- Email: [email protected]
The progression of retinopathy over 2 years: The Pittsburgh Epidemiology of Diabetes Complications (EDC) Study
The Progression of Reti .I IoDathv Over 2 Years: The Pittsburgh E Aide&iology of Diabetes Complicatiok (&ik) Study Cathy E. Lloyd Ronald Klein Raelene E. Maser Lewis I-I Kuller Dorothy 1. Becker Trevor 1. Orchard
ABSTRACT This study examined potential risk factors for the incidence and progression of retinopathy in a large representative cohort of childhood onset insulin-dependent diabetic patients. Participants in the Epidemiology of Diabetes Complications (EDC) Study underwent a full clinical examination at baseline and again at a 2-year follow-up. Retinopathy status was ascertained using stereo fundus photographs graded according to the modified Airlie House System. The study population is based on a large cohort of childhood-onset insulin-dependent diabetic patients, seen within 1 year of diagnosis at the Children’s Hospital of Pittsburgh between January 1950 and May 1980. A total of 657 subjects participated at baseline (1986-19881, with 80% of eligible survivors taking part in the follow-up examination. This report concerns risk factors associated with the progression of diabetic retinopathy over a 2-year period, and the interaction of these factors with the presence of nephropathy. Analyses showed that baseline diastolic blood pressure was significantly _--
-
-..
___.
~
-___.
.~_
Department ofEpidemiology (C.E.L., L.H.K., T.J.O.), University of Pittsburgh, Pittsburgh, Pennsylvania; Department of Opthalmology (R.K.), University of Wisconsin Medical School, Wisconsin, Wisconsin; School of Life and Health Sciences (R.E.M.), University of Delaware, Newark, Delaware; Department of Pediatrics (D.J.B.), University of Pittsburgh, Pittsburgh, Pennsylvania, USA Reprint requests to be sent to: Dr. C. E. Lloyd, Department of Epidemiology, University of Pittsburgh, Diabetes Research Center, 5th Floor Rangos Building, 3460 Fifth Avenue, Pittsburgh, PA 15213.
~ourml of Diabetes and Its Complications 1995; 9:140-148 0 Elsevier Science Inc., 1995 655 Avenue of the Americas, New York, NY 10010
associated with the incidence of any retinopathy, while glycosylated hemoglobin, baseline severity of retinopathy, serum triglycerides, and, to a lesser extent, higher levels of low-density lipoprotein (LDL) cholesterol and fibrinogen were associated with the progression of retinopathy. Progression to proliferative retinopathy was related to higher LDL cholesterol, fibrinogen, serum triglycerides, albumin excretion rate, and glycosylated hemoglobin (GHb). Risk factors varied with the presence of nephropathy. In the absence of nephropathy, GHb was a significant predictor of progression, whilst this was not the case in the presence of nephropathy. Multivariate analyses demonstrated that the incidence of new cases of retinopathy was associated only with baseline diastolic blood pressure, while both the progression of retinopathy and progression to proliferative retinopathy were significantly associated with baseline glycosylated hemoglobin and baseline severity of retinopathy. These results suggest that the risk factors for the development of retinopathy may differ at various stages of the disease and according to the presence of nephropathy. While blood pressure, even within the normal range, may be important in the early stages of retinopathy, glycemic control remains the most important factor in the long-term development of this complication. (Journal of Diabetes and Its Complications 9;3:140-148, 1995.)
SSDI
10568727/951$9.50 1056-8727(94)00039-S
] Diab Comp
RETINOPATHY
1995; 9:140-148
INTRODUCTION
R
etinopathy, the most frequent complication of diabetes, is determined to be the leading cause of blindness in some countries.’ This complication, after 20 years, may come to affect virtually all those with insulin-dependent diabetes mellitus (IDDM).‘f3 Until recently, the available evidence linking potential risk factors to the development of diabetic retinopathy has come mainly from crosssectional studies, which have often been inconclusive. Furthermore, studies have not considered all possible risk factors, the Epidemiology of Diabetes Complications (EDC) study being the first to incorporate variables such as coagulation factors and diabetic autonomic neuropathy. However, recently two trials have shown clear evidence that intensive therapy regimens designed to normalize blood sugars as far as possible, dramatically retard (but do not eliminate) the development and progression of retinopathy.4,5 These trial data (which are needed to prove causality) are consistent with prior (epidemiologic and clinical) studies of the role of glycemic control as an important predictor of both early retinopathy changes, and the later progression of retinopathy.3,6-9 Other factors may be important, and raised lipid levels and higher blood pressures have been implicated.‘OI3 Early onset of proliferative retinopathy has also been linked to the presence of nephropathy,14,15 although it is unclear how this “renal-retinal” connection operates. The aim of this study was therefore to identify possible risk factors for the development and progression of retinopathy in a representative16 cohort of childhood-onset IDDM patients, and to examine if these risk factors are affected by the presence of renal disease. The latter objective will help resolve whether the “renal-retinal” link is mediated by other factors (i.e., renal disease may be acting as a confounder) or whether it modifies the effect of other risk factors (i.e., is an effect modifier) or whether it represents an independent predictor. A recent report stresses both a strong association between renal histological lesions and retinal structural lesions, and a fair degree of discordance of renal/retinal disease.17 METHODS The study participants were all childhood-onset (<17 years) IDDM patients seen at or within 1 year of diagnosis at the Children’s Hospital of Pittsburgh between January 1950 and May 1980, and living within 100 miles of Pittsburgh at the time of the study. All potential subjects were invited by letter to attend a clinical examination and to complete a series of questionnaires which were mailed to the participant 2 weeks prior to the examination. Of the 979 eligible, 67% participated
PROGRESSION
141
in a full evaluation. The representativeness of this group had been reported previously, as have full details of both the baseline questionnaires and clinical examinations that took place between 1986 and 1988.2,‘6 This report focuses on the incidence and progression of diabetic retinopathy based upon status at entry to the study (baseline) and at follow-up 2 years later. At the baseline clinical examination, after informed written consent, subjects underwent a sitting blood pressure measurement with a random zero sphygmomanometer (Hypertension Detection and Follow-up Program ProtocoP) with the mean of the second and third blood pressures used in the analyses. Blood samples were taken for measurement of lipids, lipoprotein, glycosylated hemoglobin (GHb), and fibrinogen. Albumin excretion rates (AER) were calculated using three timed urine samples; i.e., a 24-h collection, an overnight sample and one collected during the clinic. These three samples were usually collected within a 2-week period. Overt nephropathy was defined as the presence of renal failure (on dialysis and/or postkidney transplant) or an AER of greater than 200 1.181 min in at least two of three timed urine samples. Microalbuminuria was defined as an AER between 20 and 2001uglmin. Laboratory Determinations. For the first 18 months of the study, GHb (stable levels of total hemoglobin Ai) was determined with saline-incubated blood samples and microcolumn cation-exchange chromatography (Isolab, Akron, OH). Thereafter GHb was measured using automated high-performance liquid chromatography (Diamat, Bio-Rad, Hercules, CA). The two methods were shown to be almost identical (r = 0.95), with an absolute difference of 0.158 (%I-IbAI) . The normal range for GHb was 4.9%-7.3%. Serum cholesterol and triglyceride levels were determined enzymatitally. 1g,20 High-density lipoprotein cholesterol was measured by a heparin and manganese procedure with modification of the Lipid Research Clinics methodology. 21,22Low-density lipoprotein cholesterol was calculated with the Friedewald equation if triglycerides were less than 4.5 n&I.= Fibrinogen levels were measured using a biuret colometric procedure and a clotting method. Measurement of Diabetic Retinopathy. Stereo fundus photographs (eyes dilated) were taken of fields one, two, and four, with a Zeiss camera and were read in a masked fashion by the Fundus Photography Reading Center at the University of Wisconsin, Madison. Although three fields were used rather than the standard seven, it has been previously demonstrated that this gives very satisfactory results with good sensitivity. 24 Readings were classified according to the modified Airlie House System.25*26 Subjects were clas-
142
LLOYDETAL.
] Diab Comp
1. BASELINE
TABLE -
- .---.
STATUS OF ATTENDERS VERSUS NONATTENDERS FOR CYCLE TWO (FOLLOW-UP) ..--. - ..-
Did not attend follow-up
follow-up (n = 496)
MI,
myocardiul
infarction;
DCCT,
Diabetes
Control
and Complications
--
.--
9: l40- 148
ELIGIBLE
Attended
Age yr, mean (SD) Duration of diabetes yr, mean (SD) Males (%) Somecollege education (%) (for >18 yr olds) Overt nephropathy (96) (albumin excretion rate >200 &mm) Cardiovascular disease(%) (MT, angina, or stroke) Peripheral vascular disease(%) (Ankle-arm blood pressureratio <0.8, or amputation) Distal symmetrical neuropathy (%) (DCCT protocol) Hypertension (%) (>140/90 mm Hg and/or therapy) __ -~._-~ ~.~ _~- ~
7 9%;
(n = 121)
27.8 (7.9) 19.4 (7.5) 53 65 24 4
26.3 (7.8)* 18.7 (7.5) 40* 52* 31 2
9
8 22* 16
31 17
Trial.
