The new england journal of medicine

original article

Atorvastatin in Patients with Type 2 Diabetes

Mellitus Undergoing Hemodialysis
Christoph Wanner, M.D., Vera Krane, M.D., Winfried März, M.D.,
Manfred Olschewski, M.Sc., Johannes F.E. Mann, M.D., Günther Ruf, M.D.,
and Eberhard Ritz, M.D., for the German Diabetes and Dialysis Study Investigators*

abstract

background
From the Department of Medicine, Divi- Statins reduce the incidence of cardiovascular events in persons with type 2 diabetes
sion of Nephrology, University of Würz- mellitus. However, the benefit of statins in such patients receiving hemodialysis, who
burg, Würzburg, Germany (C.W., V.K.); the
Clinical Institute of Medical and Chemical are at high risk for cardiovascular disease and death, has not been examined.
Laboratory Diagnostics, University Gen-
eral Hospital, Graz, Austria (W.M.); the methods
Department of Medical Biometrics and
Statistics, University Hospital of Freiburg, We conducted a multicenter, randomized, double-blind, prospective study of 1255
Freiburg, Germany (M.O.); Schwabing Gen- subjects with type 2 diabetes mellitus receiving maintenance hemodialysis who were
eral Hospital, Munich, Germany (J.F.E.M.); randomly assigned to receive 20 mg of atorvastatin per day or matching placebo. The
Clinical Research Department, Pfizer, Karls-
ruhe, Germany (G.R.); and the Depart- primary end point was a composite of death from cardiac causes, nonfatal myocardial
ment of Medicine, University of Heidel- infarction, and stroke. Secondary end points included death from all causes and all
berg, Heidelberg, Germany (E.R.). Address cardiac and cerebrovascular events combined.
reprint requests to Dr. Wanner at the De-
partment of Medicine, Division of Neph-
rology, University Hospital, Josef-Schneider- results
Str. 2, D-97080 Würzburg, Germany, or at After four weeks of treatment, the median level of low-density lipoprotein cholesterol
wanner_c@medizin.uni-wuerzburg.de.
was reduced by 42 percent among patients receiving atorvastatin, and among those
*Investigators and research coordinators receiving placebo it was reduced by 1.3 percent. During a median follow-up period of
participating in this study are listed in four years, 469 patients (37 percent) reached the primary end point, of whom 226 were
the Appendix.
assigned to atorvastatin and 243 to placebo (relative risk, 0.92; 95 percent confidence
N Engl J Med 2005;353:238-48. interval, 0.77 to 1.10; P=0.37). Atorvastatin had no significant effect on the individual
Copyright © 2005 Massachusetts Medical Society.
components of the primary end point, except that the relative risk of fatal stroke among
those receiving the drug was 2.03 (95 percent confidence interval, 1.05 to 3.93;
P=0.04). Atorvastatin reduced the rate of all cardiac events combined (relative risk,
0.82; 95 percent confidence interval, 0.68 to 0.99; P=0.03, nominally significant) but
not all cerebrovascular events combined (relative risk, 1.12; 95 percent confidence
interval, 0.81 to 1.55; P=0.49) or total mortality (relative risk, 0.93; 95 percent con-
fidence interval, 0.79 to 1.08; P=0.33).

conclusions
Atorvastatin had no statistically significant effect on the composite primary end point
of cardiovascular death, nonfatal myocardial infarction, and stroke in patients with
diabetes receiving hemodialysis.

238 n engl j med 353;3 www.nejm.org july 21, 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

