SPECIAL ARTICLE
Roger E. Pecoraro, MD
Pathways to Diabetic
Limb Amputation
Basis for Prevention
We defined the causal pathways responsible for 80
consecutive initial lower-extremity amputations to an
extremity in diabetic patients at the Seattle Veterans
Affairs Medical Center over a 30-mo interval from 1984
to 1987. Causal pathways, either unitary or composed of
various combinations of seven potential causes (i.e.,
ischemia, infection, neuropathy, faulty wound healing,
minor trauma, cutaneous ulceration, gangrene), were
determined empirically after a synthesis by the
investigators of various objective and subjective data.
Estimates of the proportion of amputations that could
be ascribed to each component cause were calculated.
Twenty-three unique causal pathways to diabetic limb
amputation were identified. Eight frequent constellations
of component causes resulted in 73% of the
amputations. Most pathways were composed of multiple
causes, with only critical ischemia from acute arterial
occlusions responsible for amputations as a singular
cause. The causal sequence of minor trauma, cutaneous
ulceration, and wound-healing failure applied to 72% of
the amputations, often with the additional association of
infection and gangrene. We specified precise criteria in
the definition of causal pathway to permit estimation of
the cumulative proportion of amputations due to various
causes. Forty-six percent of the amputations were
attributed to ischemia, 59% to infection, 61% to
neuropathy, 81% to faulty wound healing, 84% to
ulceration, 55% to gangrene, and 81% to initial minor
trauma. An identifiable and potentially preventable
pivotal event, in most cases an episode involving minor
From the Department of Medicine, Division of General Internal Medicine, and
Department of Orthopaedics, University of Washington School of Medicine; the
Department of Epidemiology, University of Washington School of Public Health
and Community Medicine; and the Seattle Veterans Affairs Medical Center,
Seattle, Washington.
Address correspondence and reprint requests to Roger E. Pecoraro, MD, Med-
ical Comprehensive Care Unit (111M), Seattle Veterans Affairs Medical Center,
1660 South Columbian Way, Seattle, WA 98108.
Received for publication 17 May 1989 and accepted in revised form 8 No-
vember 1989.
DIABETES CARE, VOL. 13, NO. 5, MAY 1990
Gayle E. Reiber, PhD
Ernest M. Burgess, MD
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
trauma that caused cutaneous injury, preceded 69 of 80
amputations. Defining causal pathways that predispose
to diabetic limb amputation suggests practical
interventions that may be effective in preventing
diabetic limb loss. Diabetes Care 13:513-21, 1990
D
iabetes is associated with >50% of the 120,000
lower-limb amputations performed annually in
the United States for indications other than
trauma (L. Ceiss, unpublished observations). The
overall risk for amputation is increased in diabetes 15-
fold beyond that for nondiabetic people (1). This is likely
to reflect, in part, the prevalence of potent pathophys-
iological risk factors for amputation, including periph-
eral neuropathy and severe arteriosclerosis obliterans,
which have been suggested in case series of diabetic
amputations. Additional possible risk factors in diabetic
individuals involve more complex pathology, i.e., fail-
ure of cutaneous wound healing and gangrene, or pre-
disposing environmental events, i.e., minor trauma that
causes skin ulceration.
Most diabetic lower-limb amputations probably result
from combinations of contributing causes rather than
from unitary causes. This contrasts with lower-limb am-
putations among nondiabetic people, which are likely
to result from severe peripheral vascular disease (2,3).
Definition and quantitation of the major independent
risk factors for amputation and description of their com-
mon interactions must precede the implementation of
strategies for prevention of lower-extremity amputations
in diabetes.
Most of the available information about risk factors
and potential causes for amputation depends on obser-
vations from surgical case series (4-6). Because case
513
CAUSAL PATHWAYS TO AMPUTATION
series usually represent patients referred for heteroge-
neous indications and procedures, control groups are
seldom assembled. Therefore, the prevalence and se-
verity of separate potential risk factors observed in those
series are not adequate to estimate either the relative
risks or the proportion of amputations that result from
the occurrence of particular factors among the diabetic
population. Controlled epidemiological studies are nec-
essary to quantify the risk due to specific potential causes.
Cohort studies can define rates of disease and relative
risks but are difficult, costly, and often impractical. The
case-control design is efficient and provides a mecha-
nism for estimating relative risk due to various expo-
sures. Unfortunately, only one controlled study, an
analysis of selected characteristics of Pima Indians with
diabetes, has been published describing relative risks for
amputation (7).
A model to define the causes of diabetic amputation
that can account for interactions of multiple synergistic
causes, both concurrent and sequential, would be useful
to anticipate clinical circumstances that convey high risk.
