SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Objective

A high frequency of infections complicating rheumatoid arthritis (RA) has been described in reports of case series. This retrospective longitudinal cohort study was undertaken to compare the frequency of infections in a population-based incidence cohort of RA patients with that in a group of individuals without RA from the same population.

Methods

RA patients included all members of an incidence cohort of Rochester, Minnesota residents ages ≥18 years who were first diagnosed as having RA between 1955 and 1994. One age- and sex-matched subject without RA was selected for each patient with RA. Study subjects were followed up by review of their entire medical record until death, migration from the area, or study end (January 1, 2000), and details of all documented infections, along with information on potential risk factors for infection, were recorded. Hazard ratios for infections were estimated using stratified Andersen-Gill proportional hazards models, with adjustment for potential confounders.

Results

The 609 RA patients and 609 non-RA study subjects (mean age 58.0 years; 73.1% female) were followed up for a mean of 12.7 years and 15.0 years, respectively, reflecting higher mortality among the group with RA. Hazards ratios for objectively confirmed infections, infections requiring hospitalization, and any documented infection in patients with RA were 1.70 (95% confidence interval [95% CI] 1.42–2.03), 1.83 (95% CI 1.52–2.21), and 1.45 (95% CI 1.29–1.64), respectively, after adjustment for age, sex, smoking status, leukopenia, corticosteroid use, and diabetes mellitus. Sites of infection with the highest risk ratios were bone, joints, skin, soft tissues, and the respiratory tract.

Conclusion

In this study, patients with RA were at increased risk of developing infections compared with non-RA subjects. This may be due to immunomodulatory effects of RA, or to agents with immunosuppressive effects used in its treatment.

A high frequency of infections complicating rheumatoid arthritis (RA) has been reported during the last 40 years. In particular, high rates of septic arthritis and pulmonary infections have been described (1, 2). Reports in the literature suggested, even in the pre-steroid era, that patients with RA may have an increased susceptibility to infection (3). In further support of the notion of increased infection risk in patients with RA is the finding that up to 40% of patients with septic arthritis have RA (4). Patients with RA have increased mortality compared with the general population, and at least part of this excess mortality appears to be due to infectious diseases, in particular, genitourinary and bronchopulmonary infections (5–8). Recently, the question of whether patients with RA are at increased risk of developing infections compared with the general population has gained interest in light of reports of serious infections in patients receiving biologic therapies for the disease (9, 10). The objectives of the present study were to characterize the occurrence of infections in members of a population-based RA incidence cohort and to determine whether these individuals are at increased risk of developing infections compared with age- and sex-matched subjects without RA from the same general population.

PATIENTS AND METHODS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Study population.

We performed a retrospective longitudinal cohort study comparing infection rates in a group of patients with RA with those in a group of individuals without RA. The former group included the members of a population-based RA inception cohort (11). These RA cases were identified using the data resources of the Rochester Epidemiology Project, a diagnostic indexing and medical records linkage system that exists at the Mayo Clinic and affords access to medical records from all sources of care for community residents. Population-based epidemiologic research in Rochester, Minnesota is possible because of its relative geographic isolation from other urban centers and the fact that nearly all medical care is delivered to local residents by a small number of providers. These providers include the Mayo Clinic, Olmsted Medical Group, and a few private practitioners. Each provider uses a comprehensive medical record system whereby all data collected on an individual are assembled in a single record. Medical records are available for all residents dating back to 1910. Medical diagnoses and other key information are routinely abstracted in a summary record (“master sheet”) and entered into computerized indices. The medical records linkage system composed of these indices facilitates identification of all cases of a given condition. Thus, this system ensures virtually complete ascertainment of all clinically diagnosed cases in a geographically defined community. The potential utility of this data retrieval system for population-based studies has been described in detail previously (12, 13).

To identify individuals with RA for this study, all potential cases of RA were identified by searching the computerized diagnostic index of the Rochester Epidemiology Project for any diagnosis of arthritis (excluding degenerative arthritis or osteoarthritis) made between January 1, 1955, and December 31, 1994, among Rochester residents ≥18 years of age. The complete medical records were reviewed by a team that included 3 trained nurse abstractors and 1 physician (MFD), using a pretested data collection form. The diagnosis was confirmed or rejected based on the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 diagnostic criteria for RA (14).