matopoietic disease or systemic disease unrelated

p rimary and secondary prevention

trials, including those involving persons
with diabetes mellitus, have documented
substantial cardiovascular benefit from the admin-
istration of statins.1,2 The recent Collaborative
to end-stage renal disease; vascular intervention,
congestive heart failure, or myocardial infarction
within the three months preceding the period of en-
rollment; unsuccessful kidney transplantation; and
Atorvastatin Diabetes Study (CARDS) reported a de- hypertension resistant to therapy (i.e., systolic blood
crease in deaths from cardiovascular causes among pressure continuously greater than 200 mm Hg or
persons with type 2 diabetes mellitus in the ab- diastolic blood pressure greater than 110 mm Hg).
sence of marked renal insufficiency.3 There are no On enrollment, lipid-lowering medications were
prospective data on the effects of statins in patients discontinued, and patients received placebo during
with end-stage renal disease with type 2 diabetes the four-week run-in phase of the study. Thereafter,
mellitus who are receiving hemodialysis, although eligible patients were randomly assigned to double-
type 2 diabetes is the most common diagnosis blind treatment with either atorvastatin at a dose of
among patients at excessive risk of cardiovascular 20 mg once daily or matching placebo. Data were
events4 whose condition requires hemodialysis in recorded at four weeks and then every six months.
both Germany5 and the United States.6 Abnormali- The protocol was approved by the ethics committee
ties in serum lipid levels that are associated with re- at the coordinating center and the 29 regional insti-
nal disease rank high among the factors implicated tutional review boards. Specifically, the ethical im-
in accelerated atherosclerosis.7 However, not all plications of the inclusion of a placebo group —
the observational data on patients receiving hemo- that is, of not providing lipid-lowering medications
dialysis link dyslipidemia with reduced rates of to those randomly assigned to the control group —
survival; indeed, opposite trends have been noted.8 were taken into account and considered accept-
An observational retrospective analysis of patients able. Written informed consent was obtained from
receiving hemodialysis, the U.S. Renal Data System all patients.
Morbidity and Mortality Study, Wave 2,9 reported Academic investigators led, managed, and co-
that the risk of death from cardiovascular causes ordinated the study. The principal investigators
was decreased by 36 percent among patients re- wrote the protocol and prepared the manuscript.
ceiving statins, as compared with those who did The data were monitored and collected by two con-
not receive statins. There has been concern about tract research organizations supported by Pfizer,
the side effects of statins in patients receiving he- one of which (Datamap) holds the data. A university-
modialysis,10 but data from small cohorts appeared based, independent statistician performed the sta-
to be reassuring.11 The present investigator-initiat- tistical analyses. The plan for the statistical analysis
ed, prospective, randomized, placebo-controlled was completed before the database was locked and
study of patients with type 2 diabetes mellitus re- unblinded.
ceiving hemodialysis was designed to answer these A computer-generated randomization code was
questions. prepared by a central Pfizer unit that was indepen-
dent of local study personnel. Medication was pre-
methods packaged on the basis of a block size of four sub-
jects at each center. Each consecutive subject was
study design given the next consecutive randomization number,
Subjects with type 2 diabetes mellitus 18 to 80 years and eligible patients were assigned in a 1:1 ratio to
of age who had been receiving maintenance hemo- receive the study drug or placebo. Lipid levels mea-
dialysis for less than two years were enrolled at 178 sured after randomization were not released to the
centers in Germany. Exclusion criteria included lev- clinical sites. If LDL cholesterol levels fell below
els of fasting serum low-density lipoprotein (LDL) 50 mg per deciliter (1.3 mmol per liter), the dose of
cholesterol of less than 80 mg per deciliter (2.1 atorvastatin was reduced to 10 mg per day. To main-
mmol per liter) or more than 190 mg per deciliter tain blinding, a randomly selected subject from the
(4.9 mmol per liter), triglyceride levels greater than placebo group received an identical dose reduction.
1000 mg per deciliter (11.3 mmol per liter); liver- One person in the central laboratory who had access
function values more than three times the upper to the randomization code controlled the changes
limit of normal or equal to those in patients with in dose. After a patient reached a primary end point,
symptomatic hepatobiliary cholestatic disease; he- the study drug could be replaced by treatment with

n engl j med 353;3 www.nejm.org july 21, 2005 239

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

an active statin. Details of the study design have Secondary end points included death from all
been described previously.12,13 causes, all cardiac events combined, and all cere-
brovascular events combined. Death from any cause
end points other than cardiac disease or cerebrovascular dis-
The study end points and serious adverse events ease was treated as a competing risk.
were continuously monitored and reported to the A central laboratory performed all the analyses.
contract research organization. Every end point was LDL cholesterol was measured directly by agarose-
adjudicated by three members of the end-point gel electrophoresis with subsequent enzymatic
committee, on the basis of predefined criteria that staining for cholesterol with the use of the rapid
are part of the study protocol. All analyses of pri- electrophoresis system (Helena Diagnostika). This
mary and secondary end points were based on the method produces more accurate measurements of
classification by the end-point committee that was LDL cholesterol than ultracentrifugation and pre-
agreed on by consensus or majority vote. All com- cipitation combined in samples with elevated tri-
mittee members were blinded to the treatment as- glyceride concentrations.14
signments until August 13, 2004. The primary end
point was a composite of death from cardiac causes, statistical analysis
fatal stroke, nonfatal myocardial infarction, or non- The study was designed to have 90 percent power
fatal stroke, whichever occurred first. Only one to detect a 27 percent reduction in the incidence of
event per subject was included in the analysis. Myo- the composite primary end point at an alpha level
cardial infarction was diagnosed when two of the of 0.05 in a two-sided test, adjusted for one pre-
following three criteria were met: typical symptoms; planned interim analysis according to an alpha-
elevated levels of cardiac enzymes (i.e., a level of spending function based on the O’Brien–Fleming
creatine kinase MB above 5 percent of the total level method, yielding a nominal level of significance for
of creatine kinase, a level of lactic dehydrogenase the final analysis of 0.045.15 The alpha-spending
1.5 times the upper limit of normal, or a level of function would have allowed for additional interim
troponin T greater than 2 ng per milliliter); or diag- analyses, if necessary. For the study to have this level
nostic changes on the electrocardiogram. A resting of power, at least 424 primary end points had to oc-
electrocardiogram was recorded every six months cur (event-driven analysis), requiring the random-
and evaluated by independent cardiologists from ization of at least 1200 patients. This calculation was
the electrocardiographic monitoring board, accord- based on observational studies.16,17 The results were
ing to the Minnesota classification system for the assessed in an intention-to-treat analysis. The pri-
electrocardiogram (codes 1-1-1 through 9-2 for mary end points were evaluated according to time-
QRS-complex, ST-segment, or T-wave changes). to-event analysis. Death from other causes was treat-
An electrocardiogram that documented silent my- ed as a competing event, and for patients who died
ocardial infarction was considered evidence of a from other causes, follow-up was censored as of the
primary end point. date of death.18 Times to an event for patients with-
Stroke was defined as a neurologic deficit last- out a primary end point or competing event were
ing longer than 24 hours. Computed tomographic treated as censored and were calculated as the time
or magnetic resonance imaging of the brain was from randomization to the date of the last contact.
recommended and available in all but 16 cases. Cumulative incidence and Kaplan–Meier curves
Death from cardiac causes comprised fatal myocar- were used only to show the survival curves within the
dial infarction (death within 28 days after a myo- treatment groups and to calculate the correspond-
cardial infarction), sudden death, death due to con- ing survival probabilities. The Cox proportional-
gestive heart failure, death due to coronary heart hazards model was used to estimate the multivari-
disease during or within 28 days after an interven- ate relative risks of the primary and secondary end
tion, and all other deaths ascribed to coronary heart points with corresponding 95 percent confidence
disease. Patients who died unexpectedly and did intervals. Adjustments were made for sex, age, and
not present with a potassium level greater than baseline status with respect to coronary heart dis-
7.5 mmol per liter before the start of the three most ease. Unless otherwise stated, the baseline lipid and
recent sessions of hemodialysis were considered to safety laboratory value was defined as the last value
have had sudden death from cardiac causes. measured during the run-in period. The baseline

