We investigated whether allergy is associated with increased all-cause mortality. Two allergy markers, peripheral blood eosinophilia (≥275 eosinophilic cells per mm3) and positive skin tests (sum score ≥3), were available for 5,383 subjects of a cohort study on asthma and chronic obstructive pulmonary disease in general population samples of Vlagtwedde and Vlaardingen, the Netherlands, in 1965–1972. During 30 years of follow-up, 1,135 subjects died. In a Cox regression analysis, eosinophilia was associated with an increased risk (relative risk = 1.4; 95% confidence interval = 1.2–1.7) of all-cause mortality, independent of gender, age, smoking habits, percentage predicted forced expiratory volume in 1 second (FEV1 % predicted), and positive skin tests at the start of the study. Subjects with positive skin tests had only an increased risk of all-cause mortality in the subgroup of subjects with FEV1 <80% of predicted (relative risk = 1.7; 95% confidence interval = 1.0–2.8). These results remained essentially unchanged after exclusion of subjects with asthma. We conclude that eosinophilia is associated with increased all-cause mortality. An increased number of peripheral blood eosinophils may reflect an increased inflammatory response, resulting in tissue injury. It is possible that the association between a low FEV1 % predicted and all-cause mortality is partly mediated by an atopic constitution.
From the Departments of 1Epidemiology and Statistics and 2Pulmonology, University of Groningen, the Netherlands, and 3Channing Laboratory, Department of Medicine, Brigham and Women’s Hospital, Harvard Medical School, Boston, MA.
Submitted October 28, 1998; final version accepted September 1, 1999.
Address reprint requests to: Jan P. Schouten, Department of Epidemiology and Statistics, University of Groningen, A. Deusinglaan 1, 9713 AV Groningen, the Netherlands.
Financial support: Netherlands Asthma Fund (grants 187 and 32.96.69) and the Ministry of Health and Environmental Hygiene of the Netherlands.
Atopic diseases are common in Western industrialized populations and have been reported to be increasing. 1 Immediate type I hypersensitivity is the type of allergy involved in these atopic diseases and can be studied by using objective markers, that is, total and specific serum immunoglobulin E levels, peripheral blood eosinophil counts, 2 and skin tests. 3 The relation of these allergy markers to mortality is unknown but of potential public health importance because of the increase in allergy prevalence.
We measured two objective allergy markers, absolute peripheral eosinophil counts and skin tests to common aeroallergens, in a study on asthma and chronic obstructive pulmonary disease in a general population sample in the Netherlands. The purpose of the present analysis is to assess whether eosinophilia and positive skin tests are associated with increased all-cause mortality. The study data also allow us to assess whether the association between eosinophilia or positive skin tests and all-cause mortality is different within various subgroups of gender, age, smoking habits, and lung function and to analyze the association of eosinophils with all-cause mortality independent of allergic sensitization for common aeroallergens. From the literature, we know that both eosinophilia and positive skin tests are associated with asthma. 4,5 Therefore, we investigated whether the association was also present in subjects without asthma.
Subjects and Methods
The Vlagtwedde-Vlaardingen Study is an epidemiologic field study on risk factors for asthma and chronic obstructive pulmonary disease in a random sample of the inhabitants of two Dutch communities. The population selection has been described previously. 6,7 Briefly, in 1965, a survey was carried out in Vlagtwedde, a rural area in the northeast, and in Vlaardingen, an urban area in the western part of the Netherlands, in subjects ages 40–64 years, followed by a survey of younger people ages 15–39 years in 1967 in Vlagtwedde and in 1969 in Vlaardingen. In some cases, data on eosinophils were not available in this first survey but were available in the first follow-up survey in 1970 in Vlagtwedde or in 1972 in Vlaardingen, in which case data from this first follow-up were used (629 cases). A total of 6,324 subjects had both a peripheral blood eosinophil count and skin tests available; 117 subjects were excluded because of missing data on lung function, smoking, or asthma status. Because of the unsatisfactory quality or lack of an assessment of the quality of the spirogram, another 825 subjects were excluded. A total of 5,383 men and women had peripheral blood eosinophil counts, skin tests, and complete data for all other covariates.
Peripheral Blood Eosinophil Count
Peripheral blood eosinophil counts were assessed in a 1:11 dilution of peripheral blood with a Bürker counting chamber. 4,7 Eosinophilia was considered to be present at 275 cells or more per mm3 of blood. This cutoff point was based on the investigation by Veening 8 and was associated with respiratory symptoms generally considered allergic in nature (persistent wheeze and asthma attacks). 4
Allergen Skin Testing
Four allergens were applied intracutaneously to the volar forearm: house dust, mixed pollen, epidermal products, and mixed molds 4,7 (from Diephuis, Groningen, the Netherlands). Wheal diameters for each allergen were measured to the nearest half millimeter and coded on a six-point scale (0 = 0–5 mm, 1 = >5–7.5 mm, 2 = >7.5–10 mm, 3 = >10–12.5 mm, 4 = >12.5–15 mm, and 5 = >15 mm). Scores for the four allergens were added to a skin test sum score (minimum 0, maximum 20). Positive skin tests were defined as having a skin test sum score of ≥3. In a previous analysis, the prevalence of positive skin tests by this definition was found to be associated with respiratory symptoms generally considered allergic in nature (persistent wheeze and asthma attacks). 4 A histamine biphosphate solution was used as a positive control. 7,9 Eight subjects had no data of the positive control, and 391 subjects had no reaction to the positive control. The majority of these subjects were tested in 1965 (379 subjects) and lived in Vlagtwedde (350 subjects).
Data on age, sex, smoking habits, and respiratory symptoms were collected by means of a Dutch version of the British Medical Research Council’s standard questionnaire. 10–12 Interviews were performed by trained interviewers. Ex-smokers had stopped smoking at least 1 month before the examination, and current smokers smoked at least one cigarette per day. Pipe and cigar smokers were also considered smokers. Asthma was considered to be present if an affirmative answer was given to the question whether a subject had ever experienced attacks of shortness of breath with wheezing at rest.
Measurements of forced expiratory volume in 1 second (FEV1) were assessed with a water-sealed spirometer (Lode Spirograph D53, Lode Instruments, Groningen, the Netherlands). Measurement of inspiratory vital capacity after a deep expiration was followed by measurement of FEV1. Subjects performed the maneuver until two technically satisfactory tracings were produced, the highest value of these tracings being taken as baseline measurement. 7,13 Percentage predicted reference values were calculated with regression coefficients derived from analysis of our own population data. In all asymptomatic subjects, regardless of their smoking habits, who took part in 1965/1967/1969, we computed regression equations for FEV1 as a function of age and height, with an age cutoff of 21 years for both sexes separately. The actual measured FEV1 value was expressed as the percentage of the predicted FEV1 (FEV1 % predicted).
Until March 10, 1995, subjects were traced for their vital status (alive, dead, lost to follow-up) with 99% success (4,173 alive and 1,135 dead) by manual review of files; only 75 subjects (1%) were lost to follow-up. Survival time was calculated for each subject from the date of entry into the study (between 1965 and 1972) until the end of follow-up on (1) March 10, 1995, for subjects registered at the municipalities as being alive; (2) the date of death for subjects identified in the Death Register of the municipalities; or (3) the last registration of subjects lost to follow-up, for example, date of last survey attended or date of move when the new address could not be traced.
We calculated mortality rates according to the allergy markers eosinophilia (Table 2A) and positive skin tests (Table 2B) and according to confounders (male gender, age, classes of FEV1 % predicted, and smoking habits). In Table 2C, we used the distribution of the confounders in the eosinophilia group as the standard for standardization of the mortality rates. In Table 2D, we did the same for positive skin tests. The associations between the allergy markers and all-cause mortality were estimated with the proportional hazard model of Cox. 14 Time was defined from the initial examination until death, end of follow-up in 1995, or last registration if subjects were lost to follow-up. A proportional hazards model accounts for varying intervals in follow-up between subjects and allows for control for potential confounding effects of other risk factors. We controlled for gender, age, smoking habits, and FEV1 % predicted, which are known to be associated with positive skin tests (gender and age) and eosinophilia (gender, smoking, and lung function), 9,13 and with all-cause mortality, 15,16 and stratified for these confounders. We categorized gender as 0 (female) and 1 (male). We used age as a continuous variable, expressed per 10-year increase. Smoking was categorized into four categories: nonsmokers, ex-smokers, smoking <15 cigarettes per day, and smoking ≥15 cigarettes per day. We used FEV1 in the analyses as <80% for subjects with reduced and ≥80% of predicted for subjects with (near) normal lung function, based on Lange et al. 17 The latter group was divided into groups with near-normal lung function (FEV1 % predicted 80–100%) and normal lung function (FEV1 % predicted >100%). Finally, we repeated these analyses for subjects without asthma (see Table 4). Asthma was categorized as 0 (no asthma) and 1 (asthma).
Table 1 shows population characteristics at the start of the study. The population was relatively young, with a mean age of 36 years and a large age range (14–65 years at the start of the study). More than half of the population smoked (55%). Twelve per cent (662 subjects) had eosinophilia, and 15% (834 subjects) had positive skin tests. Only 161 subjects (3%) had both eosinophilia and positive skin tests, and 182 subjects had a history of asthma attacks (3%). Table 1 also shows that both the eosinophilia and the positive skin test groups contained more males, more smokers, and more subjects with asthma (8%vs 3% for both groups). A lower FEV1 % predicted was found at the start of the study in the group with eosinophilia than in the group without eosinophilia. The mean age was 6 years lower in the group with positive skin tests than in the group without positive skin tests. Table 1 also shows follow-up data. The average follow-up was 24.8 ± 5.5 years (mean ± standard deviation; range, 0–29.4 years). During follow-up, 1,135 deaths occurred, 175 in the group with eosinophilia and 102 in the group with positive skin tests. A higher percentage of deaths was found in the group with eosinophilia (26%) than in the group without eosinophilia (20%) and fewer deaths occurred in the group with positive skin tests (12%) than in the group without positive skin tests (23%).
Table 2, A and B, shows that the mortality rate in subjects with eosinophilia was 0.0111 per year compared with 0.0082 per year in those without eosinophilia, whereas the mortality rate was 0.0049 per year in those with positive skin tests compared with 0.0092 per year in those without positive skin tests. Table 2C shows that standardized mortality rates were increased for all strata in those with eosinophilia compared with those without eosinophilia, whereas Table 2D shows that standardized mortality rates were decreased for all strata in those with positive skin tests compared with those without positive skin tests. Table 3A shows the results of Cox regression models stratified for important confounders. The relative risks (RRs) for eosinophilia hardly changed for different categories of different confounders, except for smoking habits. In contrast, the reduced mortality risk for those with positive skin tests disappeared after adjustment for important confounders. Furthermore, there was a markedly increased mortality risk associated with positive skin tests in those with an FEV1 % predicted <80%. Table 3B shows that the associations between eosinophilia and positive skin tests were independent, because the RRs associated with both allergy markers separately in the model hardly changed when both allergy markers were included.
Table 4 shows the results of two separate Cox regression models. The first model shows that eosinophilia was associated with a 1.43 increased risk [95% confidence interval (CI) = 1.21–1.68] of all-cause mortality. The association of peripheral blood eosinophilia with all-cause mortality did not differ by gender, age category, smoking group, or level of FEV1 % predicted. In contrast, subjects with positive skin tests only had an increased risk of death (RR = 0.92 × 1.84 = 1.69; 95% CI = 1.04–2.75) compared with subjects without positive skin tests in the subgroup with values of FEV1 % predicted <80%. Male gender, older age, current smoking, and lower FEV1 % predicted at the start of the study were all predictors of all-cause mortality. Results were essentially unchanged when we repeated the analyses excluding subjects with asthma (model 2).
Survival curves show that subjects with eosinophilia had poorer survival than those without eosinophilia (Figure 1). Figure 2 shows that, within the subgroup with FEV1 % predicted <80%, subjects with positive skin tests had greater mortality than those without positive skin tests. In the subgroup with FEV1 % predicted <80%, 69 subjects had positive skin tests; of these, 18 died during follow-up.
The following additional analyses did not alter the results: changing the cutoff point of eosinophilia to as low as 200 or as high as 450 cells per mm3, the presence of positive skin tests in the model as the total sum score (range, 0–20) instead of a dichotomous variable, additional adjustment of residence and its interactions with the allergy markers, inclusion of 795 subjects with an unsatisfactory spirogram, or exclusion of 391 subjects without a positive control for skin tests.
To our knowledge, this is the first large-scale population study showing that peripheral blood eosinophilia is associated with increased all-cause mortality (RR = 1.43). Positive skin tests are only associated with all-cause mortality in the subgroup of subjects with FEV1 % predicted <80% (RR = 1.7). These risks are independent of gender, age, smoking habits, and FEV1 % predicted, which all contributed to all-cause mortality in this study. The above-mentioned results remain unchanged after exclusion of subjects with asthma.
One single measure of eosinophilia predicted all-cause mortality (RR = 1.43) after a maximum follow-up of almost 30 years. This increased all-cause mortality risk was comparable with that associated with smoking (RR = 1.46). We found that with a cutoff point of 275 cells per mm3, 12% of our population had eosinophilia. Relatively few subjects had extreme values for eosinophils; only one-quarter of the subjects with eosinophilia had more than 450 cells per mm3 (3% of all subjects; data not shown). With a cutoff point of 200 cells per mm3, we still found an increased risk for all-cause mortality associated with eosinophilia. Thus, a moderately elevated peripheral eosinophil count was associated with an increased all-cause mortality risk. Total leukocyte count has been associated with all-cause mortality. 18,19 It is impossible for us to assess in our data whether the effect of eosinophils on mortality is independent of total leukocyte count, as we measured eosinophils directly and have no measure of total leukocytes.
We only can speculate as to the mechanism by which moderately elevated eosinophil levels relate to total mortality. Eosinophils normally circulate less than 24 hours in the blood before they attach to vascular endothelium and migrate into tissues with an epithelial surface with the environment such as the respiratory, gastrointestinal, and lower genitourinary tracts. They carry out a wide range of functions, mainly associated with allergic, parasitic, and chronic inflammatory responses. Eosinophils can degranulate and secrete major basic protein, eosinophilic cationic protein, eosinophil-derived neurotoxin, and eosinophil peroxidase and secrete newly synthesized constituents from lipid components, oxygen radicals, cytokines, and neuropeptides in areas of acute and chronic inflammation. 20,21 These eosinophil products can damage tracheal epithelial cells, cause direct histamine release from human basophils, inhibit lymphocyte proliferation, and form a potent cytotoxic effector system in the presence of hydrogen peroxide, resulting in toxicity vs helminths and augmentation of the effects of mast cells. 20,22–28 Several investigators 22,29,30 showed that a marked accumulation of eosinophils occurs in several important disorders, such as allergic disease, parasitic infections, cancer, skin disease, gastrointestinal disease, immunological disease. In addition, Sweetnam et al31 showed that subjects with an increased eosinophil count had an increased incidence of ischemic heart disease, a major cause of death in Western countries. Because eosinophils seem to be present in a very large number of human diseases, the eosinophil may have a much more general role in inflammation. 30 In addition, our results suggest that eosinophils or their above-mentioned products may contribute to inflammatory processes that are, in the long term, related to mortality.
Another possibility is that increased numbers of eosinophils are associated with a third factor, which explains the association with all-cause mortality. In this study, the increased risk persisted after we controlled for smoking and lung function, both known to be associated with eosinophilia 9,13 and all-cause mortality. 15,16 Also, the increased all-cause mortality risk for subjects with eosinophilia did not change with adjustment for skin test positivity, which suggests that its effect was independent of allergic sensitization. Eosinophilia is associated with specific diseases such as asthma. 5 We showed that the all-cause mortality risk associated with eosinophilia, however, was not limited to subjects with asthma. Ulrik and Frederiksen 32 demonstrated that eosinophilia (>450 cells per mm3) was associated with a highly increased risk (RR = 7.4) of asthma death in asthma patients. In contrast to the study of Ulrik and Frederiksen, 32 our study was not restricted to patients with clinical asthma, and a history of asthma attacks was relatively uncommon (3%) in our general population sample.
Positive skin tests were associated with all-cause mortality only within the subgroup of subjects with FEV1 % predicted <80%. There is only indirect support in the literature that this association was not merely attributable to chance. Two population studies reported 33,34 that positive skin tests were associated with an accelerated decline in FEV1. FEV1 % predicted <80% by itself was independent of positive skin tests associated with a higher mortality risk. Therefore, the association between FEV1 % predicted <80% and all-cause mortality was only partly mediated by positive skin tests. This observed association not only reflected the presence of subjects with asthma in this population, because analyses without subjects with asthma showed comparable RRs and CIs.
The strengths of the current epidemiologic study are that (1) we adjusted for gender, age, smoking, and FEV1 % predicted, which are known predictors of mortality; (2) we had a large, young study population (5,383 subjects) with a large number of deaths (1,135); (3) we had a very long follow-up time (maximum almost 30 years); and (4) we had only 75 subjects lost to follow-up (1%).
In summary, we conclude that peripheral blood eosinophilia was independently associated with total mortality in the general population. This effect was not different in subgroups of gender, age, smoking habits, or levels of FEV1 % predicted. Eosinophilia may reflect endothelial damage that can be a cause or a consequence of disease, and it may integrate several environmental and physiologic risk factors. Nonallergic causes of eosinophilia appear to be important for all-cause mortality, because the RR remained unchanged after additional adjustment for positive skin tests and asthma. The higher mortality risk associated with positive skin tests was limited to the subgroup with values of FEV1 % predicted <80%. It seems that the association between FEV1 % predicted and all-cause mortality was partly mediated by an atopic constitution and, therefore, may suggest that lung function impairment must be avoided in subjects with positive skin tests.
We thank Joke Noordhof for her assistance with data collection and management and the departments of civil affairs of the municipalities of Vlagtwedde and Vlaardingen for their assistance with tracing the vital status of the participants.
1. Burney PGJ. Epidemiologic trends. In: Barnes PJ, Grunstein MM, Leff AR, Woolcock AJ, eds. Asthma. Philadelphia: Lippincott-Raven, 1997; 35–47.
2. Desreumaux P, Capron M. Eosinophils in allergic reactions. Curr Opin Immunol 1996; 8:790–795.
3. Barbee RA. Methodological issues in assessment of atopy. In: Weiss ST, Sparrow D, eds. Airway Responsiveness and Atopy in the Development of Chronic Lung Disease. New York: Raven Press, 1989; 121–156.
4. Mensinga TT, Schouten JP, Rijcken B, Weiss ST, Speizer FE, Van der Lende R. The relationship of eosinophilia and positive skin test reactivity to respiratory symptom prevalence in a community-based population study. J Allergy Clin Immunol 1990; 86:99–107.
5. Bousquet J, Chanez P, Lacoste JY, Barnéon G, Ghavanian N, Enander I, Venge P, Ahlstedt S, Simony-Lafontaine J, Godard P, Michel F-B. Eosinophilic inflammation in asthma. N Engl J Med 1990; 323:1033–1039.
6. Van der Lende R, Kok TJ, Peset Reig R, Quanjer PH, Schouten JP, Orie NGM. Decreases in VC and FEV1 with time: indicators for effects of smoking and air pollution. Bull Eur Physiopathol Respir 1981; 17:775–792.
7. Van der Lende R. Epidemiology of Chronic Non-Specific Lung Disease (Chronic Bronchitis). Assen, the Netherlands: Van Gorcum, 1969.
8. Veening GJJ. Waarnemingen over het aantal eosinofiele granulocyten in het bloed by astmatici en normalen. Den Haag, the Netherlands: Excelsior, 1958.
9. Mensinga TT, Schouten JP, Rijcken B, Weiss ST, Van der Lende R. Host factors and environmental determinants associated with skin test reactivity and eosinophilia in a community-based population study. Ann Epidemiol 1994; 4:382–392.
10. Van der Lende R, Orie NGM. The MRC-ECCS questionnaire on respiratory symptoms (use in epidemiology). Scand J Respir Dis 1972; 53:218–226.
11. British Medical Research Council’s Committee on Research into Chronic Bronchitis. Instructions for the Use of the Questionnaire on Respiratory Symptoms. London: Medical Research Council, 1966.
12. Groupe de travail sur la bronchite et l’emphysème de la Haute Autorité de la Communauté Européene du Charbon et de l’Acier (CECA). Questionnaire pour l’étude de la bronchite chronique et de l’emphysème pulmonaire. Luxembourg: CECA, 1967.
13. Mensinga TT, Schouten JP, Weiss ST, Van der Lende R. Relationship of skin test reactivity and eosinophilia to level of pulmonary function in a community-based population study. Am Rev Respir Dis 1992; 146:638–643.
14. Cox DR. Regression models and life tables. J R Stat Soc 1972; 34:187–220.
15. Beaty TH, Newill CA, Cohen BH, Tockman MS, Bryant SH, Spurgeon HA. Effects of pulmonary function on mortality. J Chron Dis 1985; 38:703–710.
16. Freund KM, Belanger AJ, D’Agostino RB, Kannel WB. The health risks of smoking. The Framingham Study: 34 years of follow-up. Ann Epidemiol 1993; 3:417–424.
17. Lange P, Nyboe J, Appleyard M, Jensen G, Schnohr P. Relation of ventilatory impairment and of chronic mucus hypersecretion to mortality from obstructive lung disease and from all causes. Thorax 1990; 45:579–585.
18. Weiss ST, Segal MR, Sparrow D, Wager C. Relation of FEV1 and peripheral blood leukocyte count to total mortality. The Normative Aging Study. Am J Epidemiol 1995; 142:493–498.
19. de Labry LO, Campion EW, Glynn RJ, Vokkonas PS. White blood cell count as a predictor of mortality: results over 18 years from the Normative Aging Study. J Clin Epidemiol 1990; 43:153–157.
20. Kroegel C, Virchow J Jr, Luttmann W, Walker C, Warner JA. Pulmonary immune cells in health and disease: the eosinophil leucocyte (part 1). Eur Respir J 1994; 7:519–543.
21. Spry CJF. Eosinophils: A Comprehensive Review, and Guide to the Scientific and Medical Literature. Oxford: Oxford University Press, 1988.
22. Weller PF. The immunobiology of eosinophils. N Engl J Med 1991; 324:1110–1118.
23. Gleich GJ, Frigas E, Loegering D, Wassom DL, Steinmuller D. Cytotoxic properties of the eosinophil major basic protein. J Immunol 1979; 123:2925–2927.
24. Ayars GH, Altman LC, Gleich GJ, Loegering DA, Blaker CB. Eosinophil- and eosinophil granule-mediated pneumocyte injury. J Allergy Clin Immunol 1985; 76:595–604.
25. O’Donnell M, Ackerman S, Gleich G, Thomas L. Activation of basophil and mast cell histamine release by eosinophil granule major basic protein. J Exp Med 1983; 157:1981–1991.
26. Zheutlin L, Ackerman S, Gleich G, Thomas L. Stimulation of basophil and rat mast cell histamine release by eosinophil granule-derived cationic proteins. J Immunol 1984; 133:2180–2185.
27. Peterson CG, Skoog V, Venge P. Human eosinophil cationic proteins (ECP and EPX) and their suppressive effects on lymphocyte proliferation. Immunobiology 1986; 171:1–13.
28. Henderson WR, Chi EY, Klebanoff J. Eosinophil peroxidase-induced mast cell secretion. J Exp Med 1980; 152:265–279.
29. Rothenberg ME. Eosinophilia. N Engl J Med 1998; 338:1592–1600.
30. Lombardi C, Senna GE. Association between eosinophilia and diseases in humans. Allergy 1998; 53:126.
31. Sweetnam PM, Thomas HF, Yarnell JWG, Baker IA, Elwood PC. Total and differential leukocyte counts as predictors of ischemic heart disease: the Caerphilly and Speedwell Studies. Am J Epidemiol 1997; 145:416–421.
32. Ulrik CS, Frederiksen J. Mortality and markers of risk of asthma death among 1,075 outpatients with asthma. Chest 1995; 108:10–15.
33. Gottlieb DJ, Sparrow D, O’Connor GT, Weiss ST. Skin test reactivity to common aeroallergens and decline of lung function. The Normative Aging Study. Am J Respir Crit Care Med 1996; 153:561–566.
34. Villar MTA, Dow L, Coggon D, Lampe FC, Holgate ST. The influence of increased bronchial responsiveness, atopy, and serum IgE on decline in FEV1: a longitudinal study in the elderly. Am J Respir Crit Care Med 1995; 151:656–662.
Keywords:© 2000 Lippincott Williams & Wilkins, Inc.
asthma; atopy; cohort studies; eosinophilia; hypersensitivity; mortality; population; skin tests; gender; smoking; lung function