* p < 0.05. sified into one of the following
categories based on the findings in the worst eye: no retinopathy (level lo), minimal background retinopathy (level 21), mild background retinopathy (level 31), moderate nonproliferative retinopathy (level 41), severe nonproliferative retinopathy (level 51), and proliferative retinopathy (level 60 or above). Both the incidence rates and progression of retinopathy were calculated in the same manner as the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR),27 as follows. Incidence of any retinopathy was calculated from all persons who had no retinopathy at baseline examination and who took part in the follow-up examination (i.e., moved from no retinopathy at baseline to levels 21/<21 or worse at follow-up). Progression of any retinopathy was defined as an increase in severity of two steps or more on the modified Airlie House system, and was calculated from all those with no or nonproliferative retinopathy at baseline. Progression to proliferative retinopathy was estimated from all subjects with no or nonproliferative retinopathy (i.e., classified at level 51/51 or better) at baseline who had deteriorated to proliferative retinopathy (level 60+ /<60 or worse) by follow-up. Statistical analyses were performed using the statistical package for the social sciences (SPSSX), and included Student’s t test, analysis of variance, and x2. Values given in tables are either means f SD or percent (%). Multiple logistic regression analyses were carried out using BMDP, where the goodness of fit of the specified models was determined. Using this statistical package it is possible to estimate the relative risk for each significant variable in order to illustrate the increase in risk of developing a particular end point
(in this case retinopathy) without the risk factor.
in those with versus those
RESULTS Response Rate. Of the 657 people who were examined at baseline, 27 had moved out of the area and were no longer eligible, and 13 had died, leaving 617 eligible, of whom 496 (80.4%) attended follow-up. Of those who returned, one-third already had had proliferative retinopathy at baseline and so were therefore ineligible for these analyses, leaving 322 (65%) at risk of the development or progression of retinopathy. Mean duration of diabetes at this time was 21.7 f 7.7 years, and the average interval between baseline and follow-up was 24.9 + 4.5 months. At follow-up, participants did not differ from nonparticipants in terms of income level but were more likely to have a college education (Table 1). A greater proportion of males returned than females. Although those who took part in both phases of the study were more likely to have had neuropathy at baseline, there were no differences in follow-up attendance in terms of other major complication at entry into the study. Progression of Diabetic Retinopathy. At the time of follow-up, a total of 23% (74/322) of subjects at risk of progression (i.e., without proliferative disease at baseline) had progressed by two steps or more, and 10% (33/322) had progressed to proliferative retinopathy. At follow-up examination, 20 subjects showed some improvement of retinopathy; these cases were analyzed together with those whose retinopathy had remained the same. A total of 12% (45/376) had received panretinal photocoagulation therapy between baseline and follow-up. Duration of diabetes was significantly longer for those who had progressed to PR (18.9 f 5.8 versus 16.5 k 6.6 years; p = 0.04). How-
1 Diab Comp
RETlNOPATHY
1995; 9:140-148
Incidence of Retinopathy. An entry into this study, only 69 subjects were still at risk of developing any new retinopathy, and one-third (23/69) had done so by follow-up, giving an incidence density of 16.6 cases/ 100 person years of follow-up. An examination of potential risk factors in these participants identified only baseline diastolic blood pressure as significantly different in those who developed new retinopathy compared to those who remained free of this complication (71.4 f. 8.8 versus 63.9 f 8.9 mm Hg; p < 0.001). Incidence of retinopathy was not associated with baseline GHb (10.2% in those who developed retinopathy versus 10.5% in those who did not). These findings were confirmed in multivariate analysis adjusting for both duration of diabetes and length of follow-up, when diastolic blood pressure was shown to be a significant independent risk factor for the development of new retinopathy (relative risk 1.15; 95% confidence interval, 1.05-1.26). Risk Factors Associated with Progression of Retinopathy. The remainder of this report focuses on the overall progression of retinopathy and on progression to proliferative disease. Subjects whose retinopathy had progressed by two steps or more at follow-up had a significantly higher mean GHb at baseline compared to those who had not progressed (10.9 versus 10.0%; p < 0.001). Subjects who experienced progression to proliferative retinopathy also had a significantly higher baseline GHb, even when duration of diabetes was accounted for (11.6 versus 10.1%; p < 0.001). In order to examine this relationship further, baseline GHb was
GHb
2. P-YEAR
(O/o)
Range
1st quartile 2nd quartile 3rd quartile
G9.0 9.1-9.9 10.0-11.1 >ll.l
4th quartile GHb,
glycosylated
PROGRESSION RATES NONPROLIFERATIVE
hemoglobin; RR, relative 0.001 Mantel-Haenszel
** p < 0.02, *** p <
BY QUARTILES OF BASELINE GHb, RETINOPATHY AT BASELINE Progression
No. at risk
%
RR
79 79 80 78
14 20 28 32
1.0 1.4 2.0 2.3
risk; CL, confidence test for trend.
143
divided into quartiles and progression of retinopathy and progression to proliferative retinopathy were examined. As Table 2 shows, there was a higher relative risk (RR) for both overall progression and progression to proliferative retinopathy for subjects with a GHb in the upper quartiles of the distribution. Severity of retinopathy at baseline was also significantly associated with the progression of retinopathy during follow-up. Table 3 shows the 2-year progression of retinopathy by quartile of GHb and baseline severity. As the table demonstrates, there was overall, a consistent increase in the proportion who progressed during follow-up according to both greater severity and the higher quartile of GHb. Tests for linear trend indicated that this increase was only statistically significant in those with moderate to severe nonproliferative retinopathy at baseline, but reached borderline significance for those with minimal level of severity at baseline. Other factors were also examined in relation to both the progression and progression to proliferative retinopathy during follow-up, with analyses adjusted for duration of diabetes (see Table 4). Higher fibrinogen levels and higher low-density lipoprotein (LDL) cholesterol were both associated with the overall progression of retinopathy, but these relationships were of borderline significance. Progression of retinopathy was also associated with higher serum triglyceride levels and a higher AER. Progression to proliferative retinopathy was significantly associated with higher serum triglycerides, higher LDL cholesterol, and somewhat higher fibrinogen levels (Table 4). Progression to proliferative retinopathy was also significantly related to baseline nephropathy status. As Figure 1 shows, a significantly greater proportion of those who had microalbuminuria or overt nephropathy at entry into the study progressed to proliferative disease over the 2-year follow-up period. There was no relationship between baseline nephropathy status and overall progression of retinopathy. The association between nephropathy and proliferative disease was further explored in order to assess
ever, duration was not significantly related to the overall progression of retinopathy over the 2-year followup. Progression to proliferative retinopathy was not significantly affected by length of follow-up interval. Overall progression of retinopathy was associated with length of follow-up, albeit a small difference (25.9 f 4.7 months compared with 24.4 f 4.4 months, p = 0.019, for those whose retinopathy did not progress). Overall progression and progression to proliferative retinopathy were both unrelated to gender.
TABLE
PROGRESSION
interval.
0.6, 4.0 1.0, 5.7 1.2, 7.0**
WITH
Progression to proliferative retinopathy _ .~~.
of retinopathy 95% C.I.
FOR PERSONS
%
RR
5 5 11 20
1.0 1.0 2.2 4.0***
95% C.I.
0.3, 3.9 0.7, 6.6 1.4, 11.4
14.4
TABLE
LLOYD
ET AL.
1 Diab Camp
iio’
3. 2-YEAR PROGRESSION (o/o) OF RETINOPATHY BY QUARTILES OF GLYCOSYLATED HEMOGLOBIN (GHb) (BASELINE) AND SEVERITY OF BASELINE RETINOPATHY (NUMBER AT RISK) -Glycosylated .-.. 2nd quartile -
Severity
1st quartile % (no. at risk)
at baseline
No (10110) Minimal levels (21<21-21/21) Mild nonproliferative (levels 31<31-31/31) Moderate-severe nonproliferative (levels 41<41-51/51) __ a p < 0.7; b p < 0.05, Mantel-Haenszel
~
-.
13 (15) 27 (22)
15 (20) 12 (17)
17 (18)
11 (19)
13 (16)
37 (19)
16 (19) ___
24 (17)
33 (27)
“p
_--
.--
45 (22y -.
test for linear trend.