These features are implicit in the concept of a causal
pathway to amputation. Each diabetic amputation im-
plies the existence of a completed causal pathway of
predisposing factors. There is a need to recognize com-
monly occurring causal pathways among individuals re-
quiring amputation and to estimate their frequencies to
first identify and then test effective preventive strategies
and interventions.
This investigation was designed to address three is-
sues. First, what were the responsible causal pathways
in 80 consecutive diabetic male subjects requiring an
initial lower-extremity amputation during a 30-mo pe-
riod at a single hospital, and what proportions of the
amputations were caused by sensory neuropathy, is-
chemia, infection, minor trauma, skin ulceration, failed
wound healing, and gangrene? Second, what was the
prevalence among these cases of a limited number of
additional conditions or circumstances, e.g., negligent
self-care practices due to patient lack of knowledge or
noncompliance, which probably contributed to ampu-
tation risk but cannot necessarily be supported as es-
sential component causes? Third, could a pivotal event
be identified that initiated the causal pathway leading
to each amputation?
RESEARCH DESIGN AND METHODS
We studied 80 consecutive diabetic male veterans, aged
30-85 yr, who required first amputations of the affected
lower extremity for indications other than trauma at the
Seattle Veterans Affairs Medical Center from October
1984 to April 1987. Patients were excluded if they had
had a previous amputation involving the same extrem-
ity. A concurrently selected control population included
236 diabetic male veterans, aged 30-85 yr, scheduled
for elective surgical procedures not related to diabetes
and chosen from a predefined list of operative proce-
dures performed by surgeons from eight subspecialty
514
services. Subjects were excluded as controls if they gave
a history of previous amputation or had lower-extremity
ulcers or infections that had persisted for >4 wk. In this
study, control subjects were used exclusively for cal-
culations of population-attributable risk percent for spe-
cific measures of ischemia and neuropathy.
All patients admitted for inpatient care or outpatient
surgery during the study were screened for the diagnosis
of diabetes by one of the investigators (G.E.R.). The di-
agnosis required 7) previous diagnosis of diabetes by a
physician; 2) a history of receiving prescribed hypogly-
cemic medication; or 3) in the absence of a previous
diagnosis, laboratory determination of hemoglobin A]c
^6.3% and/or fasting plasma glucose >140 mg/dl. The
population from which diabetic amputees and control
subjects were selected included 19,508 individuals ad-
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
mitted and 1659 patients scheduled for outpatient sur-
gery. One hundred percent of the 80 diabetic individ-
uals fulfilling the case eligibility criteria agreed to
participate, whereas 236 of 239 eligible potential con-
trol subjects participated (98.7% participation rate).
The purpose of identifying causal pathways to dia-
betic amputation is to link important responsible events,
pathophysiological factors, and other conditions con-
tributing to amputation cause in a manner that may sug-
gest specific clinical strategies for anticipating and pre-
venting diabetic limb loss. An effective model should
use a definition of the causal pathway that permits as-
similation of empirical information derived from actual
representative amputation cases. The model should ac-
commodate causes for a particular amputation that might
occur early in a causal sequence, i.e., cutaneous ul-
ceration, as well as causes that represent the final pre-
cipitating circumstance necessary for a particular am-
putation to occur, e.g., gangrene. For clarity and
simplicity, the causal pathway should be concise and
exclude nonessential factors. Finally, the causal path-
way should be defined in a way that provides insight
into the potential usefulness of interventions directed
toward removal or negation of particular elements of the
causal chain.
A model for causation that meets these objectives has
been described by Rothman (8,9). Therefore, to corre-
spond with Rothman's conceptualization of "sufficient
cause," we defined the causal pathway to diabetic lower-
limb amputation as a set of minimal conditions and events
that inevitably produce disease; "minimal" implies that
none of the conditions or events is superfluous. In dis-
ease etiology, the completion of a sufficient cause may
be considered equivalent to the onset of disease (8). As
part of this construct, "component causes" have been
defined as causes of interest that are not sufficient in
themselves but are required components of one or more
distinctive sufficient causes (8,9). Consequently, the lack
of any component cause from the minimal necessary set
renders the remaining component causes insufficient (Fig.
1). The causal pathway concept is valuable in disease
prevention because the disease effect can be delayed or
prevented by removing any individual component of a
sufficient causal pathway, which then renders the joint
DIABETES CARE, VOL. 13, NO. 5, MAY 1990
 R.L PECORARO, G.E. REIBER, AND E.M. BURGESS