To select non-RA study subjects, each case was individually matched to 1 randomly selected Rochester resident with no diagnosis of inflammatory arthritis, who was of the same age (±3 years) and sex and had a similar length of enrollment in the records linkage system. This ensured that the non-RA subjects in the comparison group had a similar duration of contact with the system prior to incidence date, and were thus matched with the RA patients by calendar year. Non-RA study subjects were assigned an index date corresponding to the incidence date of their matched case.

All study subjects were followed up through their entire (inpatient and outpatient) community medical record, until death, migration from the county, or the date of study end (January 1, 2000).

Data collection.

Study personnel collected the data according to a prespecified and pretested detailed protocol. Three of the 4 abstractors were not aware of the study hypotheses, the exception being the physician-abstractor (MFD), who reviewed a small proportion of the records. Reliability testing was carried out at the outset of the abstraction process and again at a point midway through the data collection: a sample of medical records was reviewed by all abstractors to ensure good interobserver and intraobserver agreement. Regular meetings were held throughout the abstraction period to identify and correct problems in data collection, interpretation of definitions, and application of study criteria. Before commencing data analysis, we performed an extensive series of checks for data consistency, proper sequences of dates, and an evaluation of missing or incomplete data. If necessary, medical records were reviewed again, and questions were resolved by consensus of the investigative team.

Data on all episodes of infection requiring medical care occurring after the incidence date (index date for subjects without RA) were collected. Information on minor upper respiratory tract infections was not collected. The primary outcome measure was infections with objective confirmation, defined as those with positive results of either microbiologic cultures or radiologic imaging, or both. Secondary outcome measures included infections that required hospitalization and all physician-documented infections.

The operational definitions for each infection type were as follows: bacteremia/septicemia, isolation of a pathogenic microorganism from 1 or more blood cultures, with fever (>38.0°C); septic arthritis, positive microbiologic culture from joint aspirate fluid in the presence of suggestive clinical features; urinary tract infection, including pyelonephritis and urosepsis, isolation of >100,000 colony-forming units/ml of urine in the presence of suggestive clinical features; pneumonia, presence of new infiltrates, consolidation, or effusion seen by chest radiography and suggestive clinical features; osteomyelitis, clinical suspicion with confirmation by definite radiologic findings or positive bone culture. Lower respiratory tract infections, skin and soft tissue infections, and acute gastroenteritis could be included on the basis of a physician's diagnosis and relevant clinical findings alone, but microbiologic culture results were recorded if available. Skin and soft tissue infections included cellulitis, abscesses, wound infections, herpes zoster, and diabetic foot infections. Intra-abdominal infections could be included on the basis of clinical findings alone, and comprised acute cholecystitis, ascending cholangitis, suppurative appendicitis, and peritonitis. The category “other infections” included episodes of otitis media and sinusitis that required hospitalization, eye infections, male and female genital tract infections, culture-proven tuberculosis, and acute hepatitis.

For each episode of infection, we collected information on accompanying fever, leukocytosis, and findings of relevant investigations, including microbiologic cultures and radiologic findings. We also recorded whether the infection required hospitalization, and length of hospital stay. In the case of urinary tract infections (other than those classified as urosepsis/acute pyelonephritis), we recorded only the total number of culture-positive infections. Patients who fulfilled criteria for more than 1 infection simultaneously were classified in both categories, except in the case of septicemia, which was classified in a single category referred to as septicemia with a notation of the accompanying infectious condition (e.g., pneumonia with septicemia, urinary tract infection with septicemia).

Information on potential confounding factors for infection (diabetes mellitus, leukopenia, neutropenia, smoking status, alcoholism, chronic lung disease, and corticosteroid use), along with dates of onset, was ascertained through medical record review.

Data analysis.

Baseline characteristics of the study population were summarized using descriptive statistics. The significance of the difference in mortality rates between patients and non-RA study subjects was tested using the log rank test. Incidence rates for infections were calculated by dividing the total number of events by the number of person-years of followup. Rate ratios were obtained by dividing infection incidence rates in RA patients by those in subjects without RA, and 95% confidence intervals (95% CIs) for these rate ratios were calculated as described by Cox (15).