240 n engl j med 353;3 www.nejm.org july 21 , 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

data were analyzed with the use of standard de- October 2002 and were followed until their final
scriptive statistics. visit in March 2004 (Fig. 1). The two groups of pa-
tients were well matched with respect to baseline
results characteristics and concomitant therapy (Table 1).
Nineteen percent of the patients had taken statins
patients before entering the study. The mean length of fol-
A total of 1255 subjects were randomly assigned to low-up was 3.96 years in the atorvastatin group and
double-blind treatment with either atorvastatin 3.91 years in the placebo group (median, 4.0 and
(619) or placebo (636) between March 1998 and 4.08 years, respectively).

Patients entering
run-in phase (n=1522)

Patients with lipid values outside

required ranges (n=207)
Excluded on the basis of other
criteria (n=12)
Excluded for other reasons (n=29)
Withdrew consent (n=12)
Medical event required
withdrawal (n=7)

Patients randomly assigned

to treatment (n=1255)

Placebo group (n=636) Atorvastatin group (n=619)

Did not receive study drug (n=1) Did not receive study drug (n=0)

Lost to follow-up (n=1) Lost to follow-up (n=0)

Discontinued treatment before Discontinued treatment before

end of study (n=150) end of study (n=142)
Wish of patient (n=66) Wish of patient (n=60)
On request of investigator On request of investigator
for medical reason (n=32) for medical reason (n=23)
Administrative reason (n=12) Administrative reason (n=12)
Other reason (n=39) Other reason (n=47)
No reason given (n=1)
Completed treatment according
Completed treatment according to study protocol (n=477)
to study protocol (n=484)

Included in intention-to-treat Included in intention-to-treat

analysis (n=636) analysis (n=619)

Figure 1. Numbers of Patients Who Entered the Study, Were Assigned to a Study Group, and Completed the Protocol.

n engl j med 353;3 www.nejm.org july 21, 2005 241

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

Table 1. Baseline Characteristics of Patients in the Placebo and Atorvastatin Groups.*

Characteristic Placebo Group (N=636) Atorvastatin Group (N=619)

Age — yr 65.7±8.3 65.7±8.3

Female sex — no. (%) 292 (45.9) 286 (46.2)
Known duration of diabetes — yr 18.7±8.8 17.5±8.7
Time receiving dialysis — mo 8.4±6.9 8.2±6.9
Blood pressure — mm Hg
Systolic 145±22 146±22
Diastolic 76±11 76±11
Current smoker — no. (%) 58 (9.1) 50 (8.1)
Former smoker — no. (%) 188 (29.6) 211 (34.1)
History of cardiovascular disease and intervention (%)†
Myocardial infarction 17.3 17.9
Myocardial infarction, either CABG or PTCA, 28.1 30.7
or coronary heart disease‡
Myocardial infarction or either CABG or PTCA 22.5 23.7
CABG or PTCA 11.8 14.2
Congestive heart failure§ 34.9 35.9
Cardiac-valve disorder 7.7 7.3
Peripheral vascular disease 43.6 45.7
Stroke or TIA 18.2 17.4
Body-mass index¶ 27.5±5.0 27.6±4.6
Hemoglobin — g/dl 10.9±1.4 10.9±1.3
Glycosylated hemoglobin — % 6.8±1.3 6.7±1.2
Albumin — g/liter 3.8±0.3 3.8±0.3
Calcium — mg/dl 9.2±0.8 9.2±0.8
Phosphate — mg/dl 6.1±1.6 6.0±1.6

