Unchanged Incidence of Lower-Limb Amputations in a German City, 1990–1998

A detailed description of the database and the methods used, as well as an analysis of incidence rates, and relative and attributable risks of subjects with amputations in 1990 and 1991 was published previously (2). Briefly, from all three hospitals with surgical departments in Leverkusen, we obtained complete lists of lower-limb amputations performed in 1990 and 1991 and in 1994–1998 by analyzing operation theater documentation. The data of 1992 and 1993 were not included in the analysis because operation records from those years were incomplete. We then reviewed the hospital records of each patient. Only patients who were city residents were included. We determined date of birth, address, sex, amputation levels, dates of operations, whether diabetes was known to be present, and—if applicable—diabetes duration. When more than one amputation was carried out in a patient within the observation period, only the first observed amputation was counted for the analysis. Patients with previous amputations were excluded from the analysis if the known date of the first previous amputation lay outside of the observation period or was ≥1 year before the first observed amputation. We obtained population data from the city administration. The total population of the study area as of 31 December 1990 was 160,684.

The population with diabetes in each stratum was estimated by multiplying the population of the study area in the stratum by the age- and sex-specific prevalence of diabetes in East Berlin in 1988, obtained from the former East German diabetes registry (2,15), because these are the only age-specific and reliable prevalence data available for the German population. (Until 1990, a centralized state-run health care system existed in East Germany. Every person who was diagnosed with either type 1 or type 2 diabetes was reported with a basic data set to a regional diabetes office that kept the registry.)

The following stratum-specific and directly standardized incidence rates (standard: West German population of 1991) were estimated per calendar year (1990, 1991, and 1994–1998) and expressed per 100,000 person-years of observation: total incidence of amputations in residents of the study area, incidence of amputations in the estimated nondiabetic population, and incidence in persons with diabetes per estimated diabetic population. The rate in the diabetic population was also standardized to the estimated West German diabetic population of 1991 because such a rate is the only one shown in some publications.

On the basis of standardized incidence rates, we estimated relative risks of amputations for people with diabetes compared with people without diabetes, attributable risks among those exposed to diabetes, and population-attributable risks. To test for time trend, we fitted a Poisson regression model using year of registration (difference from the first year as an ordinal variable), age (categorized into four classes), diabetes, and sex as independent variables. In addition, we fitted separate Poisson models for diabetic and nondiabetic patients (16). The statistical analysis using Poisson models was repeated including only amputations above the toe level and major amputations (i.e., those above the ankle). Calculations were carried out with the SAS statistical package (version 6.12).

During the defined period, 375 patients who were residents of Leverkusen had nontraumatic lower-limb amputations in the three hospitals of Leverkusen, but 36 of them were excluded due to previous amputations. The distribution of the remaining 339 persons with respect to age, sex, diabetic status, and calendar year is shown in Tables 1 and 2. Due to the exclusion of patients with previous amputations from this analysis, the combined number of patients from 1990 and 1991 is slightly smaller than that used in our first publication (2). Of all subjects, 46.0% were female and 75.5% of all subjects were known to have diabetes. Of the diabetic patients, 180 were likely to have type 2 diabetes; only 1 patient was classified as having type 1 diabetes. In 75 cases, sufficient information on the type of diabetes was not available. There were 145 patients who were known to take insulin. Mean age was 71.3 years (SD 10.8; range 33–98). Mean age of women was 74.5 years (10.1); of men, 68.6 years (10.6). Diabetes duration could be obtained for 64 subjects (mean 16.1 years, SD 10.3, median 15, range 0–55). Amputation levels (highest level in the case of more than one amputation) were as follows: toe, 102; forefoot, 80; lower leg, 52; knee, 4; thigh, 100; and hip, 1.

Standardized incidence rates for each year are shown in Table 3. (Diabetic rates standardized to the general population are lower than diabetic rates standardized to the diabetic population because, in the latter case, higher weights are assigned to the older age-groups.) When all years are combined and standardized results are considered, the risk of having an amputation was 26-fold (CI 17–39) in the diabetic population compared with the nondiabetic population. Amputation rates per 100,000 person-years (standardized to the female or male populations of West Germany of 1991, respectively) were 154 (117–192) in diabetic women and 311 (150–472) in diabetic men. Incidence rates of major amputations (above the ankle, per 100,000 person-years, all years combined, with the general population as the standard) were 15 (13–18) in the general population and 79 (62–97) in the diabetic population. When the diabetic population was used as the standard, the incidence rate was 227 (187–267) in the diabetic population. Moreover, 96% (CI 94–97) of the amputation risk in diabetic individuals and 70% (61–77) of the amputation risk in the entire population were due to diabetes. Upon first inspection, despite some variability of incidence rates between years, no trend over time can be observed (Table 3).

The results of the Poisson model are shown in Table 4. According to this model, no change of incident amputations could be observed in the total population. When different Poisson models were calculated for diabetic and nondiabetic patients, almost identical results were obtained for the risk ratio associated with calendar year. When the analyses were repeated including the patients with previous amputations, the pattern of the results did not change. Also, the separate analyses of amputations excluding toe amputations and of major amputations (i.e., those above the ankle) did not show any significant time trend in the general, the diabetic, or the nondiabetic population. (Risk ratio per calendar year in the diabetic population for major amputations: 0.991 [CI 0.930–1.058]; P = 0.78).

We have obtained incidence rates of lower-limb amputations in one German city during most of the years from 1990 through 1998. The results of this study confirm our findings based on the analysis of the data from 1990 and 1991 (2). Great population-attributable risks indicate that improving foot care in people with diabetes appears to be the main target for the reduction of amputations in the general population.