The risks of infections between individuals with and those without RA were compared using stratified Andersen-Gill proportional hazards models (16). With these models, we were able to examine all episodes of infection occurring in each study subject. Included in these analyses were all types of infection with the exception of minor urinary tract infections, which were excluded because the dates of each of these infections in each patient were not available. A forward stepwise selection process (2-sided, α = 0.05) was used to create an optimal multivariate model of potential confounding variables. For each outcome, the hazard ratio comparing RA patients with non-RA study subjects and the 95% CI was then estimated, with adjustment for potential confounders. These potentially confounding variables were included as time-varying covariates in the analyses, with the one exception of smoking status, which was recorded at baseline only.

RESULTS

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The RA incidence cohort comprised 609 individuals, each of whom was matched to 1 (non-RA) study subject. The mean age at incidence was 58.0 years for RA patients and 58.2 years for non-RA subjects. The majority of subjects in each group were female (73.1%). The mean followup time in the non-RA comparison group (15.0 years) was longer than that in the RA group (12.7 years); this was due to the significantly higher mortality rate among RA patients (P < 0.001). Total followup in cases and controls was 7,729.7 and 9,132.7 person-years, respectively. RA patients and unaffected subjects were similar at baseline in terms of age, sex, and variables that might increase infection susceptibility, including diabetes mellitus, chronic lung disease, alcoholism, and leukopenia (Table 1), but subjects with RA were more likely to be cigarette smokers and to have had corticosteroid exposure.

Table 1. Comparison of the frequency of potential confounding variables among rheumatoid arthritis (RA) and non-RA subjects
VariableRA patients at incidence date, no. (%) (n = 609)Non-RA subjects at incidence date, no. (%) (n = 609)RA patients after incidence date, no. (rate/100 person-years)*Non-RA subjects after incidence date, no. (rate/100 person-years)*
  • *

    Includes only individuals who did not have the condition prior to incidence date. NA = not applicable.

  • Smoking status was unknown in 35 cases and 21 controls.

  • Defined as a total white blood cell count of <3.5 × 109. Leukopenia was never tested in 3 cases and 14 controls (assumed normal).

Female445 (73.1)445 (73.1)NANA
Ever smoker317 (55.2)273 (46.4)NANA
Diabetes mellitus25 (4.1)35 (5.7)38 (0.52)57 (0.69)
Chronic lung disease67 (11.0)62 (10.2)91 (1.41)61 (0.76)
Alcoholism13 (2.1)20 (3.3)29 (0.39)12 (0.14)
Leukopenia3 (0.5)1 (0.2)99 (1.44)64 (0.73)
Corticosteroid use65 (10.7)3 (0.5)181 (2.53)63 (0.69)

Infections with objective confirmation.

For infections with objective confirmation (i.e., positive culture or imaging result), the overall rate of infection per 100 person-years was higher in RA patients (19.64) than in non-RA subjects (12.87). The rate ratio for developing infections in patients with RA was 1.53 (95% CI 1.41–1.65), indicating a significantly elevated risk of ∼50% in these subjects compared with those without RA. In fact, the rate of infection in RA patients was higher than that in the non-RA group in each of the 11 infection categories examined (Table 2). The difference in all but 2 of these rate ratios (urosepsis/pyelonephritis and gastroenteritis) reached statistical significance. Infection sites that were associated with the highest rate ratios were the joints (rate ratio for septic arthritis 14.89 [95% CI 6.12–73.71]), bone (rate ratio for osteomyelitis 10.63 [95% CI 3.39–126.81]), and skin and soft tissues (rate ratio 3.28 [95% CI 2.67–4.07]). We also calculated the relative risks for infections overall after excluding all episodes of septic arthritis and osteomyelitis, and found that the risk was marginally lower, with incidence in cases being 19.05/100 person-years and that in controls being 12.84/100 person-years. The resulting rate ratio was 1.48 (95% CI 1.37–1.60). Of the 31 cases of septic arthritis in RA patients, 11 (35%) occurred in prosthetic joints. There was 1 case of tuberculosis in the RA group and 2 cases in the non-RA group.