lipid levels three years in the atorvastatin group, as compared

At randomization, the median level of LDL choles- with 11.2 percent and 30.5 percent, respectively,
in the placebo group (Fig. 3). The relative risk re-
terol was 121 mg per deciliter (3.13 mmol per liter)
in the atorvastatin group and 125 mg per deciliterduction afforded by active treatment, as compared
(3.23 mmol per liter) in the placebo group. After with placebo, was 8 percent (hazard ratio, 0.92; 95
four weeks, in the atorvastatin group, the median percent confidence interval, 0.77 to 1.10; P=0.37).
level of LDL cholesterol was 72 mg per deciliter A similar number of patients died from cardiac
(1.86 mmol per liter; median change from baseline,causes in the two groups (20 percent in the atorva-
¡42 percent). In the placebo group, the level of statin group and 23 percent in the placebo group;
LDL cholesterol remained essentially unchanged relative risk, 0.81; 95 percent confidence interval,
0.64 to 1.03; P=0.08). Eleven percent (70) of the
(120 mg per deciliter [3.10 mmol per liter]; median
change from baseline, ¡1.3 percent) (Fig. 2). patients in the atorvastatin group had a nonfatal
myocardial infarction, as compared with 12 percent
primary outcomes (79) of those in the placebo group (relative risk,
The cumulative incidence of the primary end point 0.88; 95 percent confidence interval, 0.64 to 1.21;
was 12.6 percent at one year and 31.9 percent at P=0.42). More patients (27) died of stroke in the

242 n engl j med 353;3 www.nejm.org july 21 , 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

Table 1. (Continued.)

Characteristic Placebo Group (N=636) Atorvastatin Group (N=619)

Lipid values — mg/dl

Total cholesterol 220±42 218±43
LDL cholesterol 127±30 125±29
HDL cholesterol 36±14 36±13
Triglycerides 267±168 261±165
LDL cholesterol levels — no. (%)
<100 120 (18.9) 122 (19.7)
100–129 241 (37.9) 252 (40.7)
130–159 186 (29.2) 169 (27.3)
≥160 89 (14.0) 76 (12.3)
Antihypertensive medication — %
ACE inhibitors 47 49
Angiotensin II–receptor antagonists 12 12
Beta-blockers 38 37
Calcium antagonists 40 41
Antiplatelet therapy 50 54
Use of erythropoietin — % 81 81
Dose per wk — IU 6.225 6.202

* Plus–minus values are means ±SD. To convert hemoglobin values to millimoles per liter, multiply by 0.6206. To convert
values for calcium to millimoles per liter, multiply by 0.250. To convert values for phosphate to millimoles per liter, mul-
tiply by 0.3229. To convert values for total cholesterol, low-density lipoprotein (LDL) cholesterol, and high-density lipo-
protein (HDL) cholesterol to millimoles per liter, multiply by 0.02586. To convert values for triglycerides to millimoles per
liter, multiply by 0.01129. CABG denotes coronary-artery bypass grafting, PTCA percutaneous transluminal coronary an-
gioplasty, TIA transient ischemic attack, and ACE angiotensin-converting enzyme.
† Types of disease and intervention are not mutually exclusive.
‡ Disease was documented by coronary angiography.
§ Most of the patients had New York Heart Association class II heart failure.
¶ The body-mass index is the weight in kilograms divided by the square of the height in meters.

atorvastatin group than in the placebo group (13; 33 percent, as compared with 39 percent in the pla-
relative risk, 2.03; 95 percent confidence interval, cebo group (relative risk, 0.82; 95 percent confi-
1.05 to 3.93; P=0.04). Nonfatal stroke was distrib- dence interval, 0.68 to 0.99; P=0.03) (Table 2). This
uted equally in the two groups (33 patients in the result was driven mainly by differences in the rates
atorvastatin group and 32 patients in the placebo of coronary-artery bypass grafting and percutane-
group; relative risk, 1.04; 95 percent confidence in- ous transluminal coronary angioplasty. The inci-
terval, 0.64 to 1.69; P=0.89) (Table 2). dence of all cerebrovascular events combined in
the atorvastatin group was not different from that
secondary outcomes in the placebo group (relative risk, 1.12; 95 percent
Death from all causes was similar in the two groups confidence interval, 0.81 to 1.55; P=0.49) (Table 2).
(48 percent in the atorvastatin group, as compared
with 50 percent in the placebo group; relative risk, adherence, tolerability,
0.93; 95 percent confidence interval, 0.79 to 1.08; and adverse events
P=0.33). Of nominal significance, the risk of all The mean (±SD) duration of exposure to placebo
cardiac events combined was reduced by 18 percent was 27.2±17.9 months (range, 0.03 to 70.2), and
in the atorvastatin group, with a total event rate of to atorvastatin, 28.5±18.6 months (range, 0.07 to

n engl j med 353;3 www.nejm.org july 21, 2005 243

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

130

120

110

100 Placebo

Median LDL Cholesterol (mg/dl)

90

80

70
Atorvastatin
60

50

40

30

20

10

0
Baseline 6 12 18 24 30 36 42 48 54 60

Month
No. at Risk
Placebo 636 611 544 493 427 327 264 208 147 105 60 37
Atorvastatin 619 597 539 484 413 343 279 218 157 117 74 44

Figure 2. Median Level of Low-Density Lipoprotein (LDL) Cholesterol from Baseline to the End of the Study.
To convert values for LDL cholesterol to millimoles per liter, multiply by 0.02586.