Consideration was limited to seven potential causes,

SUFFICIENT CAUSE :
based on available historical and quantitative data. These
seven causes were suggested by previous literature on
— INEVITABLY PRODUCES
THE EFFECT diabetic amputations. They included the four major
pathophysiological mechanisms implicated in diabetic-
— RESTRICTED TO THE limb loss, neuropathy, ischemia, infection, and wound-
MINIMAL NUMBER OF
COMPONENT CAUSES healing failure (10-13); two common soft tissue com-
REQUIRED FOR plications, cutaneous ulceration and gangrene; and
CAUSATION minor trauma, which represents a frequent initiating
environmental event encompassing direct accidental in-
COMPONENT CAUSE : jury or trauma from repetitive mechanical pressure from
— NOT SUFFICIENT IN ITSELF
footwear.
Subjects were interviewed to obtain a detailed stan-
— REMOVAL OR BLOCKING RENDERS ACTION OF dardized medical history including demographic infor-
OTHER COMPONENTS INSUFFICIENT
mation, general medical history, specific diabetes his-

Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023

FIG. 1. Diagram of sufficient and component causes. A-
E represent causes that are not sufficient in themselves
but that are required components of a sufficient cause that
inevitably produces effect. Adapted from Rothman (8,9).
action of the other components insufficient. Figure 2
shows a representation of a hypothetical causal pathway
to an individual amputation according to this model.
tory, history of the current lesion and previous foot and
leg complications, health care, self-care, and life-style.
Hospital and outpatient medical records were abstracted
for the 5 yr preceding study enrollment. Before surgery
the lower extremities of all subjects were examined
to document cutaneous lesions and skin or soft tissue
infections, with findings recorded on a wound-coding
sheet. All subjects were evaluated directly by one or

NEUROPATHY MINOR TRAUMA ULCERATION FAULTY HEALING GANGRENE

BASELINE 4. ENVIRON- 4. SKIN 4. INTERCURRENT .L INTERCURRENT

nr
PATHOLOGY MENTAL • LESION PATHO- PATHO-
EVENT PHYSIOLOGY PHYSIOLOGY

ACCUMULATION of COMPONENT COMPLETED

C A U S E S TO FORM A SUFFICIENT CAUSE CAUSAL CHAIN
JOAMFUTATIOl

FIG. 2. Representation of causal pathway to individual amputation, which includes essential contributions from un-
derlying diabetes-related pathophysiology (neuropathy), initiating environmental event (minor trauma), formation of
foot lesion, and subsequent healing complications. Eventual occurrence of gangrene is terminal event of this causal
chain, which requires participation of all preceding components before becoming sufficient to cause amputation.
Theoretically, amputation could have been avoided by elimination of any one component cause before convergence
of causal chain.

DIABETES CARE, VOL. 13, NO. 5, MAY 1990 515

CAUSAL PATHWAYS TO AMPUTATION
more of the investigators during interviews (G.E.R.), pre-
surgical examinations and wound assessments (R.E.P.,
E.M.B., G.E.R.), and/or amputation surgery (E.M.B.,
R.E.P.).
Lesions were photographed. Sensory perception (vi-
bration, touch, two-point discrimination, and stereog-
nosis), Doppler segmental blood pressures, and trans-
cutaneous oxygen tension at standardized sites (14)
were measured on each subject by the same technician.
Blood was drawn by venipuncture after overnight fast
shortly after hospital admission for determination of rou-
tine clinical chemistry and hematology values, a panel
of plasma nutritional indicators (albumin, carotene,
zinc, ascorbic acid, pyridoxine, and zinc-protoporphyrin
heme ratio), and glycosylated hemoglobin.
Historical data, wound descriptions and coding, pho-
tographs, and objective neurological, vascular, and lab-
oratory data were collated in each case for review by
the investigators, and causes of amputation were as-
signed and recorded independently on a standardized
summary worksheet.
The criteria for assignment of causes by the investi-
gators to individual cases included a synthesis of objec-
tive and subjective considerations. Guidelines were
agreed on in advance for evaluation of data and assign-
ment oi potential causes. Assignment of neuropathy as
a cause required objective evidence of sensory neurop-
athy of the lower extremities and a judgment by the
investigators that the existence of neuropathy was an
essential component in the causal pathway to amputa-
tion. Assignment of ischemia required objective evi-
dence of substantially impaired arterial circulation of the
lower extremities from physical examination of the skin,
cutaneous appendages, and arterial pulses, usually ac-
companied by an ankle/arm index ^0.70 and/or local
transcutaneous oxygen tension <20 mmHg (14-18).
These findings had to be consistent with a clinical judg-
ment by the investigators that circulatory impairment
represented a critical contribution to the causal pathway
to amputation. Presence of ulceration or gangrene was
determined by physical examination, and to qualify as
a component cause required documentation in the med-
ical record that it was considered as an indication for
amputation. Major infection was documented by direct
physical examination, review of microbiological culture
results, and/or radiology and pathology reports consis-
tent with osteomyelitis. Furthermore, to qualify as a
component cause, infection had to be overwhelming or
unresponsive to medical and surgical treatment. Wound-
healing failure was implicated as a cause if there was
no healing progress after 6 wk in the case of major cu-
taneous ulceration, and this was judged by clinicians or
investigators to have been a direct or indirect indication
for the amputation. Minor trauma was assigned as a
cause if it was clearly documented by medical history
and was judged by the investigator to have initiated a
sequence of subsequent pathological events that termi-
nated in limb amputation.
The predominant and final criterion for assignment of
516
a component cause was a unanimous consensus by the
investigators that a synthesis of the relevant data justified
the assignment of that factor as an essential contributing
cause for the particular amputation. For each amputa-
tion, the final element in the causal chain was identified
first. If the final cause was not by itself sufficient, one
or more additional required component causes were
selected using the standardized summary worksheet.
Rarely was there disagreement among investigators re-
garding assignment of the final causes or most causal
constellations. In the few cases in which there was initial
disagreement, final consensus was achieved with a
modified Delphi procedure (19). The Delphi procedure
is a method used in arriving at consensus. It was mod-
ified by allowing face-to-face discussions after the initial
round of independent assessment until there was unan-
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
imous agreement on assignments of cause. The inves-
tigators represent complementary areas of professional
expertise, including clinical diabetology (R.E.P.), or-
thopedic and amputation surgery and rehabilitation
(E.M.B.), and nursing, public health, and epidemiology
(G.E.R.).
Statistical analysis. Results are reported with standard
descriptive statistics. The frequencies of each compo-
nent cause, final cause, and complete causal pathway
were computed. Values of population-attributable risk
percent, an expression in percentage terms of the pro-
portion of amputation cases in the total diabetic popu-
lation due to the risk factor, which can be estimated
from the odds ratio and the proportion of the population
exposed (20), were calculated for indicators of signifi-
cant ischemia among diabetic amputees and surgical
control subjects. Values were calculated independently
for measurements of transcutaneous oxygen tension,
which was <20 mmHg at the dorsal foot, and for ankle/
arm index, which was ^0.70 for the affected limb (14-
18,21). Similarly, the population-attributable risk per-
cent for significant peripheral sensory neuropathy was
calculated from the frequency of absent distal vibratory
sensation among patients and control subjects to indi-
cate impaired protective sensation.
RESULTS
The 80 subjects receiving initial lower-extremity am-
putations had a mean ± SD age of 63.4 ± 11.9 yr and
had a clinical duration of diabetes of 13.3 ± 10.5 yr.
Insulin-dependent diabetes mellitus occurred in 11 sub-
jects and non-insulin-dependent diabetes mellitus in 69.
These and other characteristics were not statistically dif-
ferent among control subjects (Table 1). The amputation
levels included 12 above knee, 35 below knee, 5 trans-
metatarsal, and 28 toe amputations.
Twenty-three unique causal pathways were repre-
sented among the 80 consecutive cases of diabetic lower-
limb amputation. Twenty-two causal pathways (96%)
were multicomponent, with only one observed pathway
DIABETES CARE, VOL. 13, NO. 5, MAY 1990
 R.E. PECORARO, G.E. REIBER, AND E.M. BURGESS

