in rat lens prevents cataract at 3 months but at later time points ‘breakthrough ’ occurs and higher doses are required to maintain benefit over untreated animals. The same may be true for other complications as suggested by AR1 studies in galactosaemic dogs (Kador at this meeting). 8. Treatment
vs. Prevention
The two extremes of view are present. To justify the cost of treatment, the sceptic likes to see drugs impact on something which patients complain about or which requires greater medical care--e.g. symptoms, need for dialysis, visual acuity. At the other extreme the view is that all diabetics go on to develop some level of complication and if an effect on early markers can be
624 achieved, this will be taken as proof of benefit for patients over the course of the disease. This leads to a debate about the relevance of early markers, e.g. hyperfiltration, microalbuminuria. The alternatives are prevention of any lesion, prevention of deterioration or reversal of disease. In order to demonstate prevention, patients need to be free of any lesion at entry into the study yet many non-insulin dependent diabetics have gross lesions at the time of diagnosis of diabetes. The only treatment population would therefore be newly diagnosed insulin dependent diabetics who will not develop complications until 5 + yr after diagnosis. This time would be considerably extended by closer monitoring of diabetic control in a controlled trial. In this group, progress to hard clinical endpoints takes 15 + yr and therefore acceptance of effects on early markers of the disease as proof of efficacy would be vital. Prevention of deterioration assumes that AR continues to be involved in the continuing damage and this has not been established in animals. Furthermore, the duration of the study is dependent on the rate of progression of the disease which in a controlled trial could be slower than projected. As with the prevention trial, a debate can develop on the magnitude of a clinically significant change unless late stage disease is studied which may well be insensitive to intervention by ARIs. The alternative is to show reversal of the damage. For example signs of nerve regeneration were shown in a sorbinil trial at 12 months but had not reached a level to be functional. Reversal of nephropathy and retinopathy is not likely, since intensified insulin therapy at best slows the progress of the disease. 9. Clinical Effects of ARIs Whilst identifying some of the problems, very few solutions have been recommended. However. in spite of the problems clinical results on ARIs have been generated. Many ARIs have been shown to affect erythrocyte sorbitol levels in diabetics. Some evidence for an effect on nerve sorbitol has been demonstrated but no effect was demonstrated on other biochemical changes thought to be critical to the development of neuropathy. The reduction in sorbitol achieved (40%) may well be below that required for long-term benefit. Effects on nerve conduction velocity in the region of 2 m see-’ and small effects on evoked potential have been shown in trials up to 12 months in duration. Vagal activity has been shown to be improved at 6 months and claims have been made for a benefit on symptoms of neuropathy. In the case of nephropathy, ARI’s have been
D. STRIBLING
reported to reverse and prevent increases in microalbuminuria in studies up to 1 yr in duration. Hyperfiltration associated with early incipient nephropathy can be reversed but no long-term studies have been completed showing an effect on declining GFR and there is no general agreement that improvement in these early markers equates with a benefit on clinical nephropathy. In retinopathy, a reduction in fluorescein leakage at 6 months has been claimed. Otherwise no doubleblind placebo-controlled trials of ARIs in retinopathy have been reported although several are in progress. As far as efficacy is concerned, therefore, short-term effects on early neuropathy and nephropathy have been achieved but have uncertain clinical significance. Looking back 5 yr, there was concern that there were class effects from inhibiting ARI-such as effects on liver function and skin rashes-but with further ARI’s being studied there is no evidence that this is the case. In a chronic disease which develops over 29 yr or more, unless the drug achieves a short-term improvement in symptoms, there will always be arguments over the significance of changes in early markers of disease in ‘short-term’ trials. Unfortunately, it may be that registration of aldose reductase inhibitors could depend on long-term studies to demonstrate an effect on preventing changes over several years treatment. References Hayman. S. and Kinoshita. J. H. (1965). Isolation and properties of lens aldose reductase. I. Biol. Chem. 240. 877-82. Kador,P. F.. Robison,W. G.and Kinoshita,J. H. ( 1985). The pharmacologyof aldosereductaseinhibitors. Annu. Rev. Pharmacol. Toxicol. 25, 691-714.