69.9). In the placebo group, 82 percent of patients

60
took the study medication without interruption, and
Cumulative Incidence of the Primary

Placebo
50 in the atorvastatin group, 80 percent of patients did
Composite End Point (%)

so. The average number of days that treatment was

40 interrupted was 12±36 in the placebo group and
Atorvastatin
13±40 in the atorvastatin group. During treatment,
30
the dose of atorvastatin or matching placebo was
20 halved when administered to 190 patients (15 per-
cent). During the study, 98 patients in the placebo
10
group (15 percent) began nonstudy statins, as com-
0 pared with 10 percent of those in the atorvastatin
0 1 2 3 4 5 6 group. The proportion of patients who continued
Year to receive the study drug at one and two years, ex-
No. at Risk pressed as a percentage of those who remained alive
Placebo 636 532 383 252 136 51 19 and free of a primary event, was 74 percent (459 pa-
Atorvastatin 619 515 378 252 136 58 29
tients) and 51 percent (317 patients), respectively,
Figure 3. Estimated Cumulative Incidence of the Composite Primary in the atorvastatin group and 74 percent (469 pa-
End Point. tients) and 48 percent (303 patients), respectively,
in the placebo group.

244 n engl j med 353;3 www.nejm.org july 21 , 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

Table 2. Rates of Primary and Secondary End Points.*

Placebo Group Atorvastatin Group

End Point (N=636) (N=619) RR (95% CI) P Value

no. (%)
Primary 243 (38) 226 (37) 0.92 (0.77–1.10) 0.37
Death from cardiac causes 149 (23) 121 (20) 0.81 (0.64–1.03) 0.08
Sudden death 83 (13) 77 (12)
Fatal myocardial infarction 33 (5) 23 (4)
Death due to congestive heart failure 24 (4) 17 (3)
Death after interventions to treat coronary 4 (0.6) 3 (0.5)
heart disease
Other death due to coronary heart disease 5 (0.8) 1 (0.2)
Nonfatal myocardial infarction 79 (12) 70 (11) 0.88 (0.64–1.21) 0.42
Silent 50 (8) 41 (7)
Nonsilent 35 (6) 33 (5)
Fatal stroke 13 (2) 27 (4) 2.03 (1.05–3.93) 0.04
Ischemic 7 (1) 18 (3)
Hemorrhagic 5 (0.8) 3 (0.5)
Other (not classified) 1 (0.2) 6 (1)
Nonfatal stroke 32 (5) 33 (5) 1.04 (0.64–1.69) 0.89
Secondary
All cardiac events combined 246 (39) 205 (33) 0.82 (0.68–0.99) 0.03
Death from cardiac causes 149 (23) 121 (20)
Nonfatal myocardial infarction 79 (12) 70 (11)
PTCA 45 (7) 34 (5)
CABG 30 (5) 24 (4)
Other interventions to treat coronary heart disease 0 1 (0.2)
All cerebrovascular events combined 70 (11) 79 (13) 1.12 (0.81–1.55) 0.49
Stroke 44 (7) 59 (10) 1.33 (0.90–1.97) 0.15
Ischemic 33 (5) 47 (8)
Hemorrhagic 8 (1) 5 (1)
Other (not classified) 6 (1) 10 (2)
TIA or PRIND 31 (5) 26 (4)
Death from all causes 320 (50) 297 (48) 0.93 (0.79–1.08) 0.33
Death from causes other than cardiovascular 158 (25) 149 (24) 0.95 (0.76–1.18) 0.62
or cerebrovascular disease
Fatal infection 68 (11) 60 (10)
Fatal cancer 19 (3) 17 (3)
Other 71 (11) 72 (12)

* The total number of patients reaching the primary end point does not equal the sum of the numbers for each component
of the primary end point, because only the first event per patient is included in the primary end point. Thus, a patient who
had a stroke and a myocardial infarction was counted once in the primary end point, but appears in the separate totals
for stroke and myocardial infarction. RR denotes relative risk, CI confidence interval, CABG coronary-artery bypass graft-
ing, TIA transient ischemic attack, and PRIND prolonged reversible ischemic neurologic deficit.

n engl j med 353;3 www.nejm.org july 21, 2005 245

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 The new england journal of medicine