The observation of incidence rates in the study area from 1990 through 1998 has shown that the intended measurable reduction in the number of amputations on a population level has not been achieved despite a number of activities focusing on more effective prevention and treatment of the diabetic foot syndrome. Over 9 years, standardized incidence rates of amputations in the entire population, the diabetic population, and the nondiabetic population remained nearly constant when the expected random variation was taken into consideration. This result was the same when only amputations above the toe level and when only major amputations (i.e., those above the ankle) were included. Therefore, there is no reason to assume that there might have been fewer major amputations and more toe amputations with the total number of amputations remaining constant.

Obviously, this study has some limitations. We had to rely on the information recorded in the hospital records. Possible sources of bias that might conceal an existing time trend must be considered. Some data from previous amputations were probably missing, in particular during the years 1990 and 1994—the first years of data collection in the two study periods. This mistake might be more likely in diabetic subjects because of a higher risk of multiple amputations. This may cause a slight overestimation of the incidence rates in those years. However, the pattern of the results was not sensitive to whether patients with known previous amputations were included or not. One might imagine that a greater proportion of people with amputations have had surgery performed in the local hospitals in recent years than in the past, although there is no empirical basis for such speculations. Another possible source of error is the estimation of the diabetic population from the registry of a different area. Due to the lack of regularly updated data on age- and sex-specific prevalence of diabetes, we had to assume that the stratum-specific prevalence of diabetes remained constant between 1988 and 1998. Prevalence estimates from other sources lack precision and are even contradictory (17,18). If this prevalence increased over this period of time, an actual reduction of the incidence of amputations in the diabetic population would be masked. An overall increase either in the prevalence of diabetes or in the incidence of amputations would be predominantly attributable to the aging of the population. However, since the incidence rates are age-adjusted, it is unlikely that such phenomena introduce considerable bias. Due to incomplete records for 1992 and 1993, we were unable to analyze the data for those years. However, it is unlikely that the pattern of the results would change if those data were included. Other limitations and possible sources of bias have been discussed in depth in our previous article (2).

We studied the incidence of amputations in a restricted area because it is usually impossible to correlate amputations in certain hospitals with a well-defined population. In our previous article, we used a correction factor (dividing by 0.89) to account for the underestimation of amputations due to operations performed outside the study area. In this case, with only minor reservations, the study population may be regarded as an almost complete sample of incident cases of amputations in the study area during the period of observation (2). (Because the focus of this article is the analysis of a potential time trend rather than a comparison of different areas, we show the results without correction factor.) The incidence rates found in our study are similar to those in the south of Germany and in the non–American Indian population of the U.S., are lower than those in Finland and in Native American diabetic populations, but are higher than those found in a recent study from Spain (2,5,6,19).

The lack of a demonstrable reduction in incident amputations might indicate that the goal set in the St. Vincent Declaration was perhaps unrealistic. In patients with type 2 diabetes, lowering of blood glucose or blood pressure alone has not had significant effects on the incidence of amputations (20,21,22). However, substantial reductions in the incidence of amputations have been reported from areas where specific intervention programs for tertiary prevention of the diabetic foot syndrome were carried out (10,11,12,14). In general, the core of these programs was a multidisciplinary team running a hospital-based diabetic foot care clinic. Close collaboration of diabetologists, surgeons, podiatrists, and other health care professionals allows for a well-defined referral procedure with a rapid service. Specialized surgical interventions, e.g., revascularization of the ischemic foot or appropriate debridement for the infected foot, result in a substantial reduction of major amputations shortly after the establishment of such programs (10,11,12,14,23).

To date, no outcome-related impact of the various activities for improvement of diabetes care in general and foot care in particular, as started in the 1990s in our study area, has been demonstrable. The intervention was based primarily on two diabetologists who opened specialized practices for diabetic patients with yearly referrals of ∼25% of the diabetic population in Leverkusen (Dr. Gerhard Willms, personal communication). In addition, they introduced quality circles on diabetes care for family physicians, with about one-fourth of the family physicians in Leverkusen actively participating. It may be possible that these activities targeting the improvement of prevention and treatment of diabetic foot problems may show their impact on the incidence of amputations only after more time has passed. However, with regard to the negative results, a crucial factor could be that during the study period, no multidisciplinary diabetic foot care team was initiated at any of the hospitals in Leverkusen and only one of the hospitals was performing vascular bypass surgery.

The results of our study, together with published data from other areas, indicate that quality-circle discussions on improvements in diabetic foot care and nontargeted measures—despite the best intentions—do not translate into a measurable reduction of lower-limb amputations across a population (6,7,8,9). Instead of targeting only some segments of the health care system (e.g., physician offices, as done in Leverkusen), programs should focus on a comprehensive effort including family physicians, diabetologists, and multidisciplinary foot care teams in specialized hospital units.

We are indebted to the P.-Klöckner-Stiftung (grant to M.B.), Gesundheitsforschung e.V., Janssen-Cilag, GlaxoWellcome, and Ratiopharm for financial support. This research project was temporarily affiliated with the Public Health Research Network North Rhine-Westphalia (Forschungsverbund Public Health Nordrhein-Westfalen).

We thank Christina Loch and Pia Vogler for collecting the raw data, Ulli Monigatti for building the database, and Nina Ennenbach for help with tables and figures and with searching the literature. We thank the following hospitals for giving us access to their patient records: Klinikum Leverkusen (Prof. Vestweber), St. Josef-Krankenhaus Leverkusen (Dr. Schratz), and Remigius-Krankenhaus Opladen (Dr. Voigt).