OUP user menu

Risk and predictors of fatigue after infectious mononucleosis in a large primary-care cohort

I. Petersen, J.M. Thomas, W.T. Hamilton, P.D. White
DOI: http://dx.doi.org/10.1093/qjmed/hci149 49-55 First published online: 5 December 2005

Abstract

Background: Fatigue has been found to complicate infectious mononucleosis (IM) when patients are directly asked about it. We do not know whether such fatigue is clinically significant, nor whether IM is a specific risk for fatigue (or whether it can follow other common infections). Various risk markers for post-infectious fatigue have been identified, but findings are inconsistent.

Aim: To determine the risk of clinically reported fatigue (compared with depression) after IM (compared with both influenza and tonsillitis) in patients attending primary care, and to examine risk markers for post-IM fatigue.

Design: Comparison of matched primary-care cohorts.

Methods: We identified 1438 adult patients with a positive heterophil antibody test for IM from the UK General Practice Research Database. These patients were individually matched on age, sex and practice to two comparison groups; one with a clinical diagnosis of influenza and the other of tonsillitis.

Results: The odds ratios (ORs) (95%CI) for reported fatigue after IM vs. influenza and tonsillitis were 4.4 (2.9–6.9) and 6.6 (4.2–10.4), respectively. Risk markers for post-IM fatigue included female sex and premorbid mood disorder. By comparison, the ORs for depression after IM vs. influenza and tonsillitis were 1.6 (0.9–2.6) and 2.3 (1.4–3.9), respectively.

Discussion: IM is a specific and significant risk for clinically reported fatigue, which is both separate from, and more common than, depression. Female sex and premorbid mood disorder are risk markers for fatigue. These can be used both to target prevention strategies and to explore aetiological mechanisms.

Introduction

Fatigue may follow infections such as infectious mononucleosis (IM),1–4 although these studies all relied on asking patients directly whether they were fatigued. We do not know whether such fatigue is clinically important to the extent of it being sufficiently severe or disabling for the patient to attend their general practitioner. Other triggering infections for fatigue include viral hepatitis,5 aseptic meningitis,6 but not common upper respiratory tract infections.2,,7 Many patients attribute fatigue to ‘flu-like illnesses‘, but there have been no prospective studies of influenza itself.8 Fatigue may be caused by depression, which is the most common differential diagnosis.1 Viruses may precipitate depressive illness, but the evidence is mixed and inconclusive.2,,9 Again, we do not know whether such depression is clinically important and reported to a doctor.

Risk markers for post-infectious fatigue may include older age, female sex, atopy, lymphadenopathy, and both pre-morbid fatigue and mood disorder,3,7,10–12 but findings have been inconsistent, being based on relatively small samples, unspecified infections, or subject to bias by retrospective recall. Risk markers for chronic fatigue syndrome (CFS), which may sometimes follow certain infections,2 may also include increased consultation rates.13

Our aims were to examine: (i) whether IM is a risk for reported fatigue more than both influenza and tonsillitis; (ii) whether IM patients were at higher risk of clinically reported fatigue after infection compared to beforehand; (iii) whether clinically reported depression was an independent sequel; and (iv) to examine predictive risk markers for post-infectious fatigue in a large cohort without the bias of recall.

Methods

Patients and reported symptoms

We studied patients from the General Practice Research Database (GPRD). Data have been collected since 1987, include records from approximately 8 million patients attending general practice throughout the UK, and have been subjected to thorough validation and stringent quality checks.14 Data were recorded as Read or OXMIS codes. This has the advantage over research interviews of not being subject to recall bias.15 A library of codes for fatigue symptoms and diagnoses was assembled. These included: ‘tiredness, malaise, lethargy, debility, and fatigue’ (symptoms), and ‘post-viral debility, post-viral fatigue syndrome, post-influenza debility and syndrome post-viral’ (diagnoses). We also created lists of codes for both depression and potential risk markers (available from the authors).