(n = 74) Length of follow-up (mo) 25.9 (4.7)b LDL cholesterol (mg/dL) 113.2 (29.1) Fibrinogen (LOG) (g/L) 2.5 (0.1) GHb (%) 10.9 (2.2) Triglycerides (LOG) 2.0 (0.3)b (m&W Albumin excretion rate 1.4 (0.8)b (LOG) Progression to proliferative retinopathy (n = 33) Length of follow-up (mo) 26.2 (5.7)” LDL cholesterol (mg/dL) 121.3 (31.1)b Fibrinogen (LOG) (g/L) 2.5 (0.1) GHb (%) 11.6 (2.3) Triglycerides (LOG) 2.0 (0.3)b (m&W Albumin excretion rate 1.7 (1.0) (LOG) GHb,
4th quartile % (no. at risk)
12 (17) 20 (26)
Progression
lipoprotein;
3rd quartile % (no. at risk)
15 (13) 0 (29)
TABLE 4. BASELINE RISK FACTORS ASSOCIATED WITH THE INCIDENCE AND PROGRESSION OF RETINOPATHY OVER TWO YEARS-ANALYSIS OF VARIANCE, ADJUSTING FOR DURATION OF DIABETES -. Group Group with without progression progression Mean ( f SD) Mean ( f SD)
lozo-density
hemoglobin
% (no. at risk)
whether other risk factors differed according to baseline nephropathy status. As Table 5 shows, in subjects without overt nephropathy at baseline, GHb was significantly higher in those who progressed to proliferative retinopathy, as was LDL cholesterol and serum triglycerides. In those with overt nephropathy a different pattern was observed, with diastolic blood pressure and fibrinogen levels significantly greater in those who progressed to proliferative disease, and the absence of a significant difference in GHb and LDL cholesterol. Similar findings were observed for overall
LDL,
1995; 9:140-
glycosylated
(n = 248) 24.4 (4.4) 106.2 (28.9) 2.4 (0.1) 10.0 (1.6) 1.9 (0.2) 1.2 (0.7) (n = 289) 24.5 (4.4) 106.1 (28.6) 2.4 (0.1) 10.1 (1.6) 1.9 (0.2) 1.2 (0.7)
_
hemoglobin.
.-
progression of retinopathy (data not shown). Multiple logistic regression models were developed in order to identify independent risk factors for overall progression and progression to proliferative retinopathy (Table 6). All significant variables from the univariate analysis were entered into the stepwise models. For each analysis, GHb was first considered as a continuous variable and then as a categorical variable according to the distribution of GHb for those subjects. When used as a continuous variable, GHb did not appear to have a significant independent effect on progression of retinopathy or on progression to proliferative disease, although the models demonstrated an improved goodness of fit on inclusion of GHb. Therefore GHb was used as a categorical variable in further analysis, and divided into quartiles in the same way as in the univariate analysis. Regression analysis revealed that greater baseline severity of retinopathy, and GHb levels in the top quartile of the distribution, were significant independent risk factors for the overall progression of retinopathy. GHb values in the top quartile of the distribution were also a significant risk factor for progression to proliferative retinopathy, along with greater baseline severity. Factors such as fibrinogen and cholesterol (for both progression models) and LDL cholesterol and baseline nephropathy status (for progression to proliferative), although improving the goodness of fit of the models, were not shown to be significant independent risk factors. When multivariate analyses were repeated after excluding those with nephropathy, GHb in the top quartile remained a significant independent risk factor for overall progression (relative risk, 2.25; p = 0.020) and almost reached statistical significance for progression to proliferative retinopathy (Relative risk, 3.76; p = 0.056). In models where only those with overt nephropathy were included, GHb did not remain a significant risk factor, and diastolic blood pressure was shown to be of borderline significance for overall progression of retinopathy (relative risk, 1.18; ~7 = 0.07) and was a significant
risk
factor
for progression
to proliferative
] Diab Comp 1995;
RETINOPATHY
9:140-148
PROGRESSION
145
l *p
1 Relationship between nephropathy and development of retinopathy.
FIGURE
Progression
Progreesion m Microaibuminuria
No nephropathy
to proliferative
I
retinopathy
Overt nephropathy
disease (relative risk, 1.14; p = 0.03). Finally, when the interaction term between nephropathy and diastolic blood pressure was included in the overall model for progression to proliferative retinopathy, the interaction term was found to be a significant independent predictor (relative risk, 1.25; p = O.OOl), somewhat reducing the significance of GHb, but also diminishing the goodness of fit. DISCUSSION
This prospective study has attempted to identify potential risk factors for both the onset and the progression of diabetic retinopathy in a large cohort of individuals with IDDM. A major advantage of our study has been the good response rate with relatively few subjects lost to follow-up over a 2-year period. Our prevalence rates have been shown to be similar to or a little higher than those of other studies,28in particular those of WESDR, whose methods of ascertainment we have used. Our incidence rate of new cases of retinopathy is also similar to have of WESDR (66% versus 59%), however our rates of progression to proliferative retinopathy may be much higher (EDC, 20%-projected from 2-year data, versus WESDR, 10.5%-actual4year TABLE
5. RISK FACTORS
data), although this can only be compared crudely at present until our 4-year follow-up data is complete. In the past, the presence of retinopathy has been linked with various factors, and most often with duration of diabetes.“-% This has not always been the case, however,5 and we were only able to show an association between duration and progression to proliferative retinopathy, and this did not remain significant in the multivariate analysis. This may have been related to the relatively short follow-up period and/or to the sample size. If the progression of retinopathy follows a linear relationship with time or duration, however, duration may not be a major factor for progression per se (i.e., all subjects are progressing) although this will of course be strongly related to the severity of retinopathy. An alternative explanation is that progression of retinopathy may not follow a linear relationship, and a peak incidence rate may be reached by a certain duration.27 This has been suggested in the past, when it was demonstrated that incidence of any retinopathy was greatest at 15-19 years, after which the rate decreased.27 In our study, we also have some evidence of this phenomenon, with rates of progression, inci-
ASSOCIATED WITH PROGRESSION BY BASELINE NEPHROPATHY No overt nepluopatby Group with progression (n = 24) Mean ( f SD)
Length of follow-up (mo) LDL cholesterol (mg/dL) Fibrinogen (LOG) (g/L) GHb (%)
Triglycerides (LOG) (mg/dL) Diastolic blood pressure (mm Hg) LDL, low-density lipoprotein; GHb, glycosylated a p < 0.1, b p < 0.05, ’ p < 0.01, d p < 0.001,
TO PROLIFERATIVE STATUS
at baseline Group without progression (n = 259) Mean (*SD)
26.0 (6.2)
24.4 (4.5)
122.3 (31.0) 2.4 (0.1)
103.5 (26.3)
11.8 (2.3)d 2.0 (0.3)b 68.3 (9.5)
10.1 (1.6)
hemoglobin. adjusted for duration
2.4 (0.1) 1.9 (0.2) 69.8 (8.6) of diabetes.
RETINOPATHY,
Overt nepbropathy Group with progression (n = 30) Mean ( f SD) 26.6 118.7 2.6 11.0 2.1 88.9
(4.3) (33.4) (O.l)b (2.4) (0.3) (11.8)
at baseline Group without progression (n = 9) Mean (rt SD) 25.4 (3.1) 127.8 (37.7) 2.5 (0.1)
10.1 (1.7) 2.0 (0.2) 75.3 (11.0)
146
LLOYD
] Diab Camp 1995; 9: 140-148
ET AL.
TABLE 6. MULTIPLE REGRESSION ANALYSES FOR THE PROGRESSION OF RETINOPATHY OVER 2 YEARS S. Coeff. Progression Duration Length/follow-up (mo) Baseline severity GHb (top quartile)
SE
RR
95% C.I.
-1.49 0.02 0.99 0.94, 1.04 2.47 0.03 1.08 1.02, 1.15 2.96 0.14 1.53 1.15, 2.03 2.15 0.37 2.19 1.07, 4.50 Goodness of fit for model = 0.850
Progression to Proliferative Retinopathy Duration 0.74 0.04 1.03 0.94, 1.12 Baseline severity 5.16 0.35 5.99 3.03, 11.9 GHb (top quartile) 2.62 0.67 5.75 1.54, 21.4 Goodness of fit for model = 0.708
-
S. Coeff., stamiardizd CL, confidence interval;
coefficient; SE, standard error; RR, relative GHb, glycosylated hemoglobin.
For each model, all significant variables were entered in a stepwise fashion.