Kinoshita, J. H. (19 74). Mechanismsinhibiting cataract formation. Invest. Ophthalmol. 13, 713-24. Kinoshita, J. H., Dvornik, D, Kraml. M. and Gabbay, K. H. (1968). The effect of an aldosereductaseinhibitor on the galactoseexposedlens.Biochim. Biophys. Acta 158. 472-5. Kinoshita.J. H.. Fukushi. S., Kador, P. F. and Merola. L. 0. (1979). Aldose reductasein diabetic complicationsof the eye. Metabolism 28 (Suppi. l), 462-9. Kinoshita, J. H.. Merola, L. 0. and Dikmak, E. (1962). Osmoticchangesin experimental galactosecataracts. Ex~. E:fe Res.1. 405-10. Kinoshita, J. H., Merola. L. 0.. Satoh, K. and Dikmak, E. (1962). Osmoticchangescausedby the accumulation of dulcitol in the lensesof rats fed with galactose.Nature 194, 1085-7. Robison,W. G.. Kador. P. F. and Kinoshita, J. H. ( 1983). Retinal capillariesbasementmembranethickening by galactosaemia prevented by an aldose reductase hibitor. Science 221, 1177-9.
in-
Clinical
Trials
with
Aldose
Reductase
Inhibitors
D. STRIBLING ICI Pharmaceuticals,
Alderley
Park, Macclesfield,
U.K.
Following the identification of potent, orally active inhibitors of aldose reductase, the next challenge comes in demonstrating clinical utility in the treatment of diabetic complications. It is not feasible to conduct pivotal studies in all indications within the same trial and neuropathy evolved as the lead indication. There was little precedent for clinical trials in any diabetic complication and one of the challenges has been the derivation of standardized methodology for quantifying changes. In order to establish the value of any intervention in a chronic disease, it is necessary to demonstrate sustained benefit in long-term trials. This has been complicated by the absence of quantitative prospective data on the natural history of the disease process. In well-controlled clinical trials, patients are likely to improve their diabetic control which will have an additional effect over and above the usual placebo effect seen in any trial. In the absence of any standard therapy, the definition of clinical as opposed to statistical significance of any effect needs to be established. In the case of neuropathy this is further complicated by the heterogeneity of the disease and questions over the relevance of small changes in electrophysiology as an index of improvement in histopathology. The majority of the animal studies have concentrated on prevention of changes rather than reversal. However, in diabetics, complications develop slowly and significant tissue damage has occurred before clinical signs can be detected. To date, all clinical studies have concentrated on treatment of complications. The dose of AR1 chosen for such studies has been based on the surrogate of erythrocyte aldose reductase activity confirmed by short-term studies on disease reversal. However, recent studies in both neuropathy and retinopathy have shown that the degree of aldose reductase inhibition required for long-term protection is much greater than that required for a short-term effect. The clinical studies which have been reported to date have not shown convincing proof of efficacy but, against this background, it is doubtful whether the aldose reductase hypothesis has really been tested. Key words: aldose reductase inhibitors : diabetes ; neuropathy : retinopathy ; nephropathy : cataract.
1. Introduction From the earliest appreciation of the possible consequences of excess sorbitol production, Jin Kinoshita
began the search for potent selective inhibitors of aldose reductase to substantiate the role of the polyol pathway in the development of sugar cataract (HaymanandKinoshita, 1965:Kinoshitaet al., 1968: Kinoshita, 1974). As an intracellular enzyme, aldose reductase did not prove to be an easy target but since those early days a number of potent and structurally diverse inhibitors have been evaluated in a wide spectrum of animal models of diabetic complications. The results indicate that excessive flux through the polyol pathway may play a key role in the development of not only increased cataracts in diabetics but also neuropathy, retinopathy, nephropathy and other debilitating sequelae of diabetes (Kador, Robison and Kinoshita, 1985). The challenge then comes in demonstrating clinical utility in a diabetic population. The pursuit of this goal has raised some key issues-some of which might have been predicted on entering any new area of medicine with no established therapies to act as benchmarks but others are unique to the diabetes field. These are discussed in turn.
In theory one could study the effect of an aldose inhibitor
00144835/90/060621+04
Catarnct
The original
work establishing
the role of aldose
reductase in the development of sugar cataract in vitro
and in vivo (Kinoshita,
Merola and Dikmak,
Kinoshita et al., 1962) did not readily translate
1962; to the
major problem affecting diabetics which is the approximate fivefold increase in senile cataract compared to the norm. A further detraction from cataracts as a target indication was the absence of reproducible methods for quantification of early cataracts whilst interruption or reversal of cataracts sufficiently established to affect visual acuity is probably an unrealistic target. With the demonstration that aldose reductase may be responsible for the acceleration of senile changes in the diabetic lens, this indication is being reconsidered and will no doubt benefit from the introduction of new methods of quantification of early cataract such as nuclear magnetic resonance. Neuropathy
2. Target Indication reductase
taneously. However, the development of the various lesions is not sufficiently closely correlated and therefore in initial clinical studies it has been necessary to recruit patients with a specified range of severity of one particular complication. This in turn led to the diffficult decision of which indication to target.
