CURRENT CONCEPTS ON PATHOGENESIS OF DIABETIC RETINOPATHY: A DYSPROTEINEMIA* BY Hunter L. Little, MD, Alvin Sacks, PhD (BY INVITATION), Arthur Vassiliadis, PhD (BY INVITATION), AND (BY INVITATION) Richard Greer, PhD INTRODUCTION
EVEN THOUGH THE
PATHOGENESIS OF DIABETIC RETINOPATHY IS POORLY UNDER-
stood, most ophthalmologists agree that the onset and the severity of diabetic retinopathy are usually related to the duration of diabetes. Most investigation on pathogenesis has been directed toward changes in the vessel wall with particular attention toward thickening of the capillary basement membrane1 and loss of the pericyte.2 In fact, some investigators have suggested that diabetic microangiopathy is a genetic disease of basement membrane related to duration and unrelated to the role of abnormal carbohydrate metabolism.1 The purpose of this presentation is to review recent evidence which strongly suggests that the microvascular complications of diabetes are related to duration of abnormal carbohydrate metabolism and to changes in the components of blood with particular reference to altered plasma proteins which induce abnormal red cell and platelet aggregation. The abnormal plasma proteins seem to be related to alterations in fibrinolysis. Other studies have shown evidence that suggests that the microvascular complications of diabetes are not caused by the duration of the disease but by the duration of poor control of diabetes. In view of the current evidence to be presented, the following hypothesis of the scheme of events leading to diabetic retinopathy is presented. 1. Insulin deficiency with hyperglycemia stimulates increased plasma levels of growth hormone.3 *From the Palo Alto Medical Research Foundation, 860 Bryant Street, Palo Alto, and Palo Alto Retinal Research Foundation, 1225 Crane Street, Menlo Park, and Palo Alto Retinal Group, Inc., 1225 Crane Street, Menlo Park, California 94025. Supported by N.I.H. Grant #EY 02177-01. TR. AM. OPHTH. Soc., vol. LXXV, 1977
398
Little et al
2. Concentrations of growth hormone and of insulin critically influence hepatic cell synthesis of plasma protein components probably accounting for increased synthesis of fibrinogen and alpha-2 globulin and decreased albumin.4'5 3. Increased concentrations of plasma (macroglobulins) leads to red cell aggregation 67'8 and platelet clumping.9"10 4. Alpha-2 globulin inhibits leukocyte proteases and results in increased thickening of capillary basement membranes.11"12 5. Clumped red cells in presence of thickened capillary basement membrane causes impaired capillary perfusion and retinal ischemia.7 6. The sequelae of retinal ischemia and hypoxia are vasodilatation, capillary leakage, microaneurysms, and retinal neovascularization. 13,14,15 ROLE OF ABNORMAL CARBOHYDRATE METABOLISM
The ability to induce diabetic microangiopathy in streptazotazine diabetic rats and alloxan diabetic dogs having no genetic background of diabetes suggests that altered carbohydrate metabolism may play a significant role in the pathogenesis of microangiopathy. 16"17 Furthermore, the presence of statistically significant greater retinal pathology in the poorly controlled diabetic dogs versus moderately well controlled dogs suggests that regulation of blood sugar may be important in minimizing the development of
retinopathy.'8 Recent studies suggest a positive relationship between poor metabolic control and the development of microangiopathy in humans. Didier and co-workers in a 3 year study compared the rate of development of microaneurysms in patients on a single injection of insulin versus patients on multiple divided insulin injections per day. 19 They observed statistically significant reductions of 24 hour glycosuria, fasting blood sugars, and numbers of new microaneurysms in the patients on multiple injections. Takazakura and associates performed serial renal biopsies or postmortem examinations on 23 subjects in a ten year study to evaluate the relationship of progression of diabetic glomerulosclerosis to control of hyperglycemia and to type of diabetes mellitus.20 The average time between renal biopsies or biopsy and postmortem examination was 52.6 months. Progression of glomerular lesions was significantly related to control of blood sugar and to type of diabetes but not to duration of diabetes. In addition to having more rapid progression in juvenile onset diabetes than adult onset or the length of follow up, they concluded that poor control of blood glucose might be important in causing the progression of glomerulosclerosis.
