Letters to the Editor
small fraction of the asthma deaths occurred among these
subjects. Had we had a sufficient number of asthma deaths
in the incident cohort, this analysis would have been performed in our initial study (4).
Nevertheless, our data showed that severe-asthma patients treated with salbutamol were preferentially switched
to fenoterol. We believe that the presence of channeling is
best investigated by measuring disease severity at the onset
of the treatment and not at any point in time during therapy.
REFERENCES
1. Pearce N, Beasley R, Crane J, et al. Re: "Confounding by
indication and channeling over time: therisksof Bj-agonists."
(Letter). Am J Epidemiol 1997; 146:885-6.
2. Beasley R, Burgess C, Pearce N, et al. Confounding by severity does not explain the association between fenoterol and
asthma death. Clin Exp Allergy 1994;24:660-8.
887
3. Blais L, Ernst P, Suissa S. Confounding by indication and
channeling over time: the risks of Bj-agonists. Am J Epidemiol 1996;144:1161-9.
4. Spitzer WO, Suissa S, Ernst P, et al. The use of B-agonists and
the risk of death and near death from asthma. N Engl J Med
1992;326:501-6.
Lucie Blais
Pierre Ernst
Samy Suissa
Department of Epidemiology and
Biostatistics
McGill Pharmacoepidemiology Research
Unit
McGill University
Montreal, Quebec, Canada H3A 1A2
RE: '•HETEROGENEITY OF HIP FRACTURE: AGE, RACE, SEX, AND GEOGRAPHIC PATTERNS OF FEMORAL
NECK AND TROCHANTERIC FRACTURES AMONG THE US ELDERLY"
We have been following with interest the correspondence
concerning the study by Karagas et al. (1) regarding the
correct classification of unspecified fractures of the proximal femur. Both Levy et al. (2) and Karagas et al. (3) were
able to show by validation studies that approximately 85
percent of unspecified fractures are transcervical. In a current analysis of routinely collected hospital discharge diagnoses for the year 1989 in the former German Democratic
Republic, covering 17 million East Germans, we found
further circumstantial evidence for a predominance of
transcervical fractures among unspecified fractures. It is
derived from similarities in in-hospital case fatality between
classified transcervical fractures and unspecified fractures
(table 1). Our analysis is restricted to closed fractures,
which comprise over 97 percent of fractures in this anatomic region.
The case fatality rate for unspecified fractures in this
region (International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), code 820.0) is
compatible with the figures for transcervical fractures (ICD-
9-CM code 820.0) but not with those for pertrochanteric
fractures (ICD-9-CM code 820.2). Overall case fatality appears high, which is partially attributable to the in-hospital
rehabilitation of the East German health care system. The
mean length of hospitalization, including transfers between
wards and hospitals after closed fractures of the proximal
femur, was 60 days.
Figure 1 puts the ratio of closed trochanteric fractures
(ICD-9-CM code 820.2) to closed cervical fractures (ICD9-CM code 820.0, including code 820.8) in perspective with
regard to the reported figures from North America (1, 2).
Despite historically dissimilar political and social environments, similarities in the ratio of transcervical fractures
to pertrochanteric fractures are striking. In accordance with
Levy et al.'s suggestion of common etiologic processes for
Canada and the United States (2), we presume that such
processes are effective in Europe also and may be found at
the biologic level, related to geographic latitude, or related
to industrialization.
REFERENCES
TABLE 1. In-hospltal case fatality for closed proximal
femoral fractures In the German Democratic Republic, 1989
No.
All closed fractures
Pertrocnanteric fractures
(ICD-9-CMt code 820.2)
Transcervical fractures
(ICD-9-CM code 820.0)
Proximal femoral fractures of
unspecified location
(ICD-9-CM code 820.8)
Inhosptal
95%Clt
0)
case
cases*
fatally
9,633
21.7
20.9-22.5
4,087
25.1
23.8-26.4
5,011
19.3
18.3-20.5
Manfred Wildner
535
17.8
14.6-21.3
* Persons aged £50 years only.
t Cl, confidence interval; ICD-9-CM, International Classification
of Diseases, Ninth Revision, Clinical Modification.
Am J Epidemiol
1. Karagas MR, Lu-Yao GL, Barrett JA, et al. Heterogeneity of
hip fracture: age, race, sex, and geographic patterns of femoral
neck and trochanteric fractures among the US elderly. Am J
Epidemiol 1996;143:677-82.