from
the univariate
risk;
analyses
dence rates and progression to proliferative retinopathy all reaching a peak between 15 and 19 years. Other factors have also been implicated in earlier studies, such as lipid levels, blood pressure, and blood Until recently, however, there glucose control. 6-9~11~12,B has been little prospective data to confirm this. More recently, a link has also been suggested between retinopathy and insulin dose.5 Our study shows that, although there were some univariate associations between lipid levels and the development of retinopathy, the more important factors, as shown in the multivariate analysis, were blood pressure and GHb. Although lipids did not remain in the multivariate model, this may have been related to their strong correlation with GHb, and thus should not be dismissed as risk factors for retinopathy. The relationship between hypertension and retinopathy has been reported previously, however, this has usually been in terms of progression rather than the incidence of retinopathy.“,‘2,3* Our findings suggest that even high-normal blood pressures may be important in the initialization of new cases of retmopathy. Chase et aL30 showed a similar association in a smaller 2-year follow-up study, when they reported a significant relationship between blood pressure and both the incidence and progression of retinopathy. These findings raise the possibility that the early manifestations of retinopathy (microaneurysms) may be a result of pressure effects on damaged vessels. The “later” manifestations of retinopathy (hemorrhage, leakage etc) may reflect a greater degree of glycemic damage to the vessels. Our findings thus underscore the importance of high-normal blood pressures even in the onset of retinopathy and the importance
of monitoring blood pressure levels in patients with even relatively short durations of IDDM. This conclusion is made with caution however, and, at this stage, it should not be assumed that lowering blood pressures in the high normal range will lead to a delay in the development of retinopathy. Although blood pressure may be an important risk factor for retinopathy, it has been argued elsewhere that it is the co-existence of nephropathy which is crucial.32 In this report, we have shown an association between progression of retinopathy to proliferative disease and blood pressure only in those with overt nephropathy. Interestingly, in those without nephropathy GHb was a more significant predictor and diastolic blood pressure was only marginally lower in those who progressed to proliferative disease. Whether or not nephiopathy is an independent risk factor for the development of retinopathy is not entirely clear. Previous cross-sectional studies have indicated that the presence of proteinuria significantly increases with increased severity of retinopathy.% Recent prospective data from the same study has confirmed this, showing that gross proteinuria remained a significant independent risk factor even after controlling for the effects of blood pressure.15 Although our prospective data have indicated that the existence of nephropathy at baseline may be associated with progression to proliferative retinopathy this was not the case for overall progression. Furthermore, the association between nephropathy and proliferative disease did not remain significant in the multivariate analysis. Other researchers have also shown similar rates of nephropathy in those with and without progression of retinopathy.12 In summary, therefore, it appears that the enhanced risk for proliferative disease seen in those with nephropathy is more closely linked to blood pressure rather than glycemic control, while the reverse is true for those without nephropathy. These comments are made with some caution, however, as the absence of a significant difference in GHb in those with nephropathy may be a result of small numbers in this analysis, further underscoring the need for even larger prospective studies, if we are to arrive at firmer conclusions. It is thus concluded that renal disease modifies the effect of blood pressure and glycemic control, and has relatively little independent effect on progression or proliferative retinopathy. Our results clearly show a significant independent effect of GHb on risk of both progression and progression to proliferative retinopathy, confirming the results of both the Diabetes Control and Complications Trial (DCCT) study and a recent report from Stockholm.4*5 Although our GHb levels appears higher than those of the DCCT and therefore not directly comparable, the DCCT study measured HbA1, levels which are lower than HbA1 (total hemoglobin Al). In fact after
1 Diab Comp
RETINOPATHY
1995; 9:140-148
adjustment for this, we have found no difference in the blood glucose levels of the DCCT and EDC studies. This study has demonstrated that the risk factors for the development and progression of diabetic retinopathy may differ at various stages of the disease. Blood pressure, even well within the normal range, may be important in the early stages of retinopathy, while glycemic control has been shown to play an important role in the later progression of this complication. The renal-retinal link is largely explained by the effects of blood pressure and glycemic control.
REFERENCES 1.
Lewitt M, Yue D: Macrovascular and microvascular disease in diabetes: An overview. World Book of Diabetes in Practice 3:278-282,
1988.
2. Orchard TJ, Dorman JS, Maser RE, Becker DJ, Drash AL, Ellis D, LaPorte RE, Kuller LH: Prevalenceof complications in IDDM by sex an duration. Pittsburgh Epidemiology of DiabetesComplications Study - II. Diabetes 39:1116-1124,1990. 3. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. II. Prevalence and risk of diabetic retinopathy when age at diagnosisislessthan 30years. Arch Opthalmol 102:520-526,1984.
PROGRESSION
147
factors for progression of background retinopathy in long-standing IDDM. Diabetes 38:460-464, 1989. 13. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: Is blood pressure a predictor of the incidence or progressionof diabetic retinopathy? Arch Intern Med 149: 2427-2432,1989. 14. Krolewski AS, Barzilay J, Warram JH, Martin BC, Pfeifer M, Rand LI: Riskof early-onsetproliferative retinopathy in IDDM is closely related to cardiovascular autonomic neuropathy. Diabetes 41:430-437,1992. 15. Klein R, Moss SE, Klein BEK: Gross proteinuria predicts the incidenceof proliferative diabetic retinopathy. Diabetes 4l(suppl 1):33A, 1992. 16. Orchard TJ, Dorman JS, Maser RE, Becker DJ, Ellis D, LaPorte RE, Kuller LH, Wolfson SK, DrashAL: Factors associatedwith the avoidance of severe complications after 25yr of IDDM. Pittsburgh Epidemiology of Diabetes Complications Study I. Diabetes Care 13:741-747, 1990.
17. Chavers BM, Mauer SM, RamsayRC, Steffes MW: Relationship between retinal and glomerular lesions in IDDM patients. Diabetes 43:441-446, 1994, 18. Borhani NO, KassEH, Langford HG, Payne GH, Reminton RD, Stamler J: The hypertension detection and follow-up program. Prev Med 5:207-215, 1976.
19. Allain CC, Poon LS, Chan CSG, Richmond W, Fu PC: Enzymatic-determinations of total serumcholesterol. 4. The DiabetesControl and ComplicationsTrial Research Clin Chem 20:470-475, 1974. Group: The effect of intensive treatment of diabeteson 20. Bucolo G, David H. Quantitative determination of sethe development and progressionof long-term complirum triglycerides by the useof enzymes. Clin Chem 19: cations in insulin-dependent diabetesmellitus. N Engl 476482, 1973. J Med 329:977-986,
1993.
21. Warnick GR, Albers JJ: Heparin-Mn2+ quantitation of high density lipoprotein cholesterol: an ultrafiltration procedure for lipemic samples.Clin Chem 24:900-904, 1978. 22. National Institutes of Health, Department of Health, 6. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: Education and Welfare: Lipid ResearchClinics ProGlycosylated hemoglobin predicts the incidence and gram. Washington, D.C., U.S. Government Printing progression of diabetic retinopathy. JAMA 260:2864Office, (NIH publication No. 75-628), 1975. 2871, 1988. 23. Friedewald WT, Levy RI, Frederickson DS: Estimation 7. Doft BH, Kingsley LA, Orchard TJ, Kuller LH, Drash of the concentration of low-density lipoprotein cholesAL, Becker DJ: The associationbetween long term diaterol in plasma,without useof the preparative ultracenbetic control and early retinopathy. Opthalmology 91: trifuge. Clin Chem 18:499-502,1972. l-7, 1984. 24. Moss SE, Meuer SM, Klein R, Hibbard LD, Brothers 8. McCance DR, Hadden DR, Atkinson AB, Archer DB: RJ, Klein BK. Are seven standard photographic fields Long-term glycaemic control and diabetic retinopathy. necessaryfor classificationof diabetic retinopathy? In5. Reichard P, N&son BY, Rosenqvist LJ: The effect of long-term intensified insulin treatment on the development of microvascular complications of diabetesmellitus. N Engl J Med 329:304-309, 1993.
Lancet ii:824-828,
1989.
9. Friberg TR, RosenstockJ, Sanborn G, Vaghefi A, Raskin P: The effect of long-term near normal glycemic control on mild diabetic retinopathy. Opthalmology 92:10511058, 1985.
vest Opthalmol
Vis Sci 30:823-828,
1989.
25. Early Treatment of Diabetic Retinopathy Study Coordinating Center Manual of Operations. Baltimore, MD, University of Maryland School of Medicine, Ch 5, 18. 1980. 10. Teuscher A, Schnell H, Wilson PWF: Incidence of dia- 26. Klein R, Klein BEK, Magli YL: An alternate method betic retinopathy and relationship to baselineplasma of grading diabetic retinopathy. Opthalmology 93:1183glucose and blood pressure. Diabetes Care 11:246-51, 1187, 1986. 1988. 27. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: 11. Bodansky HJ, Cudworth AG, Drury PL, Kohner EM: The Wisconsin Epidemiologic Study of Diabetic RetiRisk factors associatedwith severe proliferative retinopathy. IX. Four-year incidence and progression of nopathy in insulin-dependent diabetesmellitus. Diabediabetic retinopathy when age at diagnosisis lessthan tes Care 5:97-100, 1982. 30 years. Arch Opthalmol 107:237-243, 1989. 12. Janka HU, Warram JH, Rand LI, Krolewski AS: Risk 28. Lloyd CE, Orchard TJ:Insulin-dependent diabetesmel-
148
LLOYD
] DiubComp
ET AL.
1995; 9:740-248
litus in young people; the epidemiology of physical and psychosocial complications, in: Home PD, Alberti KGMM, Marshall SM, Krall L (eds), Diabetes Annual 7. Elsevier, Amsterdam, 1992.
31.
Agardh E, Torffvit 0, Agardh CD: The prevalence of retinopathy and associated medical risk factors in type 1 (insulin-dependent) diabetes mellitus. ] Intern Med 22647-52, 1989.
29.
Weber B, Burger W, Hartmann R, Hovener G, Malchus R, Oberdise U: Risk factors for the development of retinopathv in tvpe 1 (insulin-dependent) diabetes mellitus. Di~betol@ia29:!23-29,198k.
32.
Norgaard K, Feldt-Rasmussen B, Deckert T: Is hypertension a major independent risk factor for retinopathy in tvpe _* 1 diabetes? DiabeticMed 8:334-337, 1991.
33.
30.
Chase HP, Garg SK, Jackson WE, Thomas MA, Harris S, Marshall G, Crews MJ: Blood pressure and retinopathy in type 1 diabetes. Opthalmolqy 97:155-159, 1990.