399 Diabetic Retinopathy The transplantation of the pancreatic islets of Langerhan into the portal vein in diabetic humans and in streptozotocin treated rats has resulted in marked reduction in the required exogenous insulin to control blood sugar.21'22 Furthermore, studies on progression of glomerular lesions in streptozotocin diabetic rats showed reduction of glomerular pathology in animals following the islet cell transplantation.22 These studies suggest that the glomerular lesions are reversible and that the lesions are related to poor control of blood sugar. CARBOHYDRATE METABOLISM AND GROWTH HORMONE
The relationship of abnormal carbohydrate metabolism to the development of retinopathy hinges on growth hormone. Serum growth hormone levels are three to four times greater in diabetics than nondiabetics,23 24 and patients with retinopathy have higher levels of growth hormone than patients without retinopathy.25 Following rigid control of blood sugar in juvenile diabetics, Hansen noted that the hypersecretion of growth hormone was normalized.3 It is of further interest that sexual ateliotic dwarfs with deficient growth hormone but with hyperglycemia never develop retinopathy.25 Furthermore, it is well known that animals can be made diabetic following multiple injections of growth hormone.27.28 29 Thus, the development of diabetic retinopathy seems dependent on elevated levels of growth hormone, and the hypersecretion of growth hormone seems to result from insulin deficiency and hyperglycemia. Even though acromegalics have almost a 50% incidence of abnormal carbohydrate metabolism, one questions why all acromegalics do not have diabetes. The explanation may be that in acromegaly there are elevated levels of both growth hormone and insulin; whereas, in diabetes there is elevation of growth hormone but deficient levels of insulin. 12 Thus, the critical ratio of growth hormone to insulin is not changed in acromegaly, and the hepatic synthesis of proteins is not altered as in diabetes. One year ago one of the authors (H. L. L.), in his thesis for this society, presented data showing increased red cell aggregation in diabetics with retinopathy.7 Furthermore, the increased red cell aggregation and severity of retinopathy were correlated with elevated levels of plasma fibrinogen and alpha-2 globulins and with reduced levels of plasma albumin. The author suggested that red cell aggregates probably impair blood flow in the microcirculation leading to retinal hypoxia and focal retinal ischemia which are the precursors to microaneurysm formation and retinal neovascularization.5 Almer and co-workers observed abnormalities of the fibrinolytic system in diabetics with microangiopathy in contrast to almost normal fibrinolytic
Little et al
400
activity in those without retinopathy.30 Their investigations corroborate reports of others which have shown reduced fibrinolytic activity in diabetics with vascular disease and elevated plasma levels of factor VIII (anti-hemophalic A factor) and fibrinogen.32233'34'35 Platelets from diabetic subjects have been shown to display increased sensitivity to platelet aggregating agents. 10.34235 Colwell and co-worker's investigation demonstrates that the increased platelet aggregation is probably due to plasma factors present in diabetics. 10 Thus, evidence suggests that the vascular complications in diabetes are due at least in part to alteration of plasma proteins including those of the fibrinolytic system. Plasma macroglobulins probably cause red cell platelet aggregation by bridging the electronegative field which normally prevents aggregation.36 ABNORMAL HEMORRHEODYNAMICS
Investigations by the authors continue to support previous results7 and reports of others that diabetic microangiopathy is related to changes in properties of blood itself. This report gives a summary of the results of a five year study on the role of altered blood rheology in the pathogenesis of diabetic retinopathy. METHODS
A number of concurrent studies were made in order to obtain some understanding of the relationship between blood chemistry, red cell aggregation, and diabetic retinopathy. The subjects were divided into four groups:
Group # 1 2 3 4
Definition
Healthy subjects - controls Diabetics with no retinopathy Diabetics with nonproliferative retinopathy Diabetics with active proliferative retinopathy
Abbreviation (C) (NR) (NPDR)
(PDRA)
CLINICAL OBSERVATIONS
Ophthalmoscopic examinations were conducted to determine the extent of diabetic retinopathy. Studies included the following techniques:
Diabetic Retinopathy
401
1) Observations with direct, indirect, and slit lamp ophthalmoscopy were conducted with emphasis on detecting early evidence of sludged flow, retinal ischemia, venous dilatation, beading, reduplication, microaneurysms, exudates, and neovascularization. Retinopathy was graded according to the following categories: minimal to no retinopathy which includes up to five microaneurysms; nonproliferative retinopathy; and active proliferative retinopathy showing retinal neovascularization. 2) Fundus photographs using the 35 mm Zeiss fundus camera were taken. These were studied to detect vascular changes suggestive of altered rheology (sludged or aggregated erythrocytes), retinal ischemia, venous and arterial changes, microaneurysms, exudates, and neovascularization. 3) Intravenous fluorescein retinal angiography was done on at least one eye of all diabetic patients. Evaluation included detection of areas of nonperfusion, circulation time, and mapping of retinal neovascularization. BLOOD CHEMISTRY MEASUREMENTS
Blood chemical measurements consisted of eight-hour fasting glucose, blood urea nitrogen (B. U. N.), cholesterol, triglyceride, uric acid, total protein, serum protein electrophoresis, and fibrinogen determinations. Blood types were determined, and complete blood counts were done on all subjects. On a small sample, platelet counts and platelet aggregation were measured on a few patients. Blood specimens were collected in clot tubes for serum studies. Specimens for fibrinogen determination were collected in tubes with citrate added. RHEOLOGIC STUDIES
During the course of the project, we have developed and perfected four different rheological tests and the required instrumentation to determine the physical and elastic properties of red cell aggregates and their resistance to flow. All tests are conducted at 37°C and at 40% hematocrit, except for the visualization of red cell aggregates in the rheoscope which is performed with 1% suspensions of red cells in plasma. The rheological tests consist of performing the following measurements: 1. Red Cell Aggregation
Red cell aggregation was estimated by the use of rheoscopy. Fasting
402
Little et al
samples were collected for all studies in EDTA tubes. The buffy coat was removed to eliminate platelets and leukocytes. Observation and photography of red cell aggregation in the Wells-Brookfield rheoscope were made at strain rates of 12 rpm (46 1/sec), 6 rpm (23 1/sec), 3 rpm (115 1/sec), and 1.5 rpm (5.75 1/sec) on 1% suspensions of red cells in plasma. This technique eliminates hematocrit as a variable and facilitates observation of single cell, rouleaux, and aggregate formation and dispersion. In special cases, similar concentrations of red cells were observed in serum. Observations were made after running the instrument for two minutes at each setting. The degree of aggregation was expressed in terms of a classification which is based primarily on variations of light intensity across the microscopic field. 2. Red Cell Sedimentation Rates
Routine complete blood counts and one hour Wintrobe sedimentation rates were carried out on all subjects. 3. Viscosity Measurements
The Wells-Brookfield microviscometer was used for all viscosity measurements. Measurements were made at 37°C in plasma, serum and suspensions of 40 percent red cells in native plasma. 4. Filtration Tests
Basically, the filtration system consists of a Sage motor-driven syringe pump, a syringe whose cylinder is fitted with a thermistor to sense the blood temperature inside, five in-line Sartorius filter holders (specially machined on the inside), each fitted with factory light-transmission tested Swank filters with 10 , square holes. The entire system, when filled with the blood sample to be tested, is inserted into a horizontally mounted thermally wrapped, plexiglass cylinder whose temperature is controlled to maintain 37°C, and the height of blood rising in a vertical capillary tube proximal to the filters is measured. During each run, blood in the syringe is continually mixed by means of a Magmix stirring unit. With this system, we have succeeded in measuring the pressure drop required to drive the blood sample through the filters at Reynolds numbers of about 0.001 and strain rates of about 300 1/sec, which corresponds quite well with those of the microcirculation. The pressure drop reaches equilibrium in a period of about 3 minutes, and final readings are taken only after steady-state equilibrium is achieved.