TABLE 1 culation provides an estimate of the proportion of am-

Characteristics of diabetic amputees and control subjects putations that might have been prevented by the reso-
lution or removal of the particular component factor
Amputees Controls from the causal chain.
n 80 236 There was a trend for neuropathy, minor trauma, ul-
Age (yr) ceration, and faulty wound healing to be represented in
Mean ± SD 63.4 ± 1 1 . 9 61.1 ± 10.0 a higher proportion of causal pathways among amputees
Median 62.5 63.0 above the median age (62 yr) compared with the younger
Range 31-85 31-83 half of the case population (Table 2). However, this trend
Clinical duration of diabetes (yr) 13.3 ± 10.5 10.9 ± 9.7 reached statistical significance at P < 0.05 only for faulty
Annual income <$15,000 (%) 62 57.8 wound healing, which occurred in most cases but was
Education <12 yr (%) 67.5 60 more common in older subjects. The prevalence of is-
White (%) 86 82 chemia, infection, and gangrene as component causes
Smoking history (%)
was similar in both age strata.
Current 36 24
Ever 72.5 78.4 Failure of normal wound healing after cutaneous ul-
ceration, a condition well recognized in diabetes but

Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023

Never 28
attributed to a unitary cause, ischemia. The eight most
frequent pathways observed together accounted for 73%
of all diabetic amputations (Fig. 3).
The proportions of amputations attributed to individ-
ual component causes are listed in Table 2. Each pro-
portion was calculated as the sum of the amputation
fractions that resulted from causal pathways that con-
tained that component cause (9). Because of the re-
stricted definition of causal pathway adopted for this
analysis (see RESEARCH DESIGN AND METHODS), this cal-
[GJ [GJ [GJ |G
% Of 80 LEA's !
9 8 5 4
73%
FIG. 3. Most unique causal pathways required interaction
of pathophysiological components (ischemia [I], neurop-
athy [N], infection [INF], faulty wound healing [FH]), path-
ological conditions (ulceration [U], gangrene [G]), and en-
vironmental events (minor trauma [T]). Prevalent triad of
factors (dashed boxes) leading to amputation included T
causing U, which was subsequently complicated by FH.
LEA, lower-extremity amputation. These are 23 causal path-
ways and 80 LEAs.
22
with complex etiologies that are not entirely explained,
emerged in this analysis as the most prevalent patho-
physiological component cause (81%) leading to am-
putation (13). Considered an independent causal factor,
wound failure illustrated many of the interactions that
link separate potential amputation causes into critical
combinations that are sufficient in concert to bring about
amputation. Cutaneous ulcers complicated by wound
failure occurred in 58 of 80 cases (72%). Similarly,
wound-healing failure was eventually associated with
unsuccessfully treated infection or gangrene, forcing the
surgical decision for amputation in 42% of the cases.
A particular critical triad of component causes, an
initial episode of minor trauma, resulting in cutaneous
ulceration and subsequent failure to heal, preceded 72%
of amputations. Absent protective sensation was consid-
ered to be a component cause in 82% of these ampu-
tations. Infection was a common complicating factor in
the various causal pathways to amputation and most
often affected cases with chronic wound-healing failure.
cn
Sixty-one percent of the cases involving the triad of mi-
C/5
nor trauma, ulceration, and wound-healing failure also
O had a major infection that was considered essential to
m the causal pathway leading to amputation.
Ischemia was found to be a causal factor in 46% of
m the amputations. Of the seven potential causal factors
O considered, ischemia was the only cause found to be
c
C/5 singularly responsible for lower-limb amputation. Three
amputations in individuals with severe peripheral arte-
rial disease were precipitated by acute arterial occlu-
sions. Critical ischemia was associated with 62% of cases
with nonhealing ulceration, with gangrene associated
with half of those cases.
27% Gangrene as a cause for amputation occurred in com-
bination with both ischemia and infection. Gangrene
occurred in 55% of all amputated limbs and was asso-
ciated with critical ischemia, alone, or in combination
with major infection in 40% of the amputations. How-
ever, in 24% of all amputations, infection accompanied
gangrene as a component cause without involvement of
preexisting ischemia.
Another way to estimate the proportion of amputa-
tions that might have been prevented by removal of a

DIABETES CARE, VOL. 13, NO. 5, MAY 1990 517

CAUSAL PATHWAYS TO AMPUTATION
TABLE 2
inULL i.
Proportion of amputations due to individual causes and final component causes
Amputation proportion (%)