KleinR, KleinBEK, Moss SE: A population-based study of diabetic retinopathy in insuiinlusing patients diagnosed before 30 years of age. DiabetesCare8(suppl 1): 71-76, 1985.
ABSTRACT This study examined potential risk factors for the incidence and progression of retinopathy in a large representative cohort of childhood onset insulin-dependent diabetic patients. Participants in the Epidemiology of Diabetes Complications (EDC) Study underwent a full clinical examination at baseline and again at a 2-year follow-up. Retinopathy status was ascertained using stereo fundus photographs graded according to the modified Airlie House System. The study population is based on a large cohort of childhood-onset insulin-dependent diabetic patients, seen within 1 year of diagnosis at the Children’s Hospital of Pittsburgh between January 1950 and May 1980. A total of 657 subjects participated at baseline (1986-19881, with 80% of eligible survivors taking part in the follow-up examination. This report concerns risk factors associated with the progression of diabetic retinopathy over a 2-year period, and the interaction of these factors with the presence of nephropathy. Analyses showed that baseline diastolic blood pressure was significantly _--
-
-..
___.
~
-___.
.~_
Department ofEpidemiology (C.E.L., L.H.K., T.J.O.), University of Pittsburgh, Pittsburgh, Pennsylvania; Department of Opthalmology (R.K.), University of Wisconsin Medical School, Wisconsin, Wisconsin; School of Life and Health Sciences (R.E.M.), University of Delaware, Newark, Delaware; Department of Pediatrics (D.J.B.), University of Pittsburgh, Pittsburgh, Pennsylvania, USA Reprint requests to be sent to: Dr. C. E. Lloyd, Department of Epidemiology, University of Pittsburgh, Diabetes Research Center, 5th Floor Rangos Building, 3460 Fifth Avenue, Pittsburgh, PA 15213.
~ourml of Diabetes and Its Complications 1995; 9:140-148 0 Elsevier Science Inc., 1995 655 Avenue of the Americas, New York, NY 10010
associated with the incidence of any retinopathy, while glycosylated hemoglobin, baseline severity of retinopathy, serum triglycerides, and, to a lesser extent, higher levels of low-density lipoprotein (LDL) cholesterol and fibrinogen were associated with the progression of retinopathy. Progression to proliferative retinopathy was related to higher LDL cholesterol, fibrinogen, serum triglycerides, albumin excretion rate, and glycosylated hemoglobin (GHb). Risk factors varied with the presence of nephropathy. In the absence of nephropathy, GHb was a significant predictor of progression, whilst this was not the case in the presence of nephropathy. Multivariate analyses demonstrated that the incidence of new cases of retinopathy was associated only with baseline diastolic blood pressure, while both the progression of retinopathy and progression to proliferative retinopathy were significantly associated with baseline glycosylated hemoglobin and baseline severity of retinopathy. These results suggest that the risk factors for the development of retinopathy may differ at various stages of the disease and according to the presence of nephropathy. While blood pressure, even within the normal range, may be important in the early stages of retinopathy, glycemic control remains the most important factor in the long-term development of this complication. (Journal of Diabetes and Its Complications 9;3:140-148, 1995.)
SSDI
10568727/951$9.50 1056-8727(94)00039-S
] Diab Comp
RETINOPATHY
1995; 9:140-148
INTRODUCTION
R
etinopathy, the most frequent complication of diabetes, is determined to be the leading cause of blindness in some countries.’ This complication, after 20 years, may come to affect virtually all those with insulin-dependent diabetes mellitus (IDDM).‘f3 Until recently, the available evidence linking potential risk factors to the development of diabetic retinopathy has come mainly from crosssectional studies, which have often been inconclusive. Furthermore, studies have not considered all possible risk factors, the Epidemiology of Diabetes Complications (EDC) study being the first to incorporate variables such as coagulation factors and diabetic autonomic neuropathy. However, recently two trials have shown clear evidence that intensive therapy regimens designed to normalize blood sugars as far as possible, dramatically retard (but do not eliminate) the development and progression of retinopathy.4,5 These trial data (which are needed to prove causality) are consistent with prior (epidemiologic and clinical) studies of the role of glycemic control as an important predictor of both early retinopathy changes, and the later progression of retinopathy.3,6-9 Other factors may be important, and raised lipid levels and higher blood pressures have been implicated.‘OI3 Early onset of proliferative retinopathy has also been linked to the presence of nephropathy,14,15 although it is unclear how this “renal-retinal” connection operates. The aim of this study was therefore to identify possible risk factors for the development and progression of retinopathy in a representative16 cohort of childhood-onset IDDM patients, and to examine if these risk factors are affected by the presence of renal disease. The latter objective will help resolve whether the “renal-retinal” link is mediated by other factors (i.e., renal disease may be acting as a confounder) or whether it modifies the effect of other risk factors (i.e., is an effect modifier) or whether it represents an independent predictor. A recent report stresses both a strong association between renal histological lesions and retinal structural lesions, and a fair degree of discordance of renal/retinal disease.17 METHODS The study participants were all childhood-onset (<17 years) IDDM patients seen at or within 1 year of diagnosis at the Children’s Hospital of Pittsburgh between January 1950 and May 1980, and living within 100 miles of Pittsburgh at the time of the study. All potential subjects were invited by letter to attend a clinical examination and to complete a series of questionnaires which were mailed to the participant 2 weeks prior to the examination. Of the 979 eligible, 67% participated
PROGRESSION
141
in a full evaluation. The representativeness of this group had been reported previously, as have full details of both the baseline questionnaires and clinical examinations that took place between 1986 and 1988.2,‘6 This report focuses on the incidence and progression of diabetic retinopathy based upon status at entry to the study (baseline) and at follow-up 2 years later. At the baseline clinical examination, after informed written consent, subjects underwent a sitting blood pressure measurement with a random zero sphygmomanometer (Hypertension Detection and Follow-up Program ProtocoP) with the mean of the second and third blood pressures used in the analyses. Blood samples were taken for measurement of lipids, lipoprotein, glycosylated hemoglobin (GHb), and fibrinogen. Albumin excretion rates (AER) were calculated using three timed urine samples; i.e., a 24-h collection, an overnight sample and one collected during the clinic. These three samples were usually collected within a 2-week period. Overt nephropathy was defined as the presence of renal failure (on dialysis and/or postkidney transplant) or an AER of greater than 200 1.181 min in at least two of three timed urine samples. Microalbuminuria was defined as an AER between 20 and 2001uglmin. Laboratory Determinations. For the first 18 months of the study, GHb (stable levels of total hemoglobin Ai) was determined with saline-incubated blood samples and microcolumn cation-exchange chromatography (Isolab, Akron, OH). Thereafter GHb was measured using automated high-performance liquid chromatography (Diamat, Bio-Rad, Hercules, CA). The two methods were shown to be almost identical (r = 0.95), with an absolute difference of 0.158 (%I-IbAI) . The normal range for GHb was 4.9%-7.3%. Serum cholesterol and triglyceride levels were determined enzymatitally. 1g,20 High-density lipoprotein cholesterol was measured by a heparin and manganese procedure with modification of the Lipid Research Clinics methodology. 21,22Low-density lipoprotein cholesterol was calculated with the Friedewald equation if triglycerides were less than 4.5 n&I.= Fibrinogen levels were measured using a biuret colometric procedure and a clotting method. Measurement of Diabetic Retinopathy. Stereo fundus photographs (eyes dilated) were taken of fields one, two, and four, with a Zeiss camera and were read in a masked fashion by the Fundus Photography Reading Center at the University of Wisconsin, Madison. Although three fields were used rather than the standard seven, it has been previously demonstrated that this gives very satisfactory results with good sensitivity. 24 Readings were classified according to the modified Airlie House System.25*26 Subjects were clas-
142
LLOYDETAL.
] Diab Comp
1. BASELINE
TABLE -
- .---.
STATUS OF ATTENDERS VERSUS NONATTENDERS FOR CYCLE TWO (FOLLOW-UP) ..--. - ..-
Did not attend follow-up
follow-up (n = 496)
MI,
myocardiul
infarction;
DCCT,
Diabetes
Control
and Complications
--
.--
9: l40- 148
ELIGIBLE
Attended
Age yr, mean (SD) Duration of diabetes yr, mean (SD) Males (%) Somecollege education (%) (for >18 yr olds) Overt nephropathy (96) (albumin excretion rate >200 &mm) Cardiovascular disease(%) (MT, angina, or stroke) Peripheral vascular disease(%) (Ankle-arm blood pressureratio <0.8, or amputation) Distal symmetrical neuropathy (%) (DCCT protocol) Hypertension (%) (>140/90 mm Hg and/or therapy) __ -~._-~ ~.~ _~- ~
7 9%;
(n = 121)
27.8 (7.9) 19.4 (7.5) 53 65 24 4
26.3 (7.8)* 18.7 (7.5) 40* 52* 31 2
9
8 22* 16
31 17
Trial.