Diabetic Retinopathy
403
COMPUTER PROGRAMS AND CALCULATIONS
Several computer programs were written to facilitate the analysis of the data. The data that has been analyzed is based on data taken on four groups of people: 1 - Healthy subjects - controls (C) 2 - Diabetics with no retinopathy (NR) 3 - Diabetics with nonproliferative retinopathy (NPDR) 4 - Diabetics with active proliferative retinopathy (PDRA) In the process of analysis it was discovered that the blood chemistry of the males in these four groups differed considerably from that of the females and it was decided to treat each sex separately. Splitting each of the four clinical groups by sex gave rise to the eight distinct groups that are analyzed with the help of the computer programs. Table I gives the number of individuals in each group that have recently been analyzed and discussed below. Although more than 230 patients have been examined and had clinical and laboratory tests made - only the 166 noted in Table I are used in the analyses discussed below. The remainder were eliminated because they had inactive proliferative retinopathy, or the dates of clinical examination and rheological examinations differed by more than three months. Altogether, 38 measurements were made on each individual in each of the eight groups. These measurements ranged from statistics such as age and sex to such calculated quantities as the albumin to fibrinogen ratio (A/F). Four computer programs were written to do the bulk ofthe data analysis. They were written for and run on an IBM 370 Model 168 computer at the Stanford Center for Information Processing. We discuss four of them hereunder. The first program read in data punched on cards, checked for gross errors, and calculated special interest variables such as the A/F ratio. The resulting data were then re-formatted and output to a disk file to serve as imput to the other programs. The function of the second program was to provide a readily accessible summary of all the measurements for each of the individuals. The output of the program was divided into sections, one for each group. Each section contained tables which displayed the measurements for each individual in the appropriate group. Also included were statistics that suggested various characteristics of the group by summarizing the measurement values assumed by the group for each of the 58 variables. These statistics included such order statistics as the minimum, maximum, median, and upper and lower quartiles, estimators of central tendency
404
Little et al
such as the mean and . 15-trimmed mean, and estimators of variability such as the standard deviation, range, and midrange. The third program, instead of presenting the data group by group, presents it variable by variable. This is a more useful format since for each variable, the observations within each of the eight groups are ordered and presented. The output from this program enables one to get as detailed a picture as desired as to how the eigth groups differ on each of the variables. The means and standard deviation for the eight groups for a number of the measured parameters are summarized in Table II for males and Table III for females. The purpose of the fourth program is to provide a means of testing of equality of distribution between groups and to then rank the variables based on how well they discriminate between the four clinical groups by sex. Two runs of the program are then made, one for the males and one for the females. For a given run and a given variable, there are six possible ways to pair the four clinical groups and then test for equality of distribution, i.e. to examine whether the difference between groups are statistically significant. Since differences in location are the natural differences to look for between two distributions in this situation, a standardized Wilcoxon statistic (corrected for continuity and ties) was calculated for each of the pairings of the four clinical groups for each variable. For example, for the fibrinogen variable, Wilcoxon statistics were computed to see whether the control males came from the same distribution as the NR diabetic males, whether control males came from the same distribution as the NPDR diabetic males, etc. The Wilcoxon test scores calculated for the males are shown in Table IV and for the females in Table V. The columns indicate the pairs of groups for which the comparisons are made. In order to easily identify the locations where rejections of significance of p
EVEN THOUGH THE
PATHOGENESIS OF DIABETIC RETINOPATHY IS POORLY UNDER-
stood, most ophthalmologists agree that the onset and the severity of diabetic retinopathy are usually related to the duration of diabetes. Most investigation on pathogenesis has been directed toward changes in the vessel wall with particular attention toward thickening of the capillary basement membrane1 and loss of the pericyte.2 In fact, some investigators have suggested that diabetic microangiopathy is a genetic disease of basement membrane related to duration and unrelated to the role of abnormal carbohydrate metabolism.1 The purpose of this presentation is to review recent evidence which strongly suggests that the microvascular complications of diabetes are related to duration of abnormal carbohydrate metabolism and to changes in the components of blood with particular reference to altered plasma proteins which induce abnormal red cell and platelet aggregation. The abnormal plasma proteins seem to be related to alterations in fibrinolysis. Other studies have shown evidence that suggests that the microvascular complications of diabetes are not caused by the duration of the disease but by the duration of poor control of diabetes. In view of the current evidence to be presented, the following hypothesis of the scheme of events leading to diabetic retinopathy is presented. 1. Insulin deficiency with hyperglycemia stimulates increased plasma levels of growth hormone.3 *From the Palo Alto Medical Research Foundation, 860 Bryant Street, Palo Alto, and Palo Alto Retinal Research Foundation, 1225 Crane Street, Menlo Park, and Palo Alto Retinal Group, Inc., 1225 Crane Street, Menlo Park, California 94025. Supported by N.I.H. Grant #EY 02177-01. TR. AM. OPHTH. Soc., vol. LXXV, 1977