on
all
complications
$03.00/O
simul-
Neuropathy offered a more attractive prospect for evaluating ARIs in that studies of the effects of 0 1990 Academic Press Limited
622
improved diabetic control had shown that increases in nerve conduction velocity could be achieved over 1-2 months. However, the variable extent and distribution of lesions under the broad heading of diabetic polyneuropathy has presented problems. Retinopathy
Diabetic retinopathy is perhaps the hardest complication to model in animals and data illustrating chronic effects of ARIs on a recognizable model of diabetic retinopathy have only recently become available (Kinoshita et al., 1979 ; Robison, Kador and Kinoshita, 1983). A major detractor was the perceived slow evolution of retinopathy. Coupled with the low likelihood of achieving reversal, this introduced the necessity for long trials. Nephropathy
This presents perhaps the least well substantiated target for selective inhibitors of aldose reductase because of the heterogeneity of the enzymes present in the kidney. Evidence is now available from animal models that ARIs can prevent increases in proteinuria and glomerular filtration rate but mixed results have been obtained from measurements of glomerulosclerosis. There is no agreement on the relevance of an effect on early markers of the disease such as microalbuminuria and therefore definitive trials may need to be large and very long to demonstrate an effect on progression of disease to classical endpoints. Reflecting these concerns, the focus of the majority of the clinical trials has been on neuropathy followed by retinopathy with nephropathy and cataract only recently attracting attention. 3. Methods of Measurement Neuropathy
Although neurologists have been assessing the extent of diabetic neuropathy using electrophysiological techniques for many years, problems arose with standardization of methods of measurement. Without previous experience of multicentre trials, neurologists were assessing different nerves for different aspects of function such as conduction velocity or amplitude by a variety of techniques. Although multiple measures were being made, no established method of producing combined variables had been established and validated as a measure of neuropathy or its improvement. Vibratory and thermal thresholds provide objective measures of sensory deficits but standardized methods and age-related normal ranges had not been widely established prior to trials starting with ARIs. No validated scoring scale existed for symptoms of neuropathy which are heterogeneous and range from
D. STRIBLING
hypersensitivity to loss of sensation. Although it is possible to apply other scoring scales to aspects of diabetic neuropathy, no systematic questionnaire or natural history data derived from its use are available even now. A compromise has been to use visual analogue scales to assess the patient’s individual rating vs. the severity at entry into the trial. Inevitably symptom scores are subjective and in a long trial depend on the patient recalling how they felt many months previously. In the absence of a definitive measure of polyneuropathy, recourse has been made to biopsies. However, whilst methods of quantification are established, there is no information on which specific aspect of nerve structure is the most important to affect nor how this correlates to a patient benefit. Objective methods for quantifying autonomic neuropathy affecting the vagal control of the heart have been more standardized together with the development of combined variables but age-related normal ranges were equally lacking. Even when best efforts have been made to address these issues, the variability in the individual measures is still large compared to the improvement anticipated. patients exhibit a spectrum of lesions with varying intensity, and in this situation composite variables run the risk of concealing an effect. Retinopathy
In the case of retinopathy sophisticated methods of scoring colour fundus photographs have been developed with the option of using a central reading facility to standardize quantification provided photographers are trained to a high enough standard. An alternative is to use fluorescein angiography which may be more sensitive and reveals vessel leakage, areas of non-perfusion and microaneurysms. The problems encountered have related more to gaining acceptance of surrogate endpoints of retinopathy such as microaneurysms. Nephropathy
Diabetic nephropathy is most frequently monitored by determining the albumin content in urine. Originally detected by calorimetric reactions, RIA has allowed detection at lower levels, and the phase of microalbuminuria preceding gross proteinuria has been recognized. Although there is a general association between the extent of proteinuria and decline in glomerular filtration (GFR), the real endpoint is GFR since patients require renal transplants or dialysis because of inadequate GFR not excess protein excretion. Unfortunately GFR is affected by both glycaemia and blood pressure and therefore interventions which may sustain GFR in the long term are likely to cause a reduction in GFR in the short term. As for
ALDOSE
REOUCTASE
INHIBITOR
CLINICAL
TRIALS