Table 2. Objectively confirmed infections in 609 rheumatoid arthritis (RA) and 609 non-RA subjects*
Infection typePatients, no.Infections, no.Incidence/100 person-years (all events/person)Rate ratio95% confidence interval
RANon-RARANon-RARANon-RA
  • *

    Defined as infections with either a positive microbiologic culture or relevant radiologic finding.

  • Obtained by dividing infection incidence rates in RA patients by those in non-RA subjects.

  • Calculated by the method of Cox (15).

Total3893431,4811,13719.6412.871.531.41–1.65
Bacteremia/septicemia533960470.780.511.501.10–2.08
Septic arthritis2223120.400.0214.896.12–73.71
Osteomyelitis1111310.170.0110.633.39–126.81
Pneumonia1791353112184.022.391.681.46–1.95
Lower respiratory tract523583521.070.571.881.41–2.53
Urinary tract infections2342246586628.727.491.161.05–1.30
Urosepsis/pyelonephritis282938400.490.441.120.77–1.63
Skin/soft tissue13259231832.990.913.282.67–4.07
Gastroenteritis871080.130.091.460.68–3.28
Intra-abdominal1771770.220.082.761.39–6.22
Other231529170.380.191.991.22–3.36

Infections requiring hospitalization.

Infections requiring hospitalization were also significantly more frequent in RA patients (9.57/100 person-years) than in non-RA study subjects (5.09/100 person-years), with a rate ratio of 1.88 (95% CI 1.71–2.07). When infections were classified by site, RA patients again had a significantly higher risk of developing infections, in all but 3 sites (Table 3). The risk pattern was similar to that seen with objectively confirmed infections, with septic arthritis, osteomyelitis, and skin and soft tissue infections being associated with the highest rate ratios (21.66 [95% CI 7.37–257.61], 10.63 [95% CI 3.39–126.81], and 2.76 [95% CI 2.22–3.47], respectively).

Table 3. Infections requiring hospitalization in 609 rheumatoid arthritis (RA) and 609 non-RA subjects*
Infection typePatients, no.Infections, no.Incidence/100 person-years (all events/person)Rate ratio*95% confidence interval
RANon-RARANon-RARANon-RA
  • *

    Obtained by dividing infection incidence rates in RA patients by those in non-RA subjects.

  • Calculated by the method of Cox (15).

  • Any urinary tract infection that required hospitalization was classified as urosepsis/pyelonephritis. NA = not applicable.

Total2902217404659.575.091.881.71–2.07
Bacteremia/septicemia533960470.780.511.501.10–2.08
Septic arthritis2012710.350.0121.667.37–257.61
Osteomyelitis1111310.170.0110.633.39–126.81
Pneumonia155992401543.101.691.841.55–2.18
Lower respiratory tract574589661.150.721.591.22–2.08
Urinary tract infectionsNANANANANANANANA
Urosepsis/pyelonephritis272935400.450.441.040.70–1.51
Skin/soft tissue10956183782.370.852.762.22–3.47
Gastroenteritis262638350.490.381.280.87–1.89
Intra-abdominal252826290.340.321.060.68–1.65
Other241229140.380.152.401.44–4.24

All physician-documented infections.

The incidence rate for all physician-documented infections in patients with RA was 32.05/100 patient-years, compared with 24.04/100 patient-years in the non-RA subjects (rate ratio 1.33 [95% CI 1.26–1.41]) (Table 4). For physician-documented infections, the infection sites that were associated with the highest rate ratios in patients with RA were the same as those for infections with objective confirmation and those requiring hospitalization.

Table 4. All physician-documented infections in 609 rheumatoid arthritis (RA) and 609 non-RA subjects
Infection typePatients, no.Infections, no.Incidence/100 person-years (all events/person)Rate ratio*95% confidence interval
RANon-RARANon-RARANon-RA
  • *

    Obtained by dividing infection incidence rates in RA patients by those in non-RA subjects.

  • Calculated by the method of Cox (15).