Whole cohort studies

We identified 1438 patients aged 16 or above, presenting to their GP between 1989 and 2000, with a positive heterophil antibody test. This test is specific for Epstein-Barr virus infection.16 Fatigue may precede a positive test result.2 Therefore, infectious onset was defined as 1 month before the date of the positive test or diagnosis. For the study of risk markers of post-IM fatigue, we studied those 1318 patients who did not have a record of premorbid fatigue in the year before onset. Patients with premorbid fatigue were only included in the analyses when testing the association between pre- and post-morbid fatigue and when comparing the risk of fatigue before and after IM. Only patients with at least one year of premorbid and post-morbid data were studied.

Matched cohort studies

For the matched cohort studies, IM patients were individually matched on sex, age (±2 years) and general practice to two comparative cohorts. For the studies of fatigue, it was possible to obtain a match for 1177 tonsillitis and 888 influenza patients with IM patients without fatigue prior to onset. For the studies of depression, a match was obtained for 1275 tonsillitis patients and 916 influenza patients with IM patients without depression in the year prior to onset. More than 80% of the influenza diagnoses were made between October and March, but there was no seasonality for either tonsillitis or IM.

Risk markers for fatigue after IM on whole cohort of IM patients

Data were extracted for all IM cases on: age, sex, lymphadenopathy within 2 months of the positive test, and the number of consultations and sickness certificates in the year before onset. We also extracted any premorbid record of anxiety or depressive (mood) disorder, fatigue and atopy (eczema, asthma or hay fever). For premorbid mood disorders, fatigue and atopy, all data available up to the date of onset were used. The study protocol was ethically approved by the Scientific and Ethical Advisory Group of the GPRD.

Statistical methods

We used conditional logistic regressions in four matched cohort studies to examine the risk of fatigue in the year following IM vs. tonsillitis and IM vs. influenza, and depression in the year following IM vs. tonsillitis and IM vs. influenza. Those cases for whom no match was possible were omitted from this part of the analysis. In our analyses of risk of fatigue and depression before and after IM, we took into account that measurements were obtained repeatedly within individuals.

In the study of risk markers for fatigue, which included the whole cohort of IM patients, we included sex, age, lymphadenopathy, sickness certificates, atopy, premorbid mood disorder, and consultation rates in the univariate analyses. All variables were binary, except age (16–19; 20–29; 30–39; 40–73 years), which we divided into epochs because of the low median age. The number of consultations ranged from 0–45, with a median of 2, and was therefore categorized around the median (0–2, 3+).

We first examined risk markers in univariate analyses. We then entered sex and age into a multivariable logistic regression, together with all other variables that had a p value below 0.05 in the univariate analyses. In order to measure the possible confounding effects of post-infectious comorbid depression on the association between premorbid depression and post-infectious fatigue, we re-analysed the data having excluded those with a comorbid post-infectious depression.

The association between premorbid and post-infectious fatigue was separately examined in the full cohort with ‘prevalent’ rather than ‘incident’ fatigue. First, we investigated this in a univariate analysis. Thereafter, we entered sex, age and premorbid mood disorder into a multivariate model. Analyses were performed using STATA 8.2.17 Results were considered significant at the 5% level.

Results

In the IM cohort, the age ranged from 16 to 72 years, with a median age of 19 years. The sex ratio was close to 1:1. The patients in the matched cohort studies were slightly older (Tables 1 and 2).

View this table:
Table 1

Characteristics of IM cohort alone and matched cohorts for fatigue

IM cohort aloneMatched cohort studies for fatigue
IM vs. influenzaIM vs. tonsillitis
IMInfluenzaIMTonsillitis
n131888888811771177
Fatigue after onset162 (12.3%)112 (12.6%)26 (2.9%)146 (12.4%)22 (1.9%)
Age (years)
16–19717 (54%)426 (48%)381 (43%)637 (54%)635 (54%)
20–29412 (31%)304 (34%)353 (40%)377 (32%)378 (32%)
30–39112 (9%)100 (11%)92 (10%)99 (9%)101 (9%)
40–7277 (6%)58 (7%)62 (7%)64 (5%)63 (5%)
Sex
Women648 (49%)409 (46%)409 (46%)567 (48%)567 (48%)
Men670 (51%)479 (54%)479 (54%)610 (52%)610 (52%)
  • Matched cohorts were individually matched by sex, age and general practice. Data are numbers (%).