* p < 0.05. sified into one of the following
categories based on the findings in the worst eye: no retinopathy (level lo), minimal background retinopathy (level 21), mild background retinopathy (level 31), moderate nonproliferative retinopathy (level 41), severe nonproliferative retinopathy (level 51), and proliferative retinopathy (level 60 or above). Both the incidence rates and progression of retinopathy were calculated in the same manner as the Wisconsin Epidemiologic Study of Diabetic Retinopathy (WESDR),27 as follows. Incidence of any retinopathy was calculated from all persons who had no retinopathy at baseline examination and who took part in the follow-up examination (i.e., moved from no retinopathy at baseline to levels 21/<21 or worse at follow-up). Progression of any retinopathy was defined as an increase in severity of two steps or more on the modified Airlie House system, and was calculated from all those with no or nonproliferative retinopathy at baseline. Progression to proliferative retinopathy was estimated from all subjects with no or nonproliferative retinopathy (i.e., classified at level 51/51 or better) at baseline who had deteriorated to proliferative retinopathy (level 60+ /<60 or worse) by follow-up. Statistical analyses were performed using the statistical package for the social sciences (SPSSX), and included Student’s t test, analysis of variance, and x2. Values given in tables are either means f SD or percent (%). Multiple logistic regression analyses were carried out using BMDP, where the goodness of fit of the specified models was determined. Using this statistical package it is possible to estimate the relative risk for each significant variable in order to illustrate the increase in risk of developing a particular end point
(in this case retinopathy) without the risk factor.
in those with versus those
RESULTS Response Rate. Of the 657 people who were examined at baseline, 27 had moved out of the area and were no longer eligible, and 13 had died, leaving 617 eligible, of whom 496 (80.4%) attended follow-up. Of those who returned, one-third already had had proliferative retinopathy at baseline and so were therefore ineligible for these analyses, leaving 322 (65%) at risk of the development or progression of retinopathy. Mean duration of diabetes at this time was 21.7 f 7.7 years, and the average interval between baseline and follow-up was 24.9 + 4.5 months. At follow-up, participants did not differ from nonparticipants in terms of income level but were more likely to have a college education (Table 1). A greater proportion of males returned than females. Although those who took part in both phases of the study were more likely to have had neuropathy at baseline, there were no differences in follow-up attendance in terms of other major complication at entry into the study. Progression of Diabetic Retinopathy. At the time of follow-up, a total of 23% (74/322) of subjects at risk of progression (i.e., without proliferative disease at baseline) had progressed by two steps or more, and 10% (33/322) had progressed to proliferative retinopathy. At follow-up examination, 20 subjects showed some improvement of retinopathy; these cases were analyzed together with those whose retinopathy had remained the same. A total of 12% (45/376) had received panretinal photocoagulation therapy between baseline and follow-up. Duration of diabetes was significantly longer for those who had progressed to PR (18.9 f 5.8 versus 16.5 k 6.6 years; p = 0.04). How-
1 Diab Comp
RETlNOPATHY
1995; 9:140-148
Incidence of Retinopathy. An entry into this study, only 69 subjects were still at risk of developing any new retinopathy, and one-third (23/69) had done so by follow-up, giving an incidence density of 16.6 cases/ 100 person years of follow-up. An examination of potential risk factors in these participants identified only baseline diastolic blood pressure as significantly different in those who developed new retinopathy compared to those who remained free of this complication (71.4 f. 8.8 versus 63.9 f 8.9 mm Hg; p < 0.001). Incidence of retinopathy was not associated with baseline GHb (10.2% in those who developed retinopathy versus 10.5% in those who did not). These findings were confirmed in multivariate analysis adjusting for both duration of diabetes and length of follow-up, when diastolic blood pressure was shown to be a significant independent risk factor for the development of new retinopathy (relative risk 1.15; 95% confidence interval, 1.05-1.26). Risk Factors Associated with Progression of Retinopathy. The remainder of this report focuses on the overall progression of retinopathy and on progression to proliferative disease. Subjects whose retinopathy had progressed by two steps or more at follow-up had a significantly higher mean GHb at baseline compared to those who had not progressed (10.9 versus 10.0%; p < 0.001). Subjects who experienced progression to proliferative retinopathy also had a significantly higher baseline GHb, even when duration of diabetes was accounted for (11.6 versus 10.1%; p < 0.001). In order to examine this relationship further, baseline GHb was
GHb
2. P-YEAR
(O/o)
Range
1st quartile 2nd quartile 3rd quartile
G9.0 9.1-9.9 10.0-11.1 >ll.l
4th quartile GHb,
glycosylated
PROGRESSION RATES NONPROLIFERATIVE
hemoglobin; RR, relative 0.001 Mantel-Haenszel
** p < 0.02, *** p <
BY QUARTILES OF BASELINE GHb, RETINOPATHY AT BASELINE Progression
No. at risk
%
RR
79 79 80 78
14 20 28 32
1.0 1.4 2.0 2.3
risk; CL, confidence test for trend.
143
divided into quartiles and progression of retinopathy and progression to proliferative retinopathy were examined. As Table 2 shows, there was a higher relative risk (RR) for both overall progression and progression to proliferative retinopathy for subjects with a GHb in the upper quartiles of the distribution. Severity of retinopathy at baseline was also significantly associated with the progression of retinopathy during follow-up. Table 3 shows the 2-year progression of retinopathy by quartile of GHb and baseline severity. As the table demonstrates, there was overall, a consistent increase in the proportion who progressed during follow-up according to both greater severity and the higher quartile of GHb. Tests for linear trend indicated that this increase was only statistically significant in those with moderate to severe nonproliferative retinopathy at baseline, but reached borderline significance for those with minimal level of severity at baseline. Other factors were also examined in relation to both the progression and progression to proliferative retinopathy during follow-up, with analyses adjusted for duration of diabetes (see Table 4). Higher fibrinogen levels and higher low-density lipoprotein (LDL) cholesterol were both associated with the overall progression of retinopathy, but these relationships were of borderline significance. Progression of retinopathy was also associated with higher serum triglyceride levels and a higher AER. Progression to proliferative retinopathy was significantly associated with higher serum triglycerides, higher LDL cholesterol, and somewhat higher fibrinogen levels (Table 4). Progression to proliferative retinopathy was also significantly related to baseline nephropathy status. As Figure 1 shows, a significantly greater proportion of those who had microalbuminuria or overt nephropathy at entry into the study progressed to proliferative disease over the 2-year follow-up period. There was no relationship between baseline nephropathy status and overall progression of retinopathy. The association between nephropathy and proliferative disease was further explored in order to assess
ever, duration was not significantly related to the overall progression of retinopathy over the 2-year followup. Progression to proliferative retinopathy was not significantly affected by length of follow-up interval. Overall progression of retinopathy was associated with length of follow-up, albeit a small difference (25.9 f 4.7 months compared with 24.4 f 4.4 months, p = 0.019, for those whose retinopathy did not progress). Overall progression and progression to proliferative retinopathy were both unrelated to gender.
TABLE
PROGRESSION
interval.
0.6, 4.0 1.0, 5.7 1.2, 7.0**
WITH
Progression to proliferative retinopathy _ .~~.
of retinopathy 95% C.I.
FOR PERSONS
%
RR
5 5 11 20
1.0 1.0 2.2 4.0***
95% C.I.
0.3, 3.9 0.7, 6.6 1.4, 11.4
14.4
TABLE
LLOYD
ET AL.
1 Diab Camp
iio’
3. 2-YEAR PROGRESSION (o/o) OF RETINOPATHY BY QUARTILES OF GLYCOSYLATED HEMOGLOBIN (GHb) (BASELINE) AND SEVERITY OF BASELINE RETINOPATHY (NUMBER AT RISK) -Glycosylated .-.. 2nd quartile -
Severity
1st quartile % (no. at risk)
at baseline
No (10110) Minimal levels (21<21-21/21) Mild nonproliferative (levels 31<31-31/31) Moderate-severe nonproliferative (levels 41<41-51/51) __ a p < 0.7; b p < 0.05, Mantel-Haenszel
~
-.
13 (15) 27 (22)
15 (20) 12 (17)
17 (18)
11 (19)
13 (16)
37 (19)
16 (19) ___
24 (17)
33 (27)
“p
_--
.--
45 (22y -.
test for linear trend.
(n = 74) Length of follow-up (mo) 25.9 (4.7)b LDL cholesterol (mg/dL) 113.2 (29.1) Fibrinogen (LOG) (g/L) 2.5 (0.1) GHb (%) 10.9 (2.2) Triglycerides (LOG) 2.0 (0.3)b (m&W Albumin excretion rate 1.4 (0.8)b (LOG) Progression to proliferative retinopathy (n = 33) Length of follow-up (mo) 26.2 (5.7)” LDL cholesterol (mg/dL) 121.3 (31.1)b Fibrinogen (LOG) (g/L) 2.5 (0.1) GHb (%) 11.6 (2.3) Triglycerides (LOG) 2.0 (0.3)b (m&W Albumin excretion rate 1.7 (1.0) (LOG) GHb,
4th quartile % (no. at risk)
12 (17) 20 (26)
Progression
lipoprotein;
3rd quartile % (no. at risk)
15 (13) 0 (29)
TABLE 4. BASELINE RISK FACTORS ASSOCIATED WITH THE INCIDENCE AND PROGRESSION OF RETINOPATHY OVER TWO YEARS-ANALYSIS OF VARIANCE, ADJUSTING FOR DURATION OF DIABETES -. Group Group with without progression progression Mean ( f SD) Mean ( f SD)
lozo-density
hemoglobin
% (no. at risk)
whether other risk factors differed according to baseline nephropathy status. As Table 5 shows, in subjects without overt nephropathy at baseline, GHb was significantly higher in those who progressed to proliferative retinopathy, as was LDL cholesterol and serum triglycerides. In those with overt nephropathy a different pattern was observed, with diastolic blood pressure and fibrinogen levels significantly greater in those who progressed to proliferative disease, and the absence of a significant difference in GHb and LDL cholesterol. Similar findings were observed for overall
LDL,
1995; 9:140-
glycosylated
(n = 248) 24.4 (4.4) 106.2 (28.9) 2.4 (0.1) 10.0 (1.6) 1.9 (0.2) 1.2 (0.7) (n = 289) 24.5 (4.4) 106.1 (28.6) 2.4 (0.1) 10.1 (1.6) 1.9 (0.2) 1.2 (0.7)
_
hemoglobin.