Total4714562,4172,12432.0524.041.331.26–1.41
Bacteremia/septicemia533960470.780.511.501.10–2.08
Septic arthritis2223120.400.0214.896.12–73.71
Osteomyelitis1111310.170.0110.633.39–126.81
Pneumonia1791353112184.022.391.681.46–1.95
Lower respiratory tract2442276246388.076.991.161.05–1.27
Urinary tract infections2342246586628.727.491.161.05–1.30
Urosepsis/pyelonephritis282938400.490.441.120.77–1.63
Skin/soft tissue2451875373286.953.591.931.72–2.17
Gastroenteritis4988631190.821.300.630.48–0.81
Intra-abdominal282829290.380.321.180.77–1.82
Other423453400.690.441.561.11–2.22

Multivariate analyses.

The hazard ratio for the development of objectively confirmed infections in RA patients compared with non-RA subjects, after adjustment for confounding variables, was 1.70 (95% CI 1.42–2.03). These confounding variables were leukopenia, diabetes mellitus, and chronic lung disease, each of which was a statistically significant predictor of infection occurrence (P < 0.001 for each). There were no significant interactions between these variables. In multivariate analyses of the secondary outcome measures, infection that required hospitalization and any physician-documented infection, the hazard ratios for RA patients developing an infection during followup were 1.83 (95% CI 1.52–2.21) and 1.45 (95% CI 1.29–1.64), respectively.

To investigate whether the relative risk for infection has changed in more recent years, we calculated the relative risks among patients with incident RA in each of the 4 decades. The risk for infections in all 4 decades was significantly higher in RA patients compared with non-RA subjects. The confidence intervals for all of these relative risks overlapped, indicating that the excess risks were not significantly different according to decade. We also examined the infection rate per 100 person-years over time, and found that infection rates rose similarly in cases and controls. This overall increase likely reflects improvements over time in the ability to diagnose infections.

DISCUSSION

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

Our results indicate that persons with RA are at twice the risk of developing an objectively confirmed infection compared with age- and sex-matched individuals in the same community who do not have RA. This excess risk is present to a varying degree for all infections examined and if infections requiring hospitalization and all documented infections are used as outcome measures. Furthermore, the magnitude of infection risk for patients with RA was greater after adjustment for potential confounders. The sites of infection that are associated with the highest rate ratios when any of these outcomes are used are the bone, joints, skin, and soft tissues.

Only 2 published controlled studies have examined the risk of infection in RA (17, 18). Neither of these studies, which were performed at the same center, showed a difference in infection rates between cases and controls. The first of these studies demonstrated an odds ratio of 0.83 (95% CI 0.57–1.2) for genitourinary infections and an odds ratio of 0.61 (95% CI 0.39–0.96) for bronchopulmonary infections, in patients with RA compared with control subjects with either osteoarthritis or soft tissue rheumatism (17). The second study examined 448 RA patients and 185 control subjects and showed that at least 1 infection occurred in 101 RA patients (23%) versus 50 controls (27%), and that the types of infection in the 2 groups were similar (18).

Major methodologic differences may explain the discrepancy between our results and those in these 2 previous studies. One methodologic difference is that the case subjects in the 2 earlier controlled studies were prevalent cases of RA, selected from a specialized treatment center. Such cases may not be representative of RA patients in the community, since patients whose disease is in remission or those who are seen only in a primary care setting will not be included. Furthermore, the use of prevalent cases of RA may lead to underrepresentation of the number of patients with fatal infections, who died early in their disease course. Our study avoids this possibility of incidence-prevalence bias by including all incident cases of RA in a geographically defined population. Also, the control subjects in the previous studies were selected from rheumatology outpatient clinics, which they attended for osteoarthritis or soft tissue rheumatism. These chronic illnesses may have rendered the controls more susceptible than healthy individuals to infections. In our study, matched controls were selected at random from the same general population as the cases.

Another methodologic difference is that the previous investigators used questionnaire and interview techniques to ascertain infection occurrence, and thus relied on the study subjects' recall. In our study, we reviewed each patient's complete medical history, including that from primary care. Therefore, we did not rely on symptom reporting by patients and controls, which is subject to the possibility of differential recall between groups. We were also able to obtain objective confirmatory evidence of infections by examining results of microbiologic cultures and radiographic imaging. Use of medical records also enabled us to obtain information about a broad range of infection sites, whereas the previous studies were limited to genitourinary, bronchopulmonary, skin, and “miscellaneous” infections.