Component cause All cases <62 yr (n = 40)
Ischemia 46 48
Infection 59 52
Neuropathy 61 52
Faulty wound healing 81 70
Ulceration 84 75
Gangrene 55 50
Minor trauma 81 72
*Yates correction with
particular component cause is to calculate the popula-
tion-attributable risk percent from case-control data (20).
This parameter was calculated for specific measures of
ischemia and neuropathy that were performed in the
diabetic amputation cases and control subjects. These
estimates were compared with the proportion of am-
putations attributed to ischemia or neuropathy in the
causal pathway analysis for internal validation. Because
diabetic control subjects were ineligible for study par-
ticipation if they had nonhealing ulcers, persistent in-
fections, or gangrene, similar calculations could not
be performed for the remaining factors considered as
potential component causes. For both ischemia and
neuropathy, the etiologic fraction estimated by popu-
lation-attributable risk percent was comparable to the
percentage estimate from the corresponding causal
pathway analysis. The amputation proportion attribut-
able to ischemia was 46% compared with a population-
attributable risk percent of 36% calculated for ankle/
arm index <0.70, and 49% for dorsal foot transcuta-
neous oxygen tension <20 mmHg (Table 2). The pro-
portion attributed to neuropathy by causal pathway
analysis was 6 1 % compared with a population-attrib-
utable risk percent of 72% calculated for individuals
with absent distal vibratory sensation.
The final component cause in the causal pathway, the
factor that triggered the surgeon's decision to amputate,
was rarely sufficient in itself to bring about amputation,
yet is often considered the responsible cause for a par-
ticular amputation. As noted above, acute ischemia was
the solitary and therefore the final and responsible suf-
ficient cause in three cases. In the remainder of the cases,
which involved multiple component causes, the re-
sponsibility for amputation could not be due to only the
final cause, but by definition was shared by the various
contributory component causes. Furthermore, the final
cause in the causal pathway did not accurately reflect
the estimate of the proportion of amputations attributed
to that causal factor (Table 2). For example, ischemia
was considered to be the final cause in only 5% of the
cases, yet its elimination as a component cause would
518
Median age split
As final component
>62 yr (n = 40) P* in pathway (%)
45 1.00 5
65 0.36 41
70 0.17
92 0.02 14
92 0.07
60 0.50 40
90 0.09
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
have obviated the causal pathway to 46% of the am-
putations.
A pivotal antecedent event, defined as a component
cause that triggered a sequence of events culminating
in lower-extremity amputation, could be clearly iden-
tified in 86% of the cases. In 58 of 80 cases (73%) these
were traumatic events causing tissue injury. These in-
cluded shoe-related repetitive pressure leading to cuta-
neous ulceration in 36% of all cases, accidental cuts or
wounds in 8%, thermal trauma (2 cases of frostbite, 4
burns) in 8%, and decubitus ulceration in 8%. Sudden
vascular occlusions were implicated as pivotal events in
6 cases; 3 cases represented singular causes, corre-
sponding to the only unitary causal pathway observed
(Fig. 3). Miscellaneous pivotal events included iatro-
genic occurrences (2 cases), development of parony-
chiae (2 cases), and sinus tract formation during an ep-
isode of podagra (1 case).
Other factors contributed to many amputations, al-
though their participation could not be unequivocally
considered to be causal, based on current understand-
ing of their effects. Nutritional impairment believed to im-
pede normal tissue repair, indicated by plasma ascorbic
acid and/or zinc levels <2SD below the laboratory
mean (i.e., <0.36 mg/dl and <60 |xg/dl, respectively),
was found in over half the amputees (51 cases). Other
laboratory nutritional parameters indicated significant
deficiencies among amputees compared with the con-
trol population, including plasma albumin (3.59 ± 0.10
vs. 4.17 ± 0.03 g/dl, P < 0.001), carotene (61.6 ± 4.4
vs. 88.3 ± 3.5 g/dl, P < 0.001), and the zinc-proto-
porphyrin heme ratio, a sensitive measure of marrow
erythropoiesis (65.9 ± 3.9 vs. 50.9 ± 2.2 jxmol/mol,
P < 0.001) (22). Mean ± SD hospital admission weights
of the 80 amputee patients averaged 178 ± 35 lbs com-
pared with 192 ± 40 lbs for the 236 hospitalized dia-
betic control subjects. Body mass index (kg/m2) for am-
putation cases was 25.8 ± 5.6 compared with 27.6 ±
5.3 for control subjects, with P < 0.01 significantly dif-
ferent. Calculations of desirable weight and estimates of
the subjects' maximum weights did not differ between
DIABETES CARE, VOL. 13, NO. 5, MAY 1990