View this table:
Table 2

Characteristics of IM cohort alone and matched cohorts for depression

IM cohort aloneMatched cohort studies for depression
IM vs. influenzaIM vs. tonsillitis
IMInfluenzaIMTonsillitis
n137491691612451245
Depression after onset54 (3.9%)38 (4.1%)25 (2.7%)47 (3.8%)21 (1.7%)
Age (years)
16–19758 (55%)461 (50%)418 (46%)686 (55%)684 (55%)
20–29425 (31%)305 (33%)351 (38%)391 (31%)392 (31%)
30–39115 (8%)95 (11%)87 (9%)105 (9%)106 (9%)
40–7276 (6%)56 (6%)60 (7%)63 (5%)63 (5%)
Sex
Women684 (50%)429 (47%)429 (47%)608 (49%)608 (49%)
Men690 (50%)487 (53%)487 (53%)637 (51%)637 (51%)
  • Matched cohorts were individually matched by sex, age and general practice. Data are numbers (%).

Infections and fatigue

In the matched cohorts, between 12.3% and 12.6% of the patients with IM had a record of fatigue in the year after onset, depending on the particular sample, compared with 2.9% after influenza and 1.9% following tonsillitis (Table 1). The odds ratio (95%CI) for fatigue after IM was 4.4 (2.9–6.9) (p<0.001), when compared with influenza, and 6.6 (4.2–10.4) (p<0.001) when compared with tonsillitis. Most of the fatigue following IM was recorded close to onset (Figure 1); the median time between onset and report was 55 (IQR 29, 167) days. No temporal association was found between fatigue and either influenza or tonsillitis (Figure 1). The odds ratio (95%CI) for fatigue following IM was 4.0 (2.9–5.5) (p<0.001), compared with before IM.

Figure 1.

Timing of patients' first record of fatigue. (a) IM (black bars) vs. tonsillitis (white bars). (b) IM (black bars) vs. influenza (white bars).

Of the 1318 patients in the full unmatched IM cohort alone, who reported no premorbid fatigue, 162 (12%) reported ‘incident’ fatigue after onset. Of these patients, 16 (10%) had additional records of fatigue (>30 days after the first code) in the year following IM. Twenty-two patients (2% of the IM cohort) had been given a fatigue diagnosis by their GP.

Infections and depression

Between 3.8% and 4.1% of patients with IM had a record of depression in the year following onset, depending on the particular sample, compared with 2.7% after influenza and 1.9% following tonsillitis (Table 2). The odds ratio (95%CI) of depression after IM compared to influenza was 1.6 (0.9–2.6) (p = 0.09), and 2.3 (1.4–3.9) (p = 0.002) when compared to tonsillitis. In contrast to fatigue, there was no temporal association between depression and the onset of IM (Figure 2). The odds ratio (95%CI) for depression after IM was 2.0 (1.4–2.9) (p<0.001), compared with beforehand.

Figure 2.

Timing of patients’ first record of depression. (a) IM (black bars) vs. tonsillitis (white bars). (b) IM (black bars) vs. influenza (white bars).

Risk markers for fatigue after IM

In the final multivariate model of incident fatigue, fatigue was significantly more common among women than men. Patients with a premorbid mood disorder had a significantly increased risk of fatigue following IM, which remained the case after excluding the 49 patients who also had a comorbid mood disorder following IM (OR 2.3, 95%CI 1.1–4.6). The significant univariate association with premorbid GP consultation frequency became non-significant in the multivariate model (OR 1.4, 95%CI 1.0–2.0, p = 0.07). No significant association was found between post-IM fatigue and records of: atopy, lymphadenopathy, or being in receipt of a sickness certificate in the year before IM (Table 3).