Unlike the earlier controlled studies, ours had an extended length of followup after RA incidence. This prolonged followup, combined with our relatively large sample size, yielded a total of 7,729.7 person-years of followup in subjects with RA and 9,132.7 person-years in the non-RA subjects. The short followup period (1 year) in one of the previous studies (18) may not reflect the long-term risk of infection in patients with RA.

In addition to the 2 controlled studies discussed above, occurrence of infections in patients with RA has been described in reports of several case series (1, 3, 19–22). In 1 such report, based on a study of 195 consecutive RA patients seen in an outpatient clinic, the estimated rate of incidence of all infections was 17 new infections per 100 patient-years of followup (21). This is comparable with the rate of 19.64 infections per 100 patient-years that we demonstrated in the present study using the definition of objectively confirmed infections, which is closest to the definition used in that study.

In most case series reports in which infections in RA patients are described, the sites most frequently affected are the joints, respiratory tract, and skin (1, 19, 20). In a recent report documenting the incidence of infections requiring inpatient hospitalization in a series of RA patients, 32.3% of all such infections were pulmonary, 24.1% involved skin and wounds, and 17.8% involved bone and joints (22). These sites correspond to those with the highest rate ratios for patients with RA in our study. Our findings thus support earlier evidence that RA patients are more susceptible to infections of bone, joints, skin and soft tissues, and the respiratory tract.

All previous reports of infection occurrence in RA are based on studies of patients seen in hospitals, often secondary or tertiary referral centers. These studies do not provide a reliable estimate of the frequency of infections in patients with RA in the community. In the present study, we characterized the occurrence of infections in a population-based cohort of RA patients and obtained data on the frequency and sites of infections in these individuals. This should facilitate comparisons with rates of infection in RA patients who are receiving biologic and other recently introduced immunomodulatory therapies (9, 10).

There are a number of possible explanations for an increased risk of infection in patients with RA. Recent evidence suggests that RA patients have immunologic abnormalities involving the majority of circulating T cells, from an early stage in the disease course (23). Thus, the ability of the immune system to respond to novel antigenic stimuli may be compromised. Alternatively, therapy with corticosteroids and other immunosuppressive medications may also predispose RA patients to the development of sepsis (24, 25). Other factors that may influence infection risk in patients with RA are disease-related factors (immobility, joint surgery), extraarticular manifestations of RA (Felty's syndrome, rheumatoid lung disease), and comorbidities (diabetes mellitus). The relative contribution of each of these to infection risk has not been established.

One limitation of the present study is that only infections that came to medical attention could be included. For this reason, we did not include infections of the upper respiratory tract. Another limitation is that we did not record details of herpes zoster infections separately from other skin infections, so we are unable to compare rates with those in other studies that have shown high rates of this infection in patients with RA. RA patients might be more likely to seek medical care when they develop minor infections, which may lead to relative overreporting in cases. However, our findings of a higher magnitude of risk for development of infections with objective confirmation and for infections serious enough to require hospitalization in patients with RA are evidence against the existence of such a bias. Finally, because some racial and ethnic groups are underrepresented in Rochester, Minnesota, where the population in 1990 was 96% white according to US Census data, the results of our population-based study are generalizable only to the US white population.

In conclusion, we have shown that patients with RA have nearly twice the rate of infection compared with matched non-RA controls, and that this excess risk is present for objectively confirmed infections, infections requiring hospitalization, and all physician-diagnosed infections. The higher frequency of infection in RA cases compared with controls might be related to the disease itself, through either altered immunologic function or other factors such as decreased mobility and skin defects. These results underscore the need for additional research to discover the determinants of this increased infection risk in RA.

Acknowledgements

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES

The authors thank Denise Herman, RN, Deanne Stiebner, RN, and Patricia Hartkopf, RN, for their assistance with data collection, and Deborah Fogarty for assistance in preparation of the manuscript.

REFERENCES

  1. Top of page
  2. Abstract
  3. PATIENTS AND METHODS
  4. RESULTS
  5. DISCUSSION
  6. Acknowledgements
  7. REFERENCES