398
Little et al
2. Concentrations of growth hormone and of insulin critically influence hepatic cell synthesis of plasma protein components probably accounting for increased synthesis of fibrinogen and alpha-2 globulin and decreased albumin.4'5 3. Increased concentrations of plasma (macroglobulins) leads to red cell aggregation 67'8 and platelet clumping.9"10 4. Alpha-2 globulin inhibits leukocyte proteases and results in increased thickening of capillary basement membranes.11"12 5. Clumped red cells in presence of thickened capillary basement membrane causes impaired capillary perfusion and retinal ischemia.7 6. The sequelae of retinal ischemia and hypoxia are vasodilatation, capillary leakage, microaneurysms, and retinal neovascularization. 13,14,15 ROLE OF ABNORMAL CARBOHYDRATE METABOLISM
The ability to induce diabetic microangiopathy in streptazotazine diabetic rats and alloxan diabetic dogs having no genetic background of diabetes suggests that altered carbohydrate metabolism may play a significant role in the pathogenesis of microangiopathy. 16"17 Furthermore, the presence of statistically significant greater retinal pathology in the poorly controlled diabetic dogs versus moderately well controlled dogs suggests that regulation of blood sugar may be important in minimizing the development of
retinopathy.'8 Recent studies suggest a positive relationship between poor metabolic control and the development of microangiopathy in humans. Didier and co-workers in a 3 year study compared the rate of development of microaneurysms in patients on a single injection of insulin versus patients on multiple divided insulin injections per day. 19 They observed statistically significant reductions of 24 hour glycosuria, fasting blood sugars, and numbers of new microaneurysms in the patients on multiple injections. Takazakura and associates performed serial renal biopsies or postmortem examinations on 23 subjects in a ten year study to evaluate the relationship of progression of diabetic glomerulosclerosis to control of hyperglycemia and to type of diabetes mellitus.20 The average time between renal biopsies or biopsy and postmortem examination was 52.6 months. Progression of glomerular lesions was significantly related to control of blood sugar and to type of diabetes but not to duration of diabetes. In addition to having more rapid progression in juvenile onset diabetes than adult onset or the length of follow up, they concluded that poor control of blood glucose might be important in causing the progression of glomerulosclerosis.
399 Diabetic Retinopathy The transplantation of the pancreatic islets of Langerhan into the portal vein in diabetic humans and in streptozotocin treated rats has resulted in marked reduction in the required exogenous insulin to control blood sugar.21'22 Furthermore, studies on progression of glomerular lesions in streptozotocin diabetic rats showed reduction of glomerular pathology in animals following the islet cell transplantation.22 These studies suggest that the glomerular lesions are reversible and that the lesions are related to poor control of blood sugar. CARBOHYDRATE METABOLISM AND GROWTH HORMONE
The relationship of abnormal carbohydrate metabolism to the development of retinopathy hinges on growth hormone. Serum growth hormone levels are three to four times greater in diabetics than nondiabetics,23 24 and patients with retinopathy have higher levels of growth hormone than patients without retinopathy.25 Following rigid control of blood sugar in juvenile diabetics, Hansen noted that the hypersecretion of growth hormone was normalized.3 It is of further interest that sexual ateliotic dwarfs with deficient growth hormone but with hyperglycemia never develop retinopathy.25 Furthermore, it is well known that animals can be made diabetic following multiple injections of growth hormone.27.28 29 Thus, the development of diabetic retinopathy seems dependent on elevated levels of growth hormone, and the hypersecretion of growth hormone seems to result from insulin deficiency and hyperglycemia. Even though acromegalics have almost a 50% incidence of abnormal carbohydrate metabolism, one questions why all acromegalics do not have diabetes. The explanation may be that in acromegaly there are elevated levels of both growth hormone and insulin; whereas, in diabetes there is elevation of growth hormone but deficient levels of insulin. 12 Thus, the critical ratio of growth hormone to insulin is not changed in acromegaly, and the hepatic synthesis of proteins is not altered as in diabetes. One year ago one of the authors (H. L. L.), in his thesis for this society, presented data showing increased red cell aggregation in diabetics with retinopathy.7 Furthermore, the increased red cell aggregation and severity of retinopathy were correlated with elevated levels of plasma fibrinogen and alpha-2 globulins and with reduced levels of plasma albumin. The author suggested that red cell aggregates probably impair blood flow in the microcirculation leading to retinal hypoxia and focal retinal ischemia which are the precursors to microaneurysm formation and retinal neovascularization.5 Almer and co-workers observed abnormalities of the fibrinolytic system in diabetics with microangiopathy in contrast to almost normal fibrinolytic
Little et al
400