amputees and control subjects, suggesting that ampu-
tees had sustained significant weight losses (23).
Edema impaired local cutaneous blood flow of the
affected extremity in 31 cases (39%). Negligent self-care
practices that contributed to a subsequent amputation
were ascribed to lack of knowledge of appropriate foot
care principles in 19 cases (24%) and to noncompliance
with medical recommendations in 22 cases (28%). In-
adequate social support contributed to eight amputa-
tions (10%).
DISCUSSION
E
mpirically derived causal pathways represent the
assembly of essential component factors and
events that act together to culminate in individual
amputations. These causal pathways incorporate
interactions between important risk factors that would
be less obvious in traditional epidemiological study de-
signs. Causal pathways may include particular critical
exposures that may be difficult to specify in the case-
control design, particularly among control subjects, in
whom no memorable consequence is likely to be trig-
gered by the exposure (e.g., the episode of minor trauma,
which could produce ulceration in the insensitive foot).
Although analyses of the time sequences necessary for
interactions of potential risk factors may be problematic
in case-control studies, the time dimension is implicit
in the causal pathway concept. Each contributing cause
represents a logical and necessary precedent to related
sequential events that combine ultimately to constitute
a sufficient cause for amputation. The causal pathway
model can also incorporate synergistic causes that are
concurrent and independent rather than sequential. These
attributes of the model are shown in Fig. 2 to illustrate
a causal pathway that was observed in two of our cases.
Several conclusions are apparent from our analysis of
the causes for 80 consecutive diabetic amputations.
Definition of the common causal pathways that predis-
pose to diabetic limb amputation offers an intuitive con-
text for designing and implementing effective public
health and individual patient-specific interventions to
prevent amputations. The value of this approach is its
ability to estimate the proportion of cases in the popu-
lation that are due to particular factors (Table 2), and to
assess the importance of eliminating those factors as part
of a disease-prevention strategy. However, some causal
factors may be inherently difficult or impossible to elim-
inate, e.g., foot anesthesia as a result of sensory neu-
ropathy. Alternatively, measures could be instituted to
structure the environment to minimize the potential for
injury. Attempts to modify others may subject certain
patients to substantial additional risks (e.g., elective vas-
cular surgery in certain high-risk individuals). The causal
pathway concept facilitates selection of accessible,
practical, and low-risk foci for interventions that might
effectively and safely interrupt the causal chain and pre-
DIABETES CARE, VOL. 13, NO. 5, MAY 1990
R.E. PECORARO, G.E. REIBER, AND E.M. BURGESS
vent a substantial proportion of lower-extremity ampu-
tations (e.g., prevention of foot trauma, early treatment
of lower-extremity infections, improved ulcer healing).
We confirmed that multiple causes usually interact to
bring about each individual diabetic limb amputation.
The frequent participation of multiple factors, each of
which is rarely sufficient by itself to produce amputa-
tion, may explain the limited success achieved by sin-
gular intervention aimed at limb salvage. This ob-
servation supports the emphasis by experts in the
management of diabetic foot problems who recommend
a multidisciplinary team approach to prevention, diag-
nosis, and treatment (10,12,24).
The occurrence of a critical causal sequence impli-
cating minor trauma, skin ulceration, and faulty wound
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
healing in 72% of diabetic limb amputations has im-
portant ramifications. This finding documents and un-
derscores the importance of wearing appropriate pro-
tective footwear by diabetic patients with impaired
protective sensation secondary to sensory neuropathy.
Continued research toward understanding the basis for
faulty wound healing in diabetes and the development
of effective therapies to accelerate wound repair, e.g.,
the topical application of human polypeptide growth
factors to heal diabetic skin ulcers (25), may bring re-
alization of the goals of the National Diabetes Advisory
Board to reduce the amputation rate in diabetes by 40%
before the year 2000 (26).
Our results highlight other exposures and interactions
that increase amputation risk. Critical infection is a fre-
quent concomitant of the prolongation or failure of cu-
taneous wound healing. Furthermore, in many cases
major infection is a precursor to gangrene, independent
of chronic ischemia from severe peripheral vascular dis-
ease. The prevalence of limb edema associated with
diabetic amputation in our study confirms the previously
reported strong association of edema from various causes
with diabetic gangrene (27). The common occurrence
of patient lack of knowledge and noncompliance with
critical aspects of diabetic foot care principles supports
the role of professional and patient education and mo-
tivation in preventive care.
The smoking status of the 80 amputees and 236 con-
trol subjects, assessed by personal interview, indicated
the "ever smoked" prevalence was 72.5 and 78.4%,
respectively (Table 1). Similar prevalence findings for
males in these age groups were reported in the National
Health Interview Survey (28). An analysis of the data of
amputees and control subjects, removing the effects of
age, yielded no statistically significant elevated risk for
amputation associated with smoking based on "current
smoker" or "ever smoked" status (odds ratio 1.32, 95%
confidence interval [Cl] 0.70, 2.97 and odds ratio 0.75,
95% Cl 0.41, 1.37, respectively) (29).
Smoking has been described as a major risk factor for
development of atherosclerosis in nondiabetic popula-