Patients receiving hemodialysis generally have mg per deciliter (1.81 mmol per liter) recommended
many adverse and serious adverse events (Table 3), by the Third Adult Treatment Panel of the National
but no cases of rhabdomyolysis or severe liver dis- Cholesterol Education Program for persons at very
ease were detected in either group. The study med- high risk of cardiovascular disease. Despite the high
ication was discontinued by the investigators in rate of cardiovascular events and the pronounced
one patient receiving placebo because of a report of LDL cholesterol–lowering activity of atorvastatin, a
myalgia in combination with elevated creatine ki- significant reduction in the incidence of the com-
nase levels. posite primary end point was not achieved.
Of nominal significance, more cases of fatal
discussion stroke occurred in the atorvastatin group (27) than
in the placebo group (13). This finding is unex-
We examined the value of lowering the level of LDL plained and could be a chance finding, particularly
cholesterol in patients receiving hemodialysis who in view of the data from CARDS, which indicate that
have type 2 diabetes mellitus, among whom the av- atorvastatin lowers the incidence of stroke.3 That
erage annual incidence of myocardial infarction or study reported a relative risk for stroke of 0.52 (95
death from coronary heart disease is 8.2 percent. percent confidence interval, 0.31 to 0.89) in per-
This incidence rate exceeds the average annual rates sons with type 2 diabetes mellitus who were taking
of major coronary events that were reported in the atorvastatin. The rate of fatal and nonfatal stroke
placebo group of the Scandinavian Simvastatin Sur- decreased from 2.8 to 1.5 percent (39 vs. 21 pa-
vival Study (6.6 percent) and is the highest rate of tients), whereas in the present study, it increased
cardiovascular events in a long-term prospective tri- from 7.0 to 9.7 percent (44 vs. 59 patients).
al of statin therapy.19 Atorvastatin (20 mg daily) low- The complete absence of a stroke benefit and
ered LDL cholesterol levels by 42 percent, to 72 mg the increase in fatal strokes contribute considerably
per deciliter, which is close to the target value of 70 to the finding that the treatment effect on the pri-
mary end point was less than predicted. A possible
reason for the unexpected results with regard to the
Table 3. Adverse Events.* primary end point might be related to the LDL cho-
lesterol concentration at baseline. In general, the
Placebo Atorvastatin absolute risk reduction attained by lowering LDL
Event Group Group
cholesterol by a given percentage is less when pre-
no. of events treatment concentrations are low than when they
Total 2255 2276
are high.20 The baseline levels of LDL cholesterol
among patients in our study were, on average, above
Serious events 1060 1073
the target (126 mg per deciliter [3.25 mmol per
Events requiring hospitalization 942 949 liter]). Given the log-linear relation between LDL
Events requiring discontinuation of study drug 52 73 cholesterol and coronary heart disease, reducing
Drug-related serious events 1 1
levels of LDL cholesterol by 40 percent from a start-
ing level of 125 mg per deciliter would result in an
Diagnosis of cancer 44 39
approximate relative risk reduction of 30 percent
Severe hyperkalemia 9 3 or more.20 This estimate is empirically supported
Severe hypoglycemia 4 6 by the results of CARDS3 and the British Heart Pro-
Ventricular fibrillation or tachycardia 13 7
tection Study21 and is very close to our initial as-
sumption of a risk reduction of 27 percent.
Myalgia or myopathy 5 7
Since we did not fully achieve this benefit, we
Creatine kinase level speculate that the pathogenesis of vascular events
3 to 5 times the upper limit of normal 3 11 in patients with diabetes mellitus who are receiving
>5 to 10 times the upper limit of normal 1 1
hemodialysis may, at least in part, be different from
that in patients without end-stage renal disease.
Alanine aminotransferase level >4 times the upper 1 5
limit of normal
Subgroup analyses showed no difference in out-
come for any LDL cholesterol level or patients with
* Some patients had more than one event. and patients without cardiovascular disease. Inter-
estingly, there was a continuous decrease in LDL

246 n engl j med 353;3 www.nejm.org july 21 , 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

