This controlled intervention trial demonstrated a significant reduction in infection episodes, especially in respiratory infections, in the trial arm carrying out enhanced conventional hand washing with soap and water, combined with instructions on how to reduce transmission while coughing or sneezing. In contrast, hand rubbing with alcohol-based disinfectant, together with guidance on coughing or sneezing, did not reduce infection episodes compared with the control group. The influenza pandemic 2009 triggered an intense national hand washing campaign, during and after which no differences were seen between the two interventions and the control arm. This is partially explained by the observation that the occurrence of infections decreased significantly in the control arm.
Cluster effect on the distribution of individual end points is often taken in account in statistical analysis of the results of intervention trials. We compared the cumulative results of the two intervention arms with the control arm directly, ignoring possible cluster effects. Baseline occurrence of infection episodes among members of a cluster defined by a working team might be influenced by behavioural habits typical of the team influencing the transmission of infections, and in this type of trial, by cluster-specific intensity of applying instructed hand-cleaning procedures and behavioural change. However, successive infection episodes within a cluster are not a direct reflection of transmission of the infectious agent within the cluster but as likely, due to new introductions, i.e., infections contracted in non-work related contacts of individual study participants. Work-related cluster effect is unlikely to affect the occurrence of these latter infections while the incidence of community-based infections is known to vary rather randomly between different subpopulations and by time within a given subpopulation. Therefore, inclusion of the extra variability due to cluster-effect in statistical analysis might result in missing true differences in a set-up like ours where the number of clusters was rather limited for practical reasons .
Our main result of reduced respiratory infections through hand washing with soap and water combined with advice on coughing and sneezing is in agreement with several previously published studies in semi-closed populations such as hospitals and children's day care centres [15, 16]. However, there are no data (6, 7) available on this intervention in adult populations in a regular office environment where contact patterns are likely to be different from the semi-closed environments. A decrease of 50% was reported in the incidence of childhood pneumonia following intensified hand washing with soap and water in Pakistan. The absence of protection from infection episodes in the alcohol hand rub intervention group is at variance with a recent publication on a similar intervention reporting a remarkable reduction of disease episodes  and with some earlier studies [17, 18]. We do not know the reason for the discrepant results but one can speculate about the potential effects of putative differences in the disinfectant composition, varying pre-study levels of hand washing routines, and different study design. Also, we do not want to rule out possibility that alcohol hand rub would have a decreasing effect on infection episodes, if the number of follow-up persons was greater. On the other hand, the current result is in agreement with our unpublished observations on the capacity of a single round of instructed hand cleaning to remove infectious human rhinovirus administered on the skin of the back of the hand. Washing with soap and water appeared to be much more efficient than rubbing with the alcohol-based disinfectant (Savolainen-Kopra et al., unpublished observations). While participants in the control arm were not forced not to use soap and water or alcohol rub, the observed effect in the intervention cannot be due to behavioural recommendations only because the same recommendations were given to both intervention arms IR1 and IR2.
The interventions continued over two winter seasons in order to cover different seasonal virus epidemics. Structurally different viruses might differ in sensitivity to the hand washing procedures employed. However, the influenza pandemic with influenza A H1N1 2009 triggered an intense national hand washing campaign that compromised the implementation of our study, with the control clusters in our study also being exposed to the information in the public media, which was further tailored to all staff by the occupational health units in the participating corporations. Rather than stopping the trial prematurely, we decided to continue it through the planned period and analyze the results in two blocks of follow-up time, "before the pandemic" and "during and after the pandemic", respectively. Subsequently, a Eurobarometer survey in all EU countries showed that in Finland over 40% of the adult population reported to have changed their behaviour so as to improve their protection from influenza . According to that survey, the change was almost exclusively seen in improved hand hygiene, with hardly any change in behaviour related to coughing or sneezing. During and after the pandemic in our study there was no significant difference in the occurrence of infection episodes between the soap-and-water arm and the control arm. This was due to a statistically significant decrease in the number of infection episodes in the control arm, obviously due to the national hygiene campaign. There was no concomitant decrease in the occurrence of infections in the two intervention arms. Somewhat unexpectedly, even if the infection episodes were reduced in the soap and water arm, there was no reduction in the number of sick leave or absence episodes due to infectious disease. Rather, after the onset of the pandemic, the number of episodes in the soap-and-water arm was higher than in the control arm. We speculate that people in the intervention clusters may have obediently followed the overall instructions given at the beginning of the study, including the concept that coming to work with symptoms is likely to put colleagues at risk of contracting the disease and is thus not recommended.
Our study has potential weaknesses. Firstly, we used subjective reporting of disease episodes rather than professional assessment of symptoms and signs of infection. However, we believe that the written instructions, clear definitions repeated in the weekly emails, and rapid responses to any enquiries ensured sufficiently reliable lay-person diagnosis. The simple and user-friendly web-based data collection system with a short recall-time without the need for personal home diaries contributed to high reporting coverage throughout the entire study period even in the control clusters. Secondly, we had no direct measure of individuals' adherence to the given instruction in the different intervention arms. The repeated interviews on transmission-limiting habits indicated that the overall level of implementation of the recommended measures, which was fairly high already at the base line, further improved during the study in both intervention arms and moderately also in the control arm. It is likely that the national anti-pandemic campaign had a major role in the observed "leakage" of transmission limiting behaviour to the control arm. Furthermore, already participation in an intervention trial testing the role of hand hygiene, even if in the control arm without specific instructions, is likely to affect one's behaviour based on common sense and general knowledge. Some "leakage" was therefore expected. This view is also indirectly supported by the intense interest in the study by participants in all arms, recorded by the study nurse during the monthly visits throughout the study. The third interview at the end of the study, several months after the peak in media publicity of the pandemic, suggested that some changes in behaviour among the controls had been short lived, and now the difference between the controls and the intervention arms was much clearer again. A similar finding on the post-pandemic decline in hand sanitizer use was reported from New Zealand in December 2009 following the rapid decline in media coverage of the pandemic . Based on the above, we believe that the participants in our intervention arms followed the instructions fairly well.
Self-evidently a large and long-lasting intervention study among office work employees conducting their regular work requires balancing between scientific ambitions and feasibility, not forgetting costs of the study. However, given the identified limitations of this study we would have suggestions for future hand hygiene studies in order to avoid some of the problems faced in this study. Firstly, active follow up of all participants for illnesses with a mechanism to collect specimens for laboratory testing of as many illnesses as possible would enable a more precise identification of infection etiology and confirm specific effects of interventions. Secondly, it would be wise to include external objective assessments of adherence such as measurement of soap/alcohol usage in offices and observation of visits to sinks, e.g. via electronic tags.