View this table:
Table 3

Univariate and multivariate regression analyses of risk markers for fatigue after IM

Fatigue after IMTotalUnivariate OR (95%CI)pMultivariate OR (95%CI)p
Sex
Women100 (15%)6481.8 (1.3–2.5)0.0011.7 (1.2–2.3)0.006
Men62 (9%)6701.01.0
Age0.110.25
16–1982 (11%)7171.01.0
20–2947 (11%)4121.0 (0.7–1.5)1.1 (0.7–1.6)
30–3917 (15%)1121.4 (0.8–2.4)1.3 (0.7–2.3)
40–7216 (21%)772.0 (1.1–3.7)1.9 (1.0–3.5)
Lymphadenopathy
Present12 (8%)1510.6 (0.3–1.1)0.09
Absent150 (13%)11671.0
Sickness certificate in the year before IM
Present7 (19%)371.7 (0.7–3.9)0.22
Absent155 (12%)12811.0
GP consultations in the year before IM
0–269 (10%)7101.01.00
3+93 (15%)6081.7 (1.2–2.3)0.0021.4 (1.0–2.0)0.07
Premorbid records: mood disorder
Present15 (25%)592.6 (1.4–4.8)0.0021.9 (1.0–3.7)0.05
Absent147 (12%)12591.01.0
Premorbid records: atopy
Present46 (11%)4130.9 (0.6–1.2)0.39
Absent116 (13%)9051.0
  • Statistically significant results are in bold. Variables included in the final multivariable analysis included: sex, age, GP consultations, and mood disorder.

There were 120 patients who had had fatigue recorded in the year before IM. Twenty-six of these (22%) also had fatigue recorded after IM compared to 162 (12%) of those without premorbid fatigue (OR 2.0, 95%CI 1.2–3.1, p = 0.004). However, the association between premorbid and postmorbid fatigue was of only borderline significance after controlling for age, sex and premorbid mood disorder (OR 1.6, 95%CI 1.0–2.6, p = 0.06).

Discussion

IM was a specific and strong risk for fatigue, compared with both influenza and tonsillitis. Fatigue was four times more likely after IM than before IM. Risk markers for fatigue after IM included both female sex and premorbid mood disorder.

It is possible that doctors—believing that IM to be a risk factor for fatigue—were biased towards recording fatigue following IM, rather than after other infections. But conversely, some doctors may not have recorded fatigue after IM, seeing it as an expected sequel of the infection. We examined the possible effect of any doctors’ bias in recording fatigue more frequently after IM by repeating conditional logistic regressions to find out what percentage of patients with fatigue after IM would need to be changed to not being fatigued in order for the odds ratios to be only just significantly different. These analyses showed that 62% of doctors would have had to record fatigue on their basis of a biased belief before the odds ratio for IM vs. influenza became non-significant. We think this is unlikely. Only 18% of post-IM fatigue cases were attributed to the infection by a doctor using codes such as post-viral fatigue syndrome.

Infections as risk factors for fatigue

An increased risk of both acute and chronic fatigue after IM was previously found in two smaller prospective cohort studies.2,,3 The results of this study confirm IM as a strong risk factor for fatigue in a much larger cohort, even when compared with another specific viral infection, which supports its specificity for inducing fatigue. Unlike previous work, our study did not ask participants directly about fatigue. This strengthens the clinical significance of this finding: fatigue that prompts a patient to visit their GP, and their GP to record it is likely to be a significant clinical problem, even though we were unable to measure its severity directly. The close temporal relationship between the onset of IM and consequent fatigue also supports an aetiological relationship.

Influenza or a ‘flu-like’ illness has been anecdotally reported to cause fatigue and even CFS, but these attributions are retrospective, making chance associations likely.7,8,,18 This study found little evidence that influenza is a risk for fatigue. In the absence of laboratory tests, we cannot be sure that all the patients actually suffered from influenza. In the light of the clear seasonal (winter) variation, which was absent in the other two infectious groups, we can be sure that we were studying clinically diagnosed influenza, which is an accurate diagnosis in 66–85 per cent of cases.19,,20 However, other non-influenza viruses may have been included in our cohort, so it is still possible that influenza corroborated by serology is a risk for reported fatigue.