2. Levy AR, Mayo NE, Grimard G. Re: "Heterogeneity of hip
fracture: age, race, sex, and geographic patterns of femoral
neck and trochanteric fractures among the US elderly." (Letter). Am J Epidemiol 1996;144:801-3.
3. Karagas MR, Baron JA, Barrett JA. Re: "Heterogeneity of hip
fracture: age, race, sex, and geographic patterns of femoral
neck and trochanteric fractures among the US elderly." (Letter). Am J Epidemiol 1996;144:803.
Vol. 146, No. 10, 1997
Institute for Medical Informatics,
Biometry, and Epidemiology
Ludwig Maximilians University
Marchioninistrafie 15
D-81377 Munich, Germany
888
Letters to the Editor
X
Quebec
H
Quebec
* unity
women
Age (years)
FIGURE 1. Ratio of transcervical (T) hip fractures to pertrochanteric (P) hip fractures among men and women in the German Democratic
Republic (GDR) (1989), Quebec, Canada (1981-1992), and the United States (1986-1990), by age group. Data for East Germany exclude
open fractures. Data for the United States were obtained from Karagas et al. (1); data for Quebec were estimated from Levy et al. (2).
Unspecified fractures are Included under "neck fractures" for all three geographic regions.
Karl E. Bergmann
Department of Noninfectious
Disease Epidemiology
Robert Koch Institute
General Pape Strafie 62-64
D-12101 Berlin, Germany
THE AUTHORS REPLY
We appreciate the pertinent observations made by
Wildner and Bergmann (1) regarding the similarities in the
occurrence of hip fractures between the German Democratic
Republic (GDR) (1), the United States (2), and Quebec,
Canada (3). Namely, the pattern in the ratio of transcervical
hip fractures to pertrochanteric hip fractures changes over
age for women but remains relatively constant over age for
men. This observation adds more evidence for a common
underlying biologic mechanism operating in the etiology of
hip fractures.
However, there are striking differences in the in-hospital
case-fatality rates and in the length of hospital stay for hip
fracture between Quebec and the GDR. Are these also
driven by biologic differences, or are they related to health
status or health care delivery?
The overall in-hospital case-fatality rate in Quebec was
8.4 percent, as compared with 21.7 percent in the GDR. The
magnitude of this difference is compatible with the discrepancy in all-cause mortality rates observed between several
Eastern European countries, including the GDR, and countries in Western Europe (4). Higher consumption of alcohol
and use of cigarettes (4, 5) in the GDR may be implicated
in the higher case-fatality rates. These factors affect overall
health status and may explain poorer survival following hip
fracture. The intake of calcium in the GDR was lower than
that in West Germany (6), which would probably exacerbate
the risk of hip-fracture but could also conceivably affect the
severity of such fractures.
Alternatively, the causes of deaths occurring after hip
fracture may differ between the GDR and Quebec. An
indirect measure would be the timing of death in relation to
the hip fracture. In Quebec, of persons who died after a hip
fracture, approximately 30 percent died during the first 7
days of hospitalization, about 40 percent died 8-28 days
after hospitalization, and 30 percent died more than 28 days
after hospitalization. These values were relatively constant
over the categories of age, sex, and type of fracture. Deaths
within the first 7 days result from the acute trauma or
perioperative complications, whereas deaths occurring after
7 days could be due to the late effects of immobility (e.g.,
pneumonia, septicemia, or thromboembolism).
Table 1 shows that the case-fatality rates in Quebec were
about half those of the GDR. There were some similarities:
The case-fatality rates were higher for pertrochanteric fractures than for transcervical fractures. In contrast, in Quebec,
the case-fatality rates of persons with hip fractures of unspecified type were higher than those for both pertrochanteric and transcervical fractures among women and were
intermediate among men. Intriguingly, the case-fatality rate
TABLE 1. ln-ho«pltal case fatality rates (%) of persons aged
£50 years wtth hip fracture, by sex and type of fracture,
Quebec, Canada, 1989-1990
Type of fracture
Women
Men
TJanscervtcal
Portrochanterlc
Unspecified
3.8
11.3
7.9
13.7
9.2
10.S
AD
types
6.9
12.5
* There were 3,143 woman and 1,147 men with a primary discharge diagnosis of hip fracture during this period.
Am J Epidemiol
Vol. 146, No. 10, 1997