activity in those without retinopathy.30 Their investigations corroborate reports of others which have shown reduced fibrinolytic activity in diabetics with vascular disease and elevated plasma levels of factor VIII (anti-hemophalic A factor) and fibrinogen.32233'34'35 Platelets from diabetic subjects have been shown to display increased sensitivity to platelet aggregating agents. 10.34235 Colwell and co-worker's investigation demonstrates that the increased platelet aggregation is probably due to plasma factors present in diabetics. 10 Thus, evidence suggests that the vascular complications in diabetes are due at least in part to alteration of plasma proteins including those of the fibrinolytic system. Plasma macroglobulins probably cause red cell platelet aggregation by bridging the electronegative field which normally prevents aggregation.36 ABNORMAL HEMORRHEODYNAMICS
Investigations by the authors continue to support previous results7 and reports of others that diabetic microangiopathy is related to changes in properties of blood itself. This report gives a summary of the results of a five year study on the role of altered blood rheology in the pathogenesis of diabetic retinopathy. METHODS
A number of concurrent studies were made in order to obtain some understanding of the relationship between blood chemistry, red cell aggregation, and diabetic retinopathy. The subjects were divided into four groups:
Group # 1 2 3 4
Definition
Healthy subjects - controls Diabetics with no retinopathy Diabetics with nonproliferative retinopathy Diabetics with active proliferative retinopathy
Abbreviation (C) (NR) (NPDR)
(PDRA)
CLINICAL OBSERVATIONS
Ophthalmoscopic examinations were conducted to determine the extent of diabetic retinopathy. Studies included the following techniques:
Diabetic Retinopathy
401
1) Observations with direct, indirect, and slit lamp ophthalmoscopy were conducted with emphasis on detecting early evidence of sludged flow, retinal ischemia, venous dilatation, beading, reduplication, microaneurysms, exudates, and neovascularization. Retinopathy was graded according to the following categories: minimal to no retinopathy which includes up to five microaneurysms; nonproliferative retinopathy; and active proliferative retinopathy showing retinal neovascularization. 2) Fundus photographs using the 35 mm Zeiss fundus camera were taken. These were studied to detect vascular changes suggestive of altered rheology (sludged or aggregated erythrocytes), retinal ischemia, venous and arterial changes, microaneurysms, exudates, and neovascularization. 3) Intravenous fluorescein retinal angiography was done on at least one eye of all diabetic patients. Evaluation included detection of areas of nonperfusion, circulation time, and mapping of retinal neovascularization. BLOOD CHEMISTRY MEASUREMENTS
Blood chemical measurements consisted of eight-hour fasting glucose, blood urea nitrogen (B. U. N.), cholesterol, triglyceride, uric acid, total protein, serum protein electrophoresis, and fibrinogen determinations. Blood types were determined, and complete blood counts were done on all subjects. On a small sample, platelet counts and platelet aggregation were measured on a few patients. Blood specimens were collected in clot tubes for serum studies. Specimens for fibrinogen determination were collected in tubes with citrate added. RHEOLOGIC STUDIES
During the course of the project, we have developed and perfected four different rheological tests and the required instrumentation to determine the physical and elastic properties of red cell aggregates and their resistance to flow. All tests are conducted at 37°C and at 40% hematocrit, except for the visualization of red cell aggregates in the rheoscope which is performed with 1% suspensions of red cells in plasma. The rheological tests consist of performing the following measurements: 1. Red Cell Aggregation
Red cell aggregation was estimated by the use of rheoscopy. Fasting
402
Little et al
samples were collected for all studies in EDTA tubes. The buffy coat was removed to eliminate platelets and leukocytes. Observation and photography of red cell aggregation in the Wells-Brookfield rheoscope were made at strain rates of 12 rpm (46 1/sec), 6 rpm (23 1/sec), 3 rpm (115 1/sec), and 1.5 rpm (5.75 1/sec) on 1% suspensions of red cells in plasma. This technique eliminates hematocrit as a variable and facilitates observation of single cell, rouleaux, and aggregate formation and dispersion. In special cases, similar concentrations of red cells were observed in serum. Observations were made after running the instrument for two minutes at each setting. The degree of aggregation was expressed in terms of a classification which is based primarily on variations of light intensity across the microscopic field. 2. Red Cell Sedimentation Rates
Routine complete blood counts and one hour Wintrobe sedimentation rates were carried out on all subjects. 3. Viscosity Measurements
The Wells-Brookfield microviscometer was used for all viscosity measurements. Measurements were made at 37°C in plasma, serum and suspensions of 40 percent red cells in native plasma. 4. Filtration Tests
Basically, the filtration system consists of a Sage motor-driven syringe pump, a syringe whose cylinder is fitted with a thermistor to sense the blood temperature inside, five in-line Sartorius filter holders (specially machined on the inside), each fitted with factory light-transmission tested Swank filters with 10 , square holes. The entire system, when filled with the blood sample to be tested, is inserted into a horizontally mounted thermally wrapped, plexiglass cylinder whose temperature is controlled to maintain 37°C, and the height of blood rising in a vertical capillary tube proximal to the filters is measured. During each run, blood in the syringe is continually mixed by means of a Magmix stirring unit. With this system, we have succeeded in measuring the pressure drop required to drive the blood sample through the filters at Reynolds numbers of about 0.001 and strain rates of about 300 1/sec, which corresponds quite well with those of the microcirculation. The pressure drop reaches equilibrium in a period of about 3 minutes, and final readings are taken only after steady-state equilibrium is achieved.