cholesterol levels over time among patients in both the benefit of atorvastatin is limited when interven-
groups. Some malnutrition cannot be ruled out dur- tion with statins is postponed until patients have
ing the course of the study, although there was no reached end-stage renal disease. Subgroup analyses
decrease in the body-mass index. of major statin-intervention trials documented a car-
The extremely high rate of death from cardio- diovascular benefit in patients with chronic kidney
vascular causes among patients receiving dialysis22 disease (stages 1, 2, and 3 according to the classifi-
is explained by more than the traditional coronary cation of the National Kidney Foundation).28,29
risk factors. Apart from the presence of many ag- According to CARDS, lowering LDL cholesterol
gravating coexisting factors, such as inappropriate levels early during the clinical course of type 2 dia-
left ventricular hypertrophy, cardiac fibrosis, cardiac betes mellitus is of benefit.3 Third, there was no ex-
microvessel disease,23 and sympathetic overactivity, cess of serious adverse events; specifically, no cases
among others, there are also indications that ath- of rhabdomyolysis occurred, but we found a nomi-
erosclerosis itself is promoted by risk factors other nally significant increase in fatal stroke.
than the traditional cardiovascular risk factors.24,25 We conclude that in persons with type 2 diabe-
The most plausible explanation for the absence of a tes mellitus who are receiving maintenance hemo-
significant effect on mortality from cardiac causes dialysis and have LDL cholesterol values between
and cardiac end points in this study is the presence 80 and 190 mg per deciliter, routine treatment with
of additional pathogenetic pathways in cardiovas- a statin to reduce the primary composite end point
cular disease. The dose of atorvastatin in the pres- of death from cardiac causes, myocardial infarction,
ent study was 20 mg, which is lower than the high and stroke is not warranted. The initiation of lipid-
dose used in a recent study by LaRosa et al.26 in lowering therapy in patients with type 2 diabetes
which intensive lipid-lowering therapy with atorva- mellitus who already have end-stage renal disease
statin at a dose of 80 mg per day was more effective may come too late to translate into consistent im-
than a dose of 10 mg per day in patients with stable provement of the cardiovascular outcome.
coronary heart disease. However, whether such an Supported by Pfizer. The committee members and investigators
did not receive remuneration for conducting the study, except for re-
advantage would accrue if patients with type 2 dia- imbursement of costs to participate in scientific meetings.
betes who were receiving dialysis were given a high- Dr. Wanner reports having received consulting fees and lecture
er dose of atorvastatin is unknown. fees from Genzyme; Dr. März, consulting fees, lecture fees, a research
grant and stock options from Pfizer; and Dr. Mann, lecture fees
Several important conclusions can be drawn from Aventis, Roche, and Janssen Cilag. Dr. Ritz is a member of the
from this study. First, we showed that it is difficult safety board of a trial sponsored by AstraZeneca and reports having
to rely on uncontrolled observational studies that received consulting fees from the company.
We are indebted to the German Association for Clinical Nephrol-
show substantial advantages of statins in the treat- ogy (K.-W. Kühn, chair) and the Association of German Nephrology
ment of patients receiving hemodialysis.9,27 Sec- Centers (H. Kütemeyer, chair).
ond, and more important, is the conclusion that

ap p e n d i x
The following investigators and research coordinators participated in the study known as the 4D Study (a complete list is available at
www.uni-wuerzburg.de/ nephrologie): Steering committee: C. Wanner, E. Ritz. Clinical coordinator: V. Krane. Medical end-point monitors: Z. Ülger,
F. Swoboda. Data and safety monitoring committee: M. Wehling (chair), E. Keller (deceased), M. Schumacher, T. Eschenhagen. Event committee:
J. Mann (chair), J. Bommer, P. Schanzenbächer, P. Schollmeyer, M. Schartl. Electrocardiography monitoring board: F. Heinrich, H. Mörl. Biometric
and statistical analysis: University of Freiburg, M. Olschewski. Central laboratory (lipid and safety core laboratory): University of Freiburg, W. März.
Contract research organization: Kendle, Munich, S. Reichmuth (Project manager); Datamap, Freiburg, J. Lilienthal. Sponsor: Pfizer, Karlsruhe, G.
Ruf, B. Rauer (Project manager).

references
1. Cheung BM, Lauder IJ, Lau CP, Kumana ton PN, et al. Primary prevention of cardio- 5. Koch M, Thomas B, Tschöpe W, Ritz E.
CR. Meta-analysis of large randomized con- vascular disease with atorvastatin in type 2 Survival and predictors of death in dialysed
trolled trials to evaluate the impact of statins diabetes in the Collaborative Atorvastatin diabetic patients. Diabetologia 1993;36:
on cardiovascular outcomes. Br J Clin Phar- Diabetes Study (CARDS): multicentre ran- 1113-7.
macol 2004;57:640-51. domised placebo-controlled trial. Lancet 6. Renal Data System. USRDS 2003 annual
2. Armitage J, Bowman L. Cardiovascular 2004;364:685-96. data report: atlas of end-stage renal disease
outcomes among participants with diabetes 4. Schwenger V, Hofmann A, Kalifeh N, et in the United States. Bethesda, Md.: Nation-
in the recent large statin trials. Curr Opin al. Uremic patients — late referral, early al Institute of Diabetes and Digestive and
Lipidol 2004;15:439-46. death. Dtsch Med Wochenschr 2003;128: Kidney Disease, 2003. (Accessed June 27,
3. Colhoun HM, Betteridge DJ, Durring- 1216-20. (In German.) 2005, at http://www.usrds.org.)

n engl j med 353;3 www.nejm.org july 21, 2005 247

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.
 atorvastatin in type 2 diabetes and hemodialysis