.-
progression of retinopathy (data not shown). Multiple logistic regression models were developed in order to identify independent risk factors for overall progression and progression to proliferative retinopathy (Table 6). All significant variables from the univariate analysis were entered into the stepwise models. For each analysis, GHb was first considered as a continuous variable and then as a categorical variable according to the distribution of GHb for those subjects. When used as a continuous variable, GHb did not appear to have a significant independent effect on progression of retinopathy or on progression to proliferative disease, although the models demonstrated an improved goodness of fit on inclusion of GHb. Therefore GHb was used as a categorical variable in further analysis, and divided into quartiles in the same way as in the univariate analysis. Regression analysis revealed that greater baseline severity of retinopathy, and GHb levels in the top quartile of the distribution, were significant independent risk factors for the overall progression of retinopathy. GHb values in the top quartile of the distribution were also a significant risk factor for progression to proliferative retinopathy, along with greater baseline severity. Factors such as fibrinogen and cholesterol (for both progression models) and LDL cholesterol and baseline nephropathy status (for progression to proliferative), although improving the goodness of fit of the models, were not shown to be significant independent risk factors. When multivariate analyses were repeated after excluding those with nephropathy, GHb in the top quartile remained a significant independent risk factor for overall progression (relative risk, 2.25; p = 0.020) and almost reached statistical significance for progression to proliferative retinopathy (Relative risk, 3.76; p = 0.056). In models where only those with overt nephropathy were included, GHb did not remain a significant risk factor, and diastolic blood pressure was shown to be of borderline significance for overall progression of retinopathy (relative risk, 1.18; ~7 = 0.07) and was a significant
risk
factor
for progression
to proliferative
] Diab Comp 1995;
RETINOPATHY
9:140-148
PROGRESSION
145
l *p
1 Relationship between nephropathy and development of retinopathy.
FIGURE
Progression
Progreesion m Microaibuminuria
No nephropathy
to proliferative
I
retinopathy
Overt nephropathy
disease (relative risk, 1.14; p = 0.03). Finally, when the interaction term between nephropathy and diastolic blood pressure was included in the overall model for progression to proliferative retinopathy, the interaction term was found to be a significant independent predictor (relative risk, 1.25; p = O.OOl), somewhat reducing the significance of GHb, but also diminishing the goodness of fit. DISCUSSION
This prospective study has attempted to identify potential risk factors for both the onset and the progression of diabetic retinopathy in a large cohort of individuals with IDDM. A major advantage of our study has been the good response rate with relatively few subjects lost to follow-up over a 2-year period. Our prevalence rates have been shown to be similar to or a little higher than those of other studies,28in particular those of WESDR, whose methods of ascertainment we have used. Our incidence rate of new cases of retinopathy is also similar to have of WESDR (66% versus 59%), however our rates of progression to proliferative retinopathy may be much higher (EDC, 20%-projected from 2-year data, versus WESDR, 10.5%-actual4year TABLE
5. RISK FACTORS
data), although this can only be compared crudely at present until our 4-year follow-up data is complete. In the past, the presence of retinopathy has been linked with various factors, and most often with duration of diabetes.“-% This has not always been the case, however,5 and we were only able to show an association between duration and progression to proliferative retinopathy, and this did not remain significant in the multivariate analysis. This may have been related to the relatively short follow-up period and/or to the sample size. If the progression of retinopathy follows a linear relationship with time or duration, however, duration may not be a major factor for progression per se (i.e., all subjects are progressing) although this will of course be strongly related to the severity of retinopathy. An alternative explanation is that progression of retinopathy may not follow a linear relationship, and a peak incidence rate may be reached by a certain duration.27 This has been suggested in the past, when it was demonstrated that incidence of any retinopathy was greatest at 15-19 years, after which the rate decreased.27 In our study, we also have some evidence of this phenomenon, with rates of progression, inci-
ASSOCIATED WITH PROGRESSION BY BASELINE NEPHROPATHY No overt nepluopatby Group with progression (n = 24) Mean ( f SD)
Length of follow-up (mo) LDL cholesterol (mg/dL) Fibrinogen (LOG) (g/L) GHb (%)
Triglycerides (LOG) (mg/dL) Diastolic blood pressure (mm Hg) LDL, low-density lipoprotein; GHb, glycosylated a p < 0.1, b p < 0.05, ’ p < 0.01, d p < 0.001,
TO PROLIFERATIVE STATUS
at baseline Group without progression (n = 259) Mean (*SD)
26.0 (6.2)
24.4 (4.5)
122.3 (31.0) 2.4 (0.1)
103.5 (26.3)
11.8 (2.3)d 2.0 (0.3)b 68.3 (9.5)
10.1 (1.6)
hemoglobin. adjusted for duration
2.4 (0.1) 1.9 (0.2) 69.8 (8.6) of diabetes.
RETINOPATHY,
Overt nepbropathy Group with progression (n = 30) Mean ( f SD) 26.6 118.7 2.6 11.0 2.1 88.9
(4.3) (33.4) (O.l)b (2.4) (0.3) (11.8)
at baseline Group without progression (n = 9) Mean (rt SD) 25.4 (3.1) 127.8 (37.7) 2.5 (0.1)
10.1 (1.7) 2.0 (0.2) 75.3 (11.0)
146
LLOYD
] Diab Camp 1995; 9: 140-148
ET AL.
TABLE 6. MULTIPLE REGRESSION ANALYSES FOR THE PROGRESSION OF RETINOPATHY OVER 2 YEARS S. Coeff. Progression Duration Length/follow-up (mo) Baseline severity GHb (top quartile)
SE
RR
95% C.I.
-1.49 0.02 0.99 0.94, 1.04 2.47 0.03 1.08 1.02, 1.15 2.96 0.14 1.53 1.15, 2.03 2.15 0.37 2.19 1.07, 4.50 Goodness of fit for model = 0.850
Progression to Proliferative Retinopathy Duration 0.74 0.04 1.03 0.94, 1.12 Baseline severity 5.16 0.35 5.99 3.03, 11.9 GHb (top quartile) 2.62 0.67 5.75 1.54, 21.4 Goodness of fit for model = 0.708
-
S. Coeff., stamiardizd CL, confidence interval;
coefficient; SE, standard error; RR, relative GHb, glycosylated hemoglobin.
For each model, all significant variables were entered in a stepwise fashion.