Diabetic Retinopathy
403
COMPUTER PROGRAMS AND CALCULATIONS
Several computer programs were written to facilitate the analysis of the data. The data that has been analyzed is based on data taken on four groups of people: 1 - Healthy subjects - controls (C) 2 - Diabetics with no retinopathy (NR) 3 - Diabetics with nonproliferative retinopathy (NPDR) 4 - Diabetics with active proliferative retinopathy (PDRA) In the process of analysis it was discovered that the blood chemistry of the males in these four groups differed considerably from that of the females and it was decided to treat each sex separately. Splitting each of the four clinical groups by sex gave rise to the eight distinct groups that are analyzed with the help of the computer programs. Table I gives the number of individuals in each group that have recently been analyzed and discussed below. Although more than 230 patients have been examined and had clinical and laboratory tests made - only the 166 noted in Table I are used in the analyses discussed below. The remainder were eliminated because they had inactive proliferative retinopathy, or the dates of clinical examination and rheological examinations differed by more than three months. Altogether, 38 measurements were made on each individual in each of the eight groups. These measurements ranged from statistics such as age and sex to such calculated quantities as the albumin to fibrinogen ratio (A/F). Four computer programs were written to do the bulk ofthe data analysis. They were written for and run on an IBM 370 Model 168 computer at the Stanford Center for Information Processing. We discuss four of them hereunder. The first program read in data punched on cards, checked for gross errors, and calculated special interest variables such as the A/F ratio. The resulting data were then re-formatted and output to a disk file to serve as imput to the other programs. The function of the second program was to provide a readily accessible summary of all the measurements for each of the individuals. The output of the program was divided into sections, one for each group. Each section contained tables which displayed the measurements for each individual in the appropriate group. Also included were statistics that suggested various characteristics of the group by summarizing the measurement values assumed by the group for each of the 58 variables. These statistics included such order statistics as the minimum, maximum, median, and upper and lower quartiles, estimators of central tendency
404
Little et al
such as the mean and . 15-trimmed mean, and estimators of variability such as the standard deviation, range, and midrange. The third program, instead of presenting the data group by group, presents it variable by variable. This is a more useful format since for each variable, the observations within each of the eight groups are ordered and presented. The output from this program enables one to get as detailed a picture as desired as to how the eigth groups differ on each of the variables. The means and standard deviation for the eight groups for a number of the measured parameters are summarized in Table II for males and Table III for females. The purpose of the fourth program is to provide a means of testing of equality of distribution between groups and to then rank the variables based on how well they discriminate between the four clinical groups by sex. Two runs of the program are then made, one for the males and one for the females. For a given run and a given variable, there are six possible ways to pair the four clinical groups and then test for equality of distribution, i.e. to examine whether the difference between groups are statistically significant. Since differences in location are the natural differences to look for between two distributions in this situation, a standardized Wilcoxon statistic (corrected for continuity and ties) was calculated for each of the pairings of the four clinical groups for each variable. For example, for the fibrinogen variable, Wilcoxon statistics were computed to see whether the control males came from the same distribution as the NR diabetic males, whether control males came from the same distribution as the NPDR diabetic males, etc. The Wilcoxon test scores calculated for the males are shown in Table IV and for the females in Table V. The columns indicate the pairs of groups for which the comparisons are made. In order to easily identify the locations where rejections of significance of p