neuropathy, biopsies are the absolute diagnostic but no data are available to indicate on which structural changes are critical and repeat biopsies on a wide scale are not a practical proposition in an efficacy study. 4. Natural
History
Up-to-date, accurate and reliable natural history data are essential to define the size, duration and endpoints for prevention or intervention studies. In the case of all the diabetic complications, knowledge of the natural history of the disease using the sensitive techniques discussed above is lacking. Information is required from longitudinal studies covering progression from the recognition of the initial lesion of the eventual outcome. Information on progression rates and opportunity for regression are important for identifying a dormant phase or one likely to respond to therapy. The historical data currently available are not representative of present clinical practice, e.g. where BP is controlled tightly, lower protein diets are recommended or clinical practice improved through introduction of HbAlc as an integral monitor of diabetic control. The progression of the disease may also be changed when patients are entered into a controlled clinical trial. Paradoxically, some of the best natural history data are emerging from controlled trials with ARIs. 5. Placebo Effects
It is well established that diabetics improve their diabetic control and observance of diet when closely monitored. Improvement in diabetic control is likely to impact on progression rates of complications producing an apparent placebo effect. In addition, where patients are required to report on their symptoms or perception of vibration and temperature, their expectation in an efficacy trial is that the condition should improve--encouraging a placebo response. It is important therefore to run double-blind, placebocontrolled studies to be able to exlude these effects. It is also important to obtain accurate data on the progression of disease when the diabetic is closely monitored in efficacy studies in order to make realistic size predictions in the design of new studies. 6. Clinical Significance
vs. Statistical Significance
IJsing large trials or sophisticated methods of measurements it is frequently possible to demonstrate a statistically significant effect on a parameter such as nerve conduction velocity which is small compared to the level of deficit existing, e.g. 1 m set-’ vs. a 10 m see-’ deficit compared to a non-diabetic control. This challenges everyone to consider what is a clinically significant change since, in assessing a large number of variables, at least S/100 are likely to be ’ significantly’ changed at P -C 0.05 level. An alternate
approach to dealing with multiple endpoints has been to propose composite variables. Unfortunately composite variables which reduce multiple endpoints to manageable proportions can obscure real effects when the disease is heterogeneous. An alternative situation is to score responders or count numbers of patients passing a clinical milestone. Whilst a statistically significant effect on the numbers of responders is generally regarded as clinically meaningful, the definitions must be well founded on clinical endpoints, not arbitrary changes in one or more parameters, for this to be a valid criterion. 7. Dose Having selected the preferred complication, the size and duration of the trial, the issue is then centred on dose selection. There is no general agreement on the extent of inhibition of sorbitol which would be required for effect. Sorbitol is a component of the target tissues and therefore the concept of returning levels to normal was considered appropriate. Some have believed sorbitol plays a protective role during hyperglycaemia and were in favour of a partial reduction. More customarily drugs aimed at enzyme systems aim for a 9 5 % effect. The aldose reductase hypothesis has focused on sorbitol but other biochemical changes may cause the damage. The dose response for a change in sorbitol may not be the same as that for the key factor thus undermining the selection of the dose. The target tissues are not readily accessible and erythrocyte sorbitol levels have been used as surrogate to predict the state of inhibition in other tissues. There are problems, however. in that the enzyme in the red cell is different from that in the tissues and the cellular levels of drug are not necessarily parallel either in time or concentration profile. An alternative is to use short-term trials in the target indication to predict the dose for long-term effects. Using this approach, however, there is a danger that the dose may be underestimated. For example, partial inhibition of AR in rat lens prevents cataract at 3 months but at later time points ‘breakthrough ’ occurs and higher doses are required to maintain benefit over untreated animals. The same may be true for other complications as suggested by AR1 studies in galactosaemic dogs (Kador at this meeting). 8. Treatment
vs. Prevention
The two extremes of view are present. To justify the cost of treatment, the sceptic likes to see drugs impact on something which patients complain about or which requires greater medical care--e.g. symptoms, need for dialysis, visual acuity. At the other extreme the view is that all diabetics go on to develop some level of complication and if an effect on early markers can be
