The factors determining disease manifestations and severity in MP infection are only partly understood. Comparing patients with mild and more severe MP respiratory infection, we found that patients in need of hospitalization had higher MP load in throat secretions than outpatients. This difference remained statistically significant after adjustment for age and interval between disease onset and sampling. Both MP genotype 1 and 2 were detected in similar proportions during the different outbreaks. Genotype identity was neither associated with disease severity nor bacterial load.
Higher MP load in hospitalized patients than in outpatients has previously been reported by Dorigmo-Zetsma et al  in ten patients; however, no statistical adjustment was done with regard to age and disease duration at sampling which is important as the pathogen load continuously declines with time after onset of disease .
Both age and time between disease onset and sampling were adjusted for in our calculations confirming the association between bacterial load and disease severity. This suggests that the pathogenesis of MP is mainly due to direct bacterial factors, a hypothesis that is supported by the finding of a higher risk of severe MP disease in patients for whom adequate antibiotic therapy had been delayed .
However, the existence of such an association does not exclude that immune-mediated mechanisms might also be involved in pathogenesis. In patients with high MP loads, elevated amounts of antigen could trigger a more prominent inflammatory response. This may explain why patients with severe MP pneumonia appear to benefit from corticosteroids given in addition to appropriate antibiotic treatment [14, 15]. Furthermore, improved resolution of pulmonary lesions in MP-infected mice having received a combination of macrolide antibiotics and corticosteroids compared to those who only received antibiotics have been found, lending further support to this association .
Different scoring methods have been used to determine the severity of pneumonia. We used hospital admission as a marker of more severe illness. Several factors apart from the severity of the acute infection may influence the need of hospitalisation, such as age, social situation and underlying diseases. In the statistical analysis, however, we have controlled for the influence of age. None of the included patients had multiple co-existing diseases, and in fact the majority were previously healthy and of young middle age. Only one hospital was involved in the study of clinical severity, making a bias in admission policy unlikely.
CRP levels were significantly higher in admitted patients (188 mg/L vs. 20 mg/L), further supporting our grouping of patients since CRP levels have been shown to predict severity of pneumonia [17–19].
Among our hospitalized MP patients, all 23 who underwent x-ray had pulmonary infiltrates consistent with pneumonia. The true proportion of outpatients with pulmonary infiltrates is unknown since only two of them had symptoms justifying x-ray.
It has been suggested that MP may be a commensal of the respiratory tract mucosa; thus, the identification of MP among outpatients with URTI could reflect colonization rather than invasive infection . However, in our previous study using the same PCR method, we found a low rate of MP carriage (0.4%) among 237 children healthy enough to be in school during a season with high rates of MP cases . The single child diagnosed with MP in that population study was found to have typical symptoms with persistent cough and fever. Our findings in school children are in agreement with those reported by Kumar et al who did not detect MP by multiplex PCR in any nasopharyngeal swab from 129 asymptomatic children and adults . Taken together, this indicates that detection of MP DNA in a patient with URTI is likely to signify symptomatic infection.
Several factors could contribute to the differences in MP load observed between patients. Immunosuppression and smoking [22, 23] have been associated with higher MP load, but both these factors were rare among our patients. Partial immunity from prior MP infections might also explain the different bacterial loads . In an outbreak in Israel, smoking and low pre-existing MP antibody levels were associated with an increased risk of symptomatic MP infection , suggesting that pre-existing antibody levels may affect the disease course.
In contrast to our observed association between bacterial load and clinical severity, no such association was found between genotype and clinical severity. Both MP1 and MP2 had similar distribution with regard to clinical severity.
In Japan, Kenri et al  found a gradual type shift of the most prevalent MP subtype during the period 1995-2005 from MP2 to MP1 in 127 clinical isolates obtained in three separate areas. Similar shifts of predominant type over time have been reported from Germany; France and Denmark . In our study, covering a much shorter time span both genotypes were detected during each season and in two geographical locations. Although immunological pressure may favour shifts of MP genotype, a co-circulation of MP genotypes appears to be common.
With the very restricted phylogenetic differences found within each subtype, the chosen gene product was too conserved to allow for analysis of transmission chains. Genotype and exact sequence conservation was however observed within each investigated family. To our knowledge genome conformity of MP has not been studied before.
Our study has some limitations. Most patients were recruited from a department of infectious diseases at a university hospital, which may have led to a disproportionate number of cases with more severe MP infection. Most patients in our study were young middle-aged adults. Therefore, our findings may not be applicable to children, who often constitute the majority of MP cases during outbreaks. Few participants in this study had severe underlying diseases or were smokers. This could reflect a selection bias in sampling. However, in the prevalence study from which the participants in this study were recruited , MP sampling was performed on a broad range of patients with pulmonary infections and underlying conditions, suggesting that MP infection is not a common pathogen in immunosuppressed subjects. All of the hospitalized patients and 40% of the outpatients had received betalactam antibiotic therapy prior to the MP diagnosis, but since MP lacks a cell wall, on which betalactam antibiotics exercise their effect, it is improbable that this treatment has influenced our results.
Throat swabs were used to obtain samples, and the same technique was used for hospitalized subjects, outpatients and family contacts. The amount of throat secretion absorbed by the swab before dilution in 1 mL of buffered saline could not be determined, making standardization of samples difficult. Housekeeping genes such as beta-globin maybe useful to standardize results from throat swabs. This is important when intracellular agents such as viruses are looked for but may be less useful in the case of MP which is a superficial infection of the mucous membranes.
Since sampling of patients and family members was done in few sites by a limited number of trained staff, sampling problems are unlikely to explain the observed 10-fold differences in bacterial load.