from
the univariate
risk;
analyses
dence rates and progression to proliferative retinopathy all reaching a peak between 15 and 19 years. Other factors have also been implicated in earlier studies, such as lipid levels, blood pressure, and blood Until recently, however, there glucose control. 6-9~11~12,B has been little prospective data to confirm this. More recently, a link has also been suggested between retinopathy and insulin dose.5 Our study shows that, although there were some univariate associations between lipid levels and the development of retinopathy, the more important factors, as shown in the multivariate analysis, were blood pressure and GHb. Although lipids did not remain in the multivariate model, this may have been related to their strong correlation with GHb, and thus should not be dismissed as risk factors for retinopathy. The relationship between hypertension and retinopathy has been reported previously, however, this has usually been in terms of progression rather than the incidence of retinopathy.“,‘2,3* Our findings suggest that even high-normal blood pressures may be important in the initialization of new cases of retmopathy. Chase et aL30 showed a similar association in a smaller 2-year follow-up study, when they reported a significant relationship between blood pressure and both the incidence and progression of retinopathy. These findings raise the possibility that the early manifestations of retinopathy (microaneurysms) may be a result of pressure effects on damaged vessels. The “later” manifestations of retinopathy (hemorrhage, leakage etc) may reflect a greater degree of glycemic damage to the vessels. Our findings thus underscore the importance of high-normal blood pressures even in the onset of retinopathy and the importance
of monitoring blood pressure levels in patients with even relatively short durations of IDDM. This conclusion is made with caution however, and, at this stage, it should not be assumed that lowering blood pressures in the high normal range will lead to a delay in the development of retinopathy. Although blood pressure may be an important risk factor for retinopathy, it has been argued elsewhere that it is the co-existence of nephropathy which is crucial.32 In this report, we have shown an association between progression of retinopathy to proliferative disease and blood pressure only in those with overt nephropathy. Interestingly, in those without nephropathy GHb was a more significant predictor and diastolic blood pressure was only marginally lower in those who progressed to proliferative disease. Whether or not nephiopathy is an independent risk factor for the development of retinopathy is not entirely clear. Previous cross-sectional studies have indicated that the presence of proteinuria significantly increases with increased severity of retinopathy.% Recent prospective data from the same study has confirmed this, showing that gross proteinuria remained a significant independent risk factor even after controlling for the effects of blood pressure.15 Although our prospective data have indicated that the existence of nephropathy at baseline may be associated with progression to proliferative retinopathy this was not the case for overall progression. Furthermore, the association between nephropathy and proliferative disease did not remain significant in the multivariate analysis. Other researchers have also shown similar rates of nephropathy in those with and without progression of retinopathy.12 In summary, therefore, it appears that the enhanced risk for proliferative disease seen in those with nephropathy is more closely linked to blood pressure rather than glycemic control, while the reverse is true for those without nephropathy. These comments are made with some caution, however, as the absence of a significant difference in GHb in those with nephropathy may be a result of small numbers in this analysis, further underscoring the need for even larger prospective studies, if we are to arrive at firmer conclusions. It is thus concluded that renal disease modifies the effect of blood pressure and glycemic control, and has relatively little independent effect on progression or proliferative retinopathy. Our results clearly show a significant independent effect of GHb on risk of both progression and progression to proliferative retinopathy, confirming the results of both the Diabetes Control and Complications Trial (DCCT) study and a recent report from Stockholm.4*5 Although our GHb levels appears higher than those of the DCCT and therefore not directly comparable, the DCCT study measured HbA1, levels which are lower than HbA1 (total hemoglobin Al). In fact after
1 Diab Comp
RETINOPATHY
1995; 9:140-148
adjustment for this, we have found no difference in the blood glucose levels of the DCCT and EDC studies. This study has demonstrated that the risk factors for the development and progression of diabetic retinopathy may differ at various stages of the disease. Blood pressure, even well within the normal range, may be important in the early stages of retinopathy, while glycemic control has been shown to play an important role in the later progression of this complication. The renal-retinal link is largely explained by the effects of blood pressure and glycemic control.
REFERENCES 1.
Lewitt M, Yue D: Macrovascular and microvascular disease in diabetes: An overview. World Book of Diabetes in Practice 3:278-282,
1988.
2. Orchard TJ, Dorman JS, Maser RE, Becker DJ, Drash AL, Ellis D, LaPorte RE, Kuller LH: Prevalenceof complications in IDDM by sex an duration. Pittsburgh Epidemiology of DiabetesComplications Study - II. Diabetes 39:1116-1124,1990. 3. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: The Wisconsin Epidemiologic Study of Diabetic Retinopathy. II. Prevalence and risk of diabetic retinopathy when age at diagnosisislessthan 30years. Arch Opthalmol 102:520-526,1984.
PROGRESSION
147
factors for progression of background retinopathy in long-standing IDDM. Diabetes 38:460-464, 1989. 13. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: Is blood pressure a predictor of the incidence or progressionof diabetic retinopathy? Arch Intern Med 149: 2427-2432,1989. 14. Krolewski AS, Barzilay J, Warram JH, Martin BC, Pfeifer M, Rand LI: Riskof early-onsetproliferative retinopathy in IDDM is closely related to cardiovascular autonomic neuropathy. Diabetes 41:430-437,1992. 15. Klein R, Moss SE, Klein BEK: Gross proteinuria predicts the incidenceof proliferative diabetic retinopathy. Diabetes 4l(suppl 1):33A, 1992. 16. Orchard TJ, Dorman JS, Maser RE, Becker DJ, Ellis D, LaPorte RE, Kuller LH, Wolfson SK, DrashAL: Factors associatedwith the avoidance of severe complications after 25yr of IDDM. Pittsburgh Epidemiology of Diabetes Complications Study I. Diabetes Care 13:741-747, 1990.
17. Chavers BM, Mauer SM, RamsayRC, Steffes MW: Relationship between retinal and glomerular lesions in IDDM patients. Diabetes 43:441-446, 1994, 18. Borhani NO, KassEH, Langford HG, Payne GH, Reminton RD, Stamler J: The hypertension detection and follow-up program. Prev Med 5:207-215, 1976.
19. Allain CC, Poon LS, Chan CSG, Richmond W, Fu PC: Enzymatic-determinations of total serumcholesterol. 4. The DiabetesControl and ComplicationsTrial Research Clin Chem 20:470-475, 1974. Group: The effect of intensive treatment of diabeteson 20. Bucolo G, David H. Quantitative determination of sethe development and progressionof long-term complirum triglycerides by the useof enzymes. Clin Chem 19: cations in insulin-dependent diabetesmellitus. N Engl 476482, 1973. J Med 329:977-986,
1993.
21. Warnick GR, Albers JJ: Heparin-Mn2+ quantitation of high density lipoprotein cholesterol: an ultrafiltration procedure for lipemic samples.Clin Chem 24:900-904, 1978. 22. National Institutes of Health, Department of Health, 6. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: Education and Welfare: Lipid ResearchClinics ProGlycosylated hemoglobin predicts the incidence and gram. Washington, D.C., U.S. Government Printing progression of diabetic retinopathy. JAMA 260:2864Office, (NIH publication No. 75-628), 1975. 2871, 1988. 23. Friedewald WT, Levy RI, Frederickson DS: Estimation 7. Doft BH, Kingsley LA, Orchard TJ, Kuller LH, Drash of the concentration of low-density lipoprotein cholesAL, Becker DJ: The associationbetween long term diaterol in plasma,without useof the preparative ultracenbetic control and early retinopathy. Opthalmology 91: trifuge. Clin Chem 18:499-502,1972. l-7, 1984. 24. Moss SE, Meuer SM, Klein R, Hibbard LD, Brothers 8. McCance DR, Hadden DR, Atkinson AB, Archer DB: RJ, Klein BK. Are seven standard photographic fields Long-term glycaemic control and diabetic retinopathy. necessaryfor classificationof diabetic retinopathy? In5. Reichard P, N&son BY, Rosenqvist LJ: The effect of long-term intensified insulin treatment on the development of microvascular complications of diabetesmellitus. N Engl J Med 329:304-309, 1993.
Lancet ii:824-828,
1989.
9. Friberg TR, RosenstockJ, Sanborn G, Vaghefi A, Raskin P: The effect of long-term near normal glycemic control on mild diabetic retinopathy. Opthalmology 92:10511058, 1985.
vest Opthalmol
Vis Sci 30:823-828,
1989.
25. Early Treatment of Diabetic Retinopathy Study Coordinating Center Manual of Operations. Baltimore, MD, University of Maryland School of Medicine, Ch 5, 18. 1980. 10. Teuscher A, Schnell H, Wilson PWF: Incidence of dia- 26. Klein R, Klein BEK, Magli YL: An alternate method betic retinopathy and relationship to baselineplasma of grading diabetic retinopathy. Opthalmology 93:1183glucose and blood pressure. Diabetes Care 11:246-51, 1187, 1986. 1988. 27. Klein R, Klein BEK, Moss SE, Davis MD, DeMets DL: 11. Bodansky HJ, Cudworth AG, Drury PL, Kohner EM: The Wisconsin Epidemiologic Study of Diabetic RetiRisk factors associatedwith severe proliferative retinopathy. IX. Four-year incidence and progression of nopathy in insulin-dependent diabetesmellitus. Diabediabetic retinopathy when age at diagnosisis lessthan tes Care 5:97-100, 1982. 30 years. Arch Opthalmol 107:237-243, 1989. 12. Janka HU, Warram JH, Rand LI, Krolewski AS: Risk 28. Lloyd CE, Orchard TJ:Insulin-dependent diabetesmel-
148
LLOYD
] DiubComp
ET AL.
1995; 9:740-248
litus in young people; the epidemiology of physical and psychosocial complications, in: Home PD, Alberti KGMM, Marshall SM, Krall L (eds), Diabetes Annual 7. Elsevier, Amsterdam, 1992.
31.
Agardh E, Torffvit 0, Agardh CD: The prevalence of retinopathy and associated medical risk factors in type 1 (insulin-dependent) diabetes mellitus. ] Intern Med 22647-52, 1989.
29.
Weber B, Burger W, Hartmann R, Hovener G, Malchus R, Oberdise U: Risk factors for the development of retinopathv in tvpe 1 (insulin-dependent) diabetes mellitus. Di~betol@ia29:!23-29,198k.
32.
Norgaard K, Feldt-Rasmussen B, Deckert T: Is hypertension a major independent risk factor for retinopathy in tvpe _* 1 diabetes? DiabeticMed 8:334-337, 1991.
33.
30.
Chase HP, Garg SK, Jackson WE, Thomas MA, Harris S, Marshall G, Crews MJ: Blood pressure and retinopathy in type 1 diabetes. Opthalmolqy 97:155-159, 1990.
KleinR, KleinBEK, Moss SE: A population-based study of diabetic retinopathy in insuiinlusing patients diagnosed before 30 years of age. DiabetesCare8(suppl 1): 71-76, 1985.