Regular pattern of respiratory syncytial virus and rotavirus infections and relation to weather in Stockholm, 1984–1993
Abstract
Objective: To seek the possible epidemiologic relationship between the two dominant pediatric infectious agents, respiratory syncytial virus (RSV) and rotavirus, and to analyze the relationship of RSV to influenza virus infections and climate.
Methods: In the laboratory register, we retrospectively identified pediatric cases less than 5 years of age from the period 1984–93 (including the winter of 1994). RSV was diagnosed by immunofluorescence in nasopharyngeal samples and rotavirus infections by electron microscopy of feces.
Results: We observed a regular and significant pattern of early RSV epidemics (December to February), alternating every other year with later ones (March to April). There were twice as many hospital admissions during early compared to late epidemics. There was a similar but reverse pattern of early and late rotavirus seasons. Influenza A virus outbreaks occurred during the same period as early RSV epidemics. Several weather factors, such as temperature, precipitation, wind force and humidity were analyzed in relation to RSV epidemics without disclosing an important relationship. Cloudiness was, however, found to be associated with RSV peaks.
Conclusions: The possibility of predicting RSV epidemics may be useful for medical planning.
INTRODUCTION
Infections with respiratory syncytial virus (RSV) and rotavirus comprise the dominant part of pediatric infectious morbidity each year [1,2]. Both infections are also associated with a considerable mortality on a worldwide basis [3,4]. Therefore, both infections are under research for possible means of prevention [5].
There have been some studies of patient risk factors, such as prematurity and cardiac diseases [6,7], but few long-term epidemiologic studies. A better understanding of epidemiologic factors is important for medical planning of care and prevention. The epidemic occurrence of both RSV and rotavirus initiated studies on their seasonality. Climate has been one indicated factor; a correlation has been reported with low temperature and dry weather [8–14]. A statistically significant correlation has also been demonstrated between short diurnal periods of sunshine and the peak of the RSV epidemic [8,9]. Another suggested factor has been interaction with other viral infections, such as influenza A virus infection [14].
In the present study we have examined retrospectively the diagnosed cases of RSV and rotavirus infections during the 10-year period 1984–93. An attempt was made to find whether any epidemiological factors existed that were common for these two infections.
METHODS
Cases
Cases of RSV and rotavirus infection from the Department of Pediatrics, St Göran's Hospital, were identified in the register at the Central