624 achieved, this will be taken as proof of benefit for patients over the course of the disease. This leads to a debate about the relevance of early markers, e.g. hyperfiltration, microalbuminuria. The alternatives are prevention of any lesion, prevention of deterioration or reversal of disease. In order to demonstate prevention, patients need to be free of any lesion at entry into the study yet many non-insulin dependent diabetics have gross lesions at the time of diagnosis of diabetes. The only treatment population would therefore be newly diagnosed insulin dependent diabetics who will not develop complications until 5 + yr after diagnosis. This time would be considerably extended by closer monitoring of diabetic control in a controlled trial. In this group, progress to hard clinical endpoints takes 15 + yr and therefore acceptance of effects on early markers of the disease as proof of efficacy would be vital. Prevention of deterioration assumes that AR continues to be involved in the continuing damage and this has not been established in animals. Furthermore, the duration of the study is dependent on the rate of progression of the disease which in a controlled trial could be slower than projected. As with the prevention trial, a debate can develop on the magnitude of a clinically significant change unless late stage disease is studied which may well be insensitive to intervention by ARIs. The alternative is to show reversal of the damage. For example signs of nerve regeneration were shown in a sorbinil trial at 12 months but had not reached a level to be functional. Reversal of nephropathy and retinopathy is not likely, since intensified insulin therapy at best slows the progress of the disease. 9. Clinical Effects of ARIs Whilst identifying some of the problems, very few solutions have been recommended. However. in spite of the problems clinical results on ARIs have been generated. Many ARIs have been shown to affect erythrocyte sorbitol levels in diabetics. Some evidence for an effect on nerve sorbitol has been demonstrated but no effect was demonstrated on other biochemical changes thought to be critical to the development of neuropathy. The reduction in sorbitol achieved (40%) may well be below that required for long-term benefit. Effects on nerve conduction velocity in the region of 2 m see-’ and small effects on evoked potential have been shown in trials up to 12 months in duration. Vagal activity has been shown to be improved at 6 months and claims have been made for a benefit on symptoms of neuropathy. In the case of nephropathy, ARI’s have been
D. STRIBLING
reported to reverse and prevent increases in microalbuminuria in studies up to 1 yr in duration. Hyperfiltration associated with early incipient nephropathy can be reversed but no long-term studies have been completed showing an effect on declining GFR and there is no general agreement that improvement in these early markers equates with a benefit on clinical nephropathy. In retinopathy, a reduction in fluorescein leakage at 6 months has been claimed. Otherwise no doubleblind placebo-controlled trials of ARIs in retinopathy have been reported although several are in progress. As far as efficacy is concerned, therefore, short-term effects on early neuropathy and nephropathy have been achieved but have uncertain clinical significance. Looking back 5 yr, there was concern that there were class effects from inhibiting ARI-such as effects on liver function and skin rashes-but with further ARI’s being studied there is no evidence that this is the case. In a chronic disease which develops over 29 yr or more, unless the drug achieves a short-term improvement in symptoms, there will always be arguments over the significance of changes in early markers of disease in ‘short-term’ trials. Unfortunately, it may be that registration of aldose reductase inhibitors could depend on long-term studies to demonstrate an effect on preventing changes over several years treatment. References Hayman. S. and Kinoshita. J. H. (1965). Isolation and properties of lens aldose reductase. I. Biol. Chem. 240. 877-82. Kador,P. F.. Robison,W. G.and Kinoshita,J. H. ( 1985). The pharmacologyof aldosereductaseinhibitors. Annu. Rev. Pharmacol. Toxicol. 25, 691-714.
Kinoshita, J. H. (19 74). Mechanismsinhibiting cataract formation. Invest. Ophthalmol. 13, 713-24. Kinoshita, J. H., Dvornik, D, Kraml. M. and Gabbay, K. H. (1968). The effect of an aldosereductaseinhibitor on the galactoseexposedlens.Biochim. Biophys. Acta 158. 472-5. Kinoshita.J. H.. Fukushi. S., Kador, P. F. and Merola. L. 0. (1979). Aldose reductasein diabetic complicationsof the eye. Metabolism 28 (Suppi. l), 462-9. Kinoshita, J. H.. Merola, L. 0. and Dikmak, E. (1962). Osmoticchangesin experimental galactosecataracts. Ex~. E:fe Res.1. 405-10. Kinoshita, J. H., Merola. L. 0.. Satoh, K. and Dikmak, E. (1962). Osmoticchangescausedby the accumulation of dulcitol in the lensesof rats fed with galactose.Nature 194, 1085-7. Robison,W. G.. Kador. P. F. and Kinoshita, J. H. ( 1983). Retinal capillariesbasementmembranethickening by galactosaemia prevented by an aldose reductase hibitor. Science 221, 1177-9.
in-