Only one prospective primary-care study found a slightly increased risk of fatigue 6 months after any ‘viral illness’ when compared with non-infectious attenders, with a low risk of 1.4.11 The authors concluded, however, that the fatigue was related to symptom attributions and doctor behaviour, rather than to features of the viral illness. Two other prospective primary-care studies provided clear evidence against an association between common upper respiratory tract infections and fatigue.2,,7 Our study was not designed to examine the general risk of fatigue after influenza. However, we found little difference in the incidence of fatigue in the year after infection with influenza (2.9%) and tonsillitis (1.9%). Since the median age was 20 years, we cannot exclude clinically diagnosed influenza as a risk factor in the general population as a whole.

Relying on a positive heterophil antibody test to define our IM cohort will exclude patients with primary Epstein-Barr virus infections who had a negative heterophil antibody test.16 Although this will be a small minority, these patients may have a less severe illness with less sequelae.10 This may have exaggerated our contrasts to the other infections.

Risk markers for fatigue after IM

Female sex was a risk factor at some point in the year after IM, both in our study and in three previous cohort studies.3,4,,10 Both biological and social explanations might explain this, but the relative contributions of these are unclear.1 Age was not a significant variable in the final multivariate model, though the trend for increasing fatigue with increasing age replicates previous work.3 Other studies have found premorbid fatigue to be a risk factor for post-infectious fatigue.7,,11 However, we only found the association to be of borderline significance after controlling for age, sex and premorbid mood disorder. There has been some debate as to whether previous mood disorder is also a predisposition for post-IM fatigue.7,,10 Our study suggests that there is indeed an association between premorbid mood disorder and post-IM fatigue. This remains the case even after excluding patients with co-morbid mood disorders. Despite lymphadenopathy being a common feature of IM,3,10,,21 only 12% of our patients had this recorded close to onset of IM. We found no evidence for an association between atopy and fatigue after IM, reflecting the inconsistency of previous evidence.10,12,,22

Other studies have found a lack of physical fitness, inactivity and illness perceptions were significant risk factors for post-IM fatigue;3,4,10,,23 factors we were unable to measure. Two controlled trials have suggested that early physical activation may prevent post-IM chronic fatigue.24,,25 Our study results allow such intervention to be targeted to those most at risk.

Post-IM fatigue and chronic fatigue syndrome

While the results of this study confirm an immediate association between IM and fatigue, only 1:10 patients had an additional record of fatigue, implying that the majority recovered within a year. This is further supported by the fact that only 2% received a fatigue diagnosis within a year after IM. This is consistent with a prospective study of IM that showed that the median duration of an acute fatigue syndrome was 8 weeks.2 Few patients therefore will have fulfilled the criteria for chronic fatigue syndrome or myalgic encephalomyelitis (CFS/ME), one of which is a duration of 6 months.1,,22 In fact, no diagnoses of CFS or ME were recorded in this study. Acute fatigue after IM and CFS/ME may share some common risk markers, such as sex and premorbid mood disorder, but CFS/ME is more heterogeneous than post-IM fatigue, and probably does not share the same aetiology.

Viruses and depression

Prospective studies of viral infections suggest an increased but transient risk of depression, in particular after IM.2,,9 We confirmed a similarly increased albeit small risk of depression following IM, but with little risk after influenza in this matched cohort. The lack of a close temporal association between the onset of IM and depression diminishes the likelihood of a direct aetiological relationship.

Acknowledgments

We are grateful to Professor Matthew Hotopf and Arlene Gallagher for advice and stimulating discussions on the subject. Funding was provided by the Linbury Trust and the Celebrity Guild of Great Britain. WTH receives funding from RCGP/BUPA and NHS Fellowships.

Footnotes

  • *Current address: Primary Care and Population Sciences, University College London, London, UK.

References

View Abstract