7. Prichard SS. Impact of dyslipidemia in 15. O’Brien PC, Fleming TR. A multiple in chronic renal disease. Am J Kidney Dis
end-stage renal disease. J Am Soc Nephrol testing procedure for clinical trials. Biomet- 1998;32:Suppl 3:S112-S119.
2003;14:Suppl 4:S315-S320. rics 1979;35:549-56. 23. Amann K, Breitbach M, Ritz E, Mall G.
8. Iseki K, Yamazato M, Tozawa M, Tak- 16. Tschöpe W, Koch M, Thomas B, Ritz E. Myocyte/capillary mismatch in the heart of
ishita S. Hypocholesterolemia is a signifi- Serum lipids predict cardiac death in diabet- uremic patients. J Am Soc Nephrol 1998;9:
cant predictor of death in a cohort of ic patients on maintenance hemodialysis: 1018-22.
chronic hemodialysis patients. Kidney Int results of a prospective study. Nephron 24. Takayama F, Aoyama I, Tsukushi S, et al.
2002;61:1887-93. 1993;64:354-8. Immunohistochemical detection of imi-
9. Seliger SL, Weiss NS, Gillen DL, et al. 17. Koch M, Kuthuhn B, Trenkwalder E, et dazolone and N(epsilon)-(carboxymethyl)
HMG-CoA reductase inhibitors are associ- al. Apolipoprotein B, fibrinogen, HDL cho- lysine in aortas of hemodialysis patients.
ated with reduced mortality in ESRD pa- lesterol, and apolipoprotein(a) phenotypes Cell Mol Biol 1998;44:1101-9.
tients. Kidney Int 2002;61:297-304. predict coronary artery disease in hemodi- 25. Himmelfarb J, Stenvinkel P, Ikizler TA,
10. Baigent C, Burbury K, Wheeler D. Pre- alysis patients. J Am Soc Nephrol 1997;8: Hakim RM. The elephant in uremia: oxidant
mature cardiovascular disease in chronic re- 1889-98. stress as a unifying concept of cardiovascular
nal failure. Lancet 2000;356:147-52. 18. Schulgen G, Olschewski M, Krane V, disease in uremia. Kidney Int 2002;62:1524-
11. Lins RL, Matthys KE, Verpooten GA, et Wanner C, Ruf G, Schumacher M. Sample 38.
al. Pharmacokinetics of atorvastatin and its sizes for clinical trials with time-to-event 26. LaRosa JC, Grundy SM, Waters DD, et
metabolites after single and multiple dosing endpoints and competing risks. Contemp al. Intensive lipid lowering with atorvastatin
in hypercholesterolaemic haemodialysis pa- Clin Trials 2005;26:386-96. in patients with stable coronary disease.
tients. Nephrol Dial Transplant 2003;18: 19. Scandinavian Simvastatin Survival Study N Engl J Med 2005;352:1425-35.
967-76. Group. Randomised trial of cholesterol low- 27. Mason NA, Bailie GR, Satayathum S, et
12. Wanner C, Krane V, Ruf G, März W, ering in 4444 patients with coronary heart al. HMG-coenzyme A reductase inhibitor
Ritz E. Rationale and design of a trial im- disease: the Scandinavian Simvastatin Sur- use is associated with mortality reduction in
proving outcome of type 2 diabetics on he- vival Study (4S). Lancet 1994;344:1383-9. hemodialysis patients. Am J Kidney Dis
modialysis. Kidney Int Suppl 1999;71: 20. Grundy SM, Cleeman JI, Merz CNB, et 2005;45:119-26.
S222-S226. al. Implications of recent clinical trials for 28. Tonelli M, Isles C, Curhan GC, et al. Ef-
13. Wanner C, Krane V, März W, et al. Ran- the National Cholesterol Education Program fect of pravastatin on cardiovascular events
domized controlled trial on the efficacy and Adult Treatment Panel III guidelines. Circu- in people with chronic kidney disease. Cir-
safety of atorvastatin in patients with type 2 lation 2004;110:227-39. [Erratum, Circula- culation 2004;110:1557-63.
diabetes on hemodialysis (4D Study): demo- tion 2004;110:763.] 29. K/DOQI clinical practice guidelines for
graphic and baseline characteristics. Kidney 21. Heart Protection Study Collaborative chronic kidney disease: evaluation, classifi-
Blood Press Res 2004;27:259-66. Group. MRC/BHF Heart Protection Study of cation, and stratification. Am J Kidney Dis
14. Nauck M, Winkler K, März W, Wieland cholesterol lowering with simvastatin in 2002;39:Suppl 1:S1-S266.
H. Quantitative determination of high-, low-, 20,536 high-risk individuals: a randomised Copyright © 2005 Massachusetts Medical Society.
and very-low-density lipoproteins and lipo- placebo-controlled trial. Lancet 2002;360:
protein(a) by agarose gel electrophoresis 7-22.
and enzymatic cholesterol staining. Clin 22. Foley RN, Parfrey PS, Sarnak MJ. Clini-
Chem 1995;41:1761-7. cal epidemiology of cardiovascular disease

physician-journalist
The Journal is seeking a physician with substantial reporting
experience to write occasional articles on timely topics in medicine
and society for the Perspective section. Send curriculum vitae and
writing samples to Perspective Editor, New England Journal of Medicine,
10 Shattuck St., Boston, MA 02115, or at writer@nejm.org.

248 n engl j med 353;3 www.nejm.org july 21 , 2005

The New England Journal of Medicine

Downloaded from nejm.org on February 15, 2015. For personal use only. No other uses without permission.
Copyright © 2005 Massachusetts Medical Society. All rights reserved.