tions. This description has been applied to diabetic pop-
ulations, although based on less rigorous data. Although
519
CAUSAL PATHWAYS TO AMPUTATION
smoking, hypertension, and lipoprotein abnormalities
have been associated with development of atheroscle-
rosis, which is more common in diabetes, factors other
than atherosclerosis and ischemia participate in the pro-
gression of limb pathology in many cases that culminate
in amputation of the diabetic limb. Causal pathway data
presented herein suggest that even though ischemia was
a causal factor in ~50% of the amputations, other fac-
tors were more often a component cause or the final
component in the amputation pathway.
There are limitations inherent in the analysis of causal
pathways to define sufficient causes for diabetic limb
amputation. The ascertainment of causes was not en-
tirely objective, although the method was designed to
use quantitative objective data as the basis for the sub-
jective synthesis of component factors leading to a con-
sensus designation of the sufficient cause for each in-
dividual amputation. However, because control data
were not considered in the assignment of causal path-
ways, the certainty of results could not be described
statistically, e.g., with CIs, as would be done for con-
trolled epidemiological investigations.
Several features of this study, however, support the
validity of the observed frequency distribution of causal
pathways. The cases comprise a consecutive series of
incident amputations accumulated over 2.5 yr at the
same medical center that serves veterans within a de-
fined catchment region. The subjective considerations
by the investigators can be viewed as an integration of
the pertinent objective data within a framework of rel-
evant clinical experience. This permitted the definition
of causal pathways that are substantially intuitive, log-
ical, and germane to a real-world clinical context. The
study design considered important potential causes, i.e.,
pivotal episodes of minor trauma, and delineated sig-
nificant sequences of causal events that would be dif-
ficult to measure or evaluate in other study designs. Al-
though causal pathway analysis does not permit
assessment of the relative risk from independent expo-
sures, it does suggest the proportion of amputations likely
to accrue from various common constellations of com-
ponent causes, many of which could not be readily
specified in case-control studies. We calculated the
population-attributable risk percent, a standard objec-
tive epidemiological measure, for parameters of is-
chemia and neuropathy, and obtained internal valida-
tion of the amputation proportions estimated for those
component causes from their cumulative participation
in causal pathways.
Interventions incorporating these observations now
need to be assessed in controlled prospective interven-
tion trials to prevent amputations in diabetic individuals
in various populations and clinical settings.
ACKNOWLEDGMENTS
This work was supported by Veterans Affairs Rehabili-
tation R&D, Veterans Affairs Health Services R&D,
520
Washington State Diabetes Control Program, and Dia-
betes Research and Education Foundation.
REFERENCES
1. Most RS, Sinnock P: The epidemiology of lower extremity
amputations in diabetic individuals. Diabetes Care 6:87-
91, 1983
2. Ebskov B: Trends in lower extremity amputation (Den-
mark, 1978-1983). In Surgery and Lower Limb Prosthet-
ics. Murdock G, Donovan RG, Eds. Oxford, UK, Black-
well, 1988, p. 3-8
3. Liedberg E, Persson BM: Increased incidence of lower
limb amputation for arterial occlusive disease. Acta Or-
thop Scand 54:230-34, 1983
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
4. Kacy S, Wolma FJ, Flye MW: Factors affecting the results
of below knee amputation in patients with and without
diabetes. Surg Cynecol Ofastet 156:513-18, 1982
5. Towne JB: Management of foot lesions in the diabetic
patient. Vase Surg 63:669-91, 1984
6. Gibbons GW: The diabetic foot: amputation and drainage
of infection. ) Vase Surg 5:791-93, 1987
7. Nelson RG, Gohdes DM, EverhartJE, Hartner JA, Zwemer
FL, Pettitt DJ, Knowler WC: Lower-extremity amputations
in NIDDM: 12-yr follow-up study in Pima Indians. Dia-
betes Care 11:8-16, 1988
8. Rothman KJ: Modern Epidemiology. Boston, MA, Little,
Brown, 1986
9. Rothman KJ: Causes. Am / Epidemiol 104:587-92, 1976
10. Levin ME: The diabetic foot: pathophysiology, evalua-
tion, and treatment. In The Diabetic Foot. 4th ed. Levin
ME, O'Neal LW, Eds. St. Louis, MO, Mosby, 1988, p. 1 -
50
11. Kozak GP: Diabetic foot disease: a major problem. In
Management of Diabetic Foot Problems. Kozak GP, Hoar
CS, Rowbotham JL, Wheelock FC, Gibbons GW, Camp-
bell D, Eds. Philadelphia, PA, Saunders, 1984, p. 1-8
12. Boulton AJM: The diabetic foot. Med Clin North Am
72:1513-30, 1988
13. Goodson WH, Hunt TK: Wound healing and the diabetic
patient. Surg Cynecol Obstet 149:600-608, 1979
14. Wyss CR, Matsen FA, Simmons CW, Burgess EM: Trans-
cutaneous oxygen tension measurements on limbs of di-
abetic and nondiabetic patients with peripheral vascular
disease. Surgery 95:339-46, 1984
15. YaoST, HobbsJT, Irvine WT: Ankle systolic pressure mea-
surements in arterial disease affecting the lower extremi-
ties. Brj Surg 56:676-79, 1969
16. Cutajar CL, Marston A, Newcombe JF: Value of cuff oc-
clusion pressures in assessment of peripheral vascular dis-
ease. Br Med J 2:393-95, 1973
17. Carter SA: Role of pressure measurements in vascular dis-
ease. In Noninvasive Diagnostic Techniques in Vascular
Disease. Bernstein EF, Ed. St. Louis, MO, Mosby, 1985,
p. 513-44
18. Wyss CR, Harrington RM, Burgess EM, Matsen FA III:
Transcutaneous oxygen tension as a predictor of success
after an amputation. / Bone joint Surg 70:203-207, 1988
19. Linstone HA, Turoff M: The Delphi Method: Techniques
and Applications. Reading, MA, Addison-Wesley, 1975
20. Armitage P, Berry G: Statistical Methods in Medical Re-
search. 2nd ed. Oxford, UK, Blackwell, 1987, p. 469-
72
DIABETES CARE, VOL. 13, NO. 5, MAY 1990

21. Olerud JO, Pecoraro R, Burgess E, McKnight B, Wyss CR,
Reiber GE, Matsen FA III: Reliability of transcutaneous
oxygen tension (TCPO2) measurements in elderly normal
subjects. Scand) Cm Lab Invest 47:535-41, 1987
22. Labbe RF, Rettmer RL: Zinc protoporphyrin: the product
of iron-deficient erythropoeisis. Semin Hematol 26:40-
46, 1989
23. Metropolitan Life Insurance Company: 1983 height and
weight tables. Stat Bull 64:2-9, 1983
24. Edmonds ME, Blundell MP, Morris ME, Thomas EM, Cot-
ton LT, Watkins PJ: Improved survival of the diabetic foot:
the role of a specialized foot clinic. Q / Med 60:763-71,
1986
25. Knighton DR, Ciresi KF, Fiegel VD, Austin LL, Butler EL:
Classification and treatment of chronic nonhealing
DIABETES CARE, VOL. 13, NO. 5, MAY 1990
R.E. PECORARO, G.E. REIBER, AND E.M. BURGESS
wounds: successful treatment with autologous platelet-
derived wound healing factors (PDWHF). Ann Surg
204:322-33, 1986
26. National Diabetes Advisory Board: The National Long-
Range Plan to Combat Diabetes. Washington, DC, U.S.
Govt. Printing Office, 1987 (DHEW publ. no. 87-1587)
27. Lithner F, Tornblom N: Gangrene localized to the lower
limbs in diabetes. Acta Med Scand 208:315-20,
1980
28. Shopland DR, Brown C: Toward the 1990 objectives for
smoking: measuring the progress with 1985 NHIS data.
Public Health Rep 102:68-73, 1987
29. Reiber GE: A Case-Control Study of Diabetic Individuals
to Identify Risk Factors Associated With Lower Extremity
Amputations. Ann Arbor, Ml, Univ. Microfilms, 1989
Downloaded from http://diabetesjournals.org/care/article-pdf/13/5/513/439750/13-5-513.pdf by guest on 24 August 2023
521