A Cohort Study of Stress and the Common Cold

Abstract

The common cold is one of the major causes of work absenteeism. Former studies, based on artificial inoculation of rhinovirus, implicated psychological stress in the occurrence of this syndrome, either by increasing susceptibility to the virus or by causing the subject to overrate the perception of the symptoms. Nevertheless, few studies on the effect of stress on the naturally acquired common cold have been conducted. We carried out a 1-year prospective cohort study among the faculty and staff of a Spanish university (N = 1,149). By means of standardized questionnaires, validated in a random sample of the population, we assessed the relation between the occurrence of common cold episodes and exposure to four dimensions of stress: stressful life events, negative affect, positive affect, and perceived stress. All four aspects of stress were related to the occurrence of the common cold. Subjects with a high (fourth quartile) index of negative affect showed an incidence rate ratio of 3.7 (95% confidence interval = 2.2–6.2). The incidence rate ratios for the fourth quartile were 2.5 (95% confidence interval = 1.5–4.1) and 1.9 (95% confidence interval = 1.1–3.2) for perceived stress and stressful events, respectively. A high index of positive affect was associated with an incidence rate ratio of 0.6 (95% confidence interval = 0.3–1.0). These findings suggest that psychological stress is a risk factor for the common cold.

The common cold, although not a severe disease, is one of the main causes of absenteeism at work. About 30 million days of work are lost every year in the United States as a result of this syndrome, which also causes 250 million days of restricted activity. ¹

There are several reasons to consider psychological stress as a possible risk factor for the common cold. First, stress can alter immune function, and thus cause the common cold, by producing changes in the concentration of leucocyte cytokines, molecules that mediate the host response to infection and are responsible for the cellular release of compounds related to inflammation. ² Second, stress may influence health behavior. Persons under stress may take on negative habits that place them under increased risk of disease. These behavior changes include smoking, excessive drinking, and poor diet. ³ Stress may also increase the perception of disease by making the subject “overdiagnose” his or her symptoms and label them as a common cold episode. ⁴

Although there is a common intuitive understanding of the term “stress,” disagreement exists about its meaning. Numerous definitions have, so far, been advanced. There is, however, a common core concept of stress that refers to a process in which external demands (also called stressors) exceed the adaptation capacity of the organism and cause undesirable psychological and biological changes. ⁵ Three steps in the stress process have been identified. Each may be a risk factor for the common cold: the existence of undesirable events or stressors (measured by checklists of events); the subjective perception of these events by the subject as threatening or not; a perception that depends on the individual’s psychological structure (measured by the affect scales); and eventually, the biologic reaction of the organism, which may include hemodynamic and endocrinologic effects such as changes in the heart rate and blood pressure and hormone secretion.

A few studies, published in the psychology and medical literature, found a relation between psychological stress and susceptibility to the common cold. ^6–9 These studies, carried out by the same group, were based on artificial (experimental) inoculation of volunteers with a limited set of viruses, essentially rhinovirus but also influenza A virus. ⁹ Although these studies have the advantage of standardizing the conditions of exposure to virus, the information they provide is likely to be incomplete, because the cause of one-third to one-fourth of the naturally occurring common cold episodes is unknown. ¹⁰ Also, no information is available on the viral dose that is necessary to produce a common cold episode in natural conditions.

Little is known about the role of stress in naturally occurring common cold episodes. The aim of this follow-up study was to elucidate the role of psychological factors in the occurrence of naturally acquired common cold among an easily accessible population. In this investigation, we focus on the development of clinical illness, based on symptoms and signs and regardless of the immune response.

Subjects and Methods

Study Population and Data Collection

For 1 year we traced the faculty and administrative staff members of a large Spanish university. The number of subjects who agreed to participate was 1,149. At the beginning of the follow-up, each participant completed an anonymous mail questionnaire on stress, common cold symptoms, and other lifestyle variables that could be potential confounders of the relation between stress and common cold. We investigated four different dimensions of stress: stressful events, ¹¹ negative affect, ¹² positive affect, ¹² and perceived stress. ¹³ Subjects experiencing a common cold episode were at that time discontinued from follow-up. Asthmatics and persons with chronic obstructive pulmonary disease were excluded. We sent short questionnaires every 10 weeks to identify new cases of common cold and to update the stress information. To increase the response rate, both the initial and the follow-up questionnaires were anonymous. To link an initial questionnaire with its corresponding follow-up, we asked the participants to write down, on every questionnaire they received, an eight-digit code number formed by the date of birth, gender, and job category (faculty or staff). Four instances of exact duplication of the code number were excluded from follow-up. Follow-up was complete for 77.9% of the cohort subjects. The initial questionnaire needed 20 minutes on average to be completed and the follow-up questionnaires about 3 minutes.

Disease Assessment

Only the first common cold episode that occurred during the follow-up was taken into account. We asked the participants to record the presence of eight common cold symptoms and to rate their intensity from 0 (none) to 3 (very intense). These symptoms were: runny nose, sneezing, nasal congestion, headache, chills, sore throat, cough, and malaise. We asked the participants to record and rate the symptoms on a daily basis on a calendar that we provided, beginning immediately after the onset of the episode until its complete remission. Subjects were considered as having developed a common cold episode if they reported rhinorrhea for a minimum of 3 consecutive days, had a subjective sensation of a common cold, and had a score of symptoms of 12 or more out of 24 on the peak day of the episode. We used the receiver operating characteristic curve to determine the cutoff point of 12, chosen to maximize the sensitivity and specificity in the validation study explained below. In addition, we performed a sensitivity analysis by comparing the results obtained under varying cutoff points. The results were not altered substantially using different assumptions.

Exposure Assessment

We used a checklist of 16 items taken from the Schedule of Recent Events scale to measure the frequency of stressful life events, weighted by their importance, that might have occurred during the last year. ¹¹ To measure negative affect, we used a scale developed by Watson et al¹² that contained a list of five adjectives related to negative emotions. The subjects were asked to report how they globally felt during the last 10 weeks by ranking every adjective on a five-grade scale from “never” to “very often.” The scale that measured the positive affect had a similar aspect and contained 10 adjectives that indicated positive feelings. ¹² The positive affect and negative affect scales are independent and do not measure the same stress dimension. ¹⁴ Globally, positive affect is usually associated with extraversion, high levels of social contacts, and participation in new activities and is a measure of energy and concentration, whereas negative affect is a measure of distress and neuroticism. ¹²

The Perceived Stress Scale ¹³ measures the degree to which life events are perceived as stressful by the subject. We used the reduced perceived stress scale, a set of four questions that measure this perception.

Data Analysis

Each participant accumulated person-time starting with the return of the initial questionnaire and ending with the onset of a common cold episode, the termination of the study, or the loss to follow-up, whichever came first. To take into account the fact that the common cold episodes are not completely independent events, we used negative binomial regression instead of Poisson regression to obtain adjusted incidence rate ratios (IRRs) and their corresponding 95% confidence intervals. ¹⁵ The use of the negative binomial regression instead of Poisson regression yields a larger variance of the maximum likelihood estimates of the IRR. Stress scores were categorized into quartiles of distribution, with the first quartile as a reference category. As potential confounders or effect modifiers of the relation stress-common cold, we measured smoking status using the standardized World Health Organization questionnaire ¹⁶ and alcohol, vitamin C, and zinc intake using an adapted version of a Spanish food frequency questionnaire. ¹⁷ We also considered three variables related to contact with children: total number of children, number of children less than 2 years of age, and number of children who go to kindergarten. None of these variables was a confounder, and hence, they were not introduced into the model. Covariates were introduced in the final model if their inclusion changed the estimate of the crude IRR by more than 10%.

Validity and Reproducibility Study

Exposure and disease assessments were largely based on scales published in English and validated in the U.S. population. We used translation and back-translation to adapt the existing questionnaires to a Spanish environment.

To assess the intrinsic value of our questionnaires, we carried out a substudy to measure their validity and reproducibility (stability over time of the exposure measurement) in a random sample of 71 subjects of the total population.

Validation of the Outcome

Self-diagnosis of common cold by the subject is reliable owing to the fact that the manifestations are typical. ¹⁰ Nevertheless, for a more accurate measurement of the validity of our case assessment, we validated the outcome using an adapted version of the complete diagnosis method of Beare and Reed as the gold standard. ¹⁸ This method uses a 20-item checklist of common cold symptoms and physical signs. Each sign or symptom is rated from 0 to 3. In our version, we emphasized the importance of nasal symptoms and signs over other items that exist in the original checklist, which focus on the lower respiratory tract and on general manifestations such as malaise and fever. For this purpose, we weighted the total score of signs and symptoms in the following way: 60% for nasal manifestations, 25% for lower-airway manifestations, and 15% for systemic characteristics. This weight is based on the relative frequency of symptoms during common cold episodes, as described in large case series by Tyrell et al.¹⁹ In this validation study, a subject was considered as having a common cold episode if he or she scored a minimum of 150 points out of a total of 300 points.

Validation of the Psychological Stress Scales

Because any physiological or biological marker of stress (hormones, heart rate, or blood pressure) measures a state of acute stress and may be completely unrelated to our measures of chronic stress, we validated the stress measures of our questionnaire against widely used independent scales that measured each stress dimension in more depth.

To validate our measure of stressful events, we used the Life Experiences Survey, which contains 60 items. ²⁰ We validated our assessment of negative affect using the corresponding ten items of the complete Positive and Negative Affect Schedule scale. ¹² Positive affect was validated using the 20-item affect balance scale. ²¹ The reference scale used in the validation of perceived stress was the complete perceived stress scale (14 items). ¹³ In addition, to assess reproducibility of the stress measures, on two occasions 10 weeks apart, we asked the 71 participants of the validation study to respond to the same questionnaire that we used in the main study.

Results

We detected 365 cases of common cold among 1,149 subjects with a total of 28,199 person-weeks of follow-up between October 1, 1998 and September 30, 1999. This number of cases represents an overall incidence rate of 0.67 per year. Participants ranged in age from 23 to 68 years, 46% were females, and 67% were faculty members. Spearman correlation coefficients between the scores for the four dimensions of stress are shown in Table 1. Except for the moderate correlation between negative affect and perceived stress (r = 0.45), the remainder of the correlation coefficients are low and suggest that the four scales of stress used in this study are independent.

Table 1: Spearman’s Correlation Coefficients between Four Stress Scales Measured on 1,149 Subjects

Table 2 shows the IRRs, adjusted by age (entered as a continuous variable), gender, and professional activity (faculty or staff) for each quartile of distribution of the scores. The stressful events score was progressively related to increased common cold incidence rate. The two negative measures of psychological structure, negative affect and perceived stress, showed a larger effect than the stressful events score. Subjects in the highest quartile of negative affect score are nearly four times more likely to develop a common cold episode than those in the first quartile of the distribution. Those with a high index of perceived stress show a relative risk of 2.8. Positive affect is inversely associated with common cold incidence. The effect is less substantial than that of the three other stress scales but nevertheless shows a steady trend.

Table 2: Incidence Rate Ratios (IRR) of Clinical Common Cold According to Four Stress Dimensions, Adjusted by Gender, Age, and Professional Category

Further analyses adjusting for smoking status, alcohol, vitamin C, and zinc intake did not substantially alter the relative risk estimates of any stress scale used in this study. To assess possible confounding due to the similarity of symptoms of the common cold and that of other respiratory diseases, we adjusted the model by two additional variables: history of allergic rhinitis and history of any respiratory disease other than allergic rhinitis. These adjustments introduced little modification of the stress relative risk estimates, although both history variables are independent risk factors of the common cold [IRR = 2.7 (95% confidence interval = 2.3–3.2) for history of rhinitis and IRR = 3.3 (95% confidence interval 2.7–4.0) for other respiratory disease].

A total of 22.1% of the cohort dropped out before the end of the study. To assess whether the probability of loss to follow-up is related to exposure, we compared the distribution of each exposure variable (four stress variables, sex, age, and faculty/staff status) for subjects with incomplete follow-up with those who had complete follow-up. The distribution of these variables was very similar in the two groups. We also recalculated the relative risks in two extreme situations. In the first one, we assumed that all subjects lost to follow-up developed the disease subsequently; in the second one we assumed that none of the subjects developed it. For every stress variable considered, the difference between the point estimates of the relative risks of each situation was not large enough to alter the previous results.

The results of the validation study (Table 3) indicate that there is some measurement error in the assessment of stressful events. This error, however, is likely to occur independently of disease status, as information on exposure is collected before disease occurrence. Because the exposure is polytomous, the relative risk estimates may be biased either toward or away from the null. To specify the direction and magnitude of the bias due to misclassification of exposure, we reanalyzed the data entering the variable “stressful events” as dichotomous (presence/absence of any event). The IRR was 1.8 with 95% confidence interval of 1.2–2.7. With the gold standard entered as a dichotomous variable, the sensitivity and specificity of this measure were 0.85 and 0.65, respectively. These results suggest that the true relative risk is greater than what we observed. For the remainder of the stress scales the measurement error is negligible.

Table 3: Intraclass Correlation Coefficients for Validity and Reproducibility of Four Stress Scales, Measured on 71 Subjects

The sensitivity was 0.94, and specificity was 0.84, for the outcome measure. The results of the reproducibility study (Table 3) show that the negative and positive affect scales are reasonably stable over time (intraclass r = 0.60 for negative affect and 0.76 for positive affect). These figures are higher than those found by Watson et al¹² when they originally developed the scale (0.58 for negative affect and 0.48 for positive affect). Reproducibility of the perceived stress and stressful events scales is high (intraclass r = 0.80 and 0.73, respectively).

Discussion

Our findings indicate that high levels of psychological stress are moderately associated with common colds. This effect is not easily ascribed to confounding, misclassification, or bias from loss to follow-up. In this study, we focused on clinical colds. Our diagnosis was based on symptoms and validated by additional clinical signs. Our case definition deliberately excluded criteria of seropositivity. We believe that the importance of the common cold, essentially resulting from the disability caused by the syndrome, is due to the symptoms and the subjective perception of illness. Detection of infection is a condition that has little effect on the progression and consequences of the common cold, because, generally, few or no sequelae are to be seen in this disease. The economic costs of the disease are mainly due to absenteeism and to the drugs used to alleviate common cold symptoms and are independent of the individual status as seropositive or not. In former studies, common cold episodes were defined by the presence of symptoms and seropositivity, generally among artificially infected subjects. ²² Usually, only one-third of subjects exposed to viruses become infected and only one-third of infected individuals develop cold symptoms. ^23,24 Generally, because the symptoms are typical, self-diagnosis of common cold by the subject is usually correct. ¹⁰

A number of epidemiologic studies have found a positive relation between different measures of stress and common cold or, more generally, acute upper respiratory infection episodes. Strong emphasis has been placed on demonstrating that the increase in the incidence rates was due to an increase in the infection rate and not to an increase of the symptoms perception. ²³ Graham et al⁴ found in a small cohort study that distressed individuals have more acute respiratory infection episodes than those with low stress (mean of 2.71 vs 1.56). A large number of studies, based on artificial inoculation of virus to volunteers, were carried out by Cohen et al.^6–9,23 They provide evidence for an association between common cold episodes among infected subjects and long-term stressors, with a relative risk that varies from 1.5 to 3.5 according to the duration of the stressors. ⁸ Despite the differences in the design and analysis, our results concerning stressors are close to those of these authors.

Our results concerning perceived stress are also in the same range as those of Cohen et al, who provide evidence for a relation between perceived stress and infection (relative risk = 5.8) on the one hand, and perceived stress and clinical colds among infected subjects on the other (relative risk = 2.2). ²³ As for negative affect, Cohen et al⁶ distinguish two types of negative affect components: trait negative affect, a stable individual difference in affective level, and state negative affect, a transient fluctuation in mood. These authors found that trait negative affect is related to an increased rate of infection and symptoms, whereas state negative affect is related to an overreporting of symptoms only. We did not find any mention in the literature of the effect of positive affect on common cold.

In summary, we found a moderate but consistent association between psychological stress and risk of clinical common cold. The association persisted after control for several established risk factors. Our findings are strengthened by the fact that they are in general agreement with previous work and by the biological plausibility of the relation between psychological stress and the common cold.

We thank Miguel A. Hernán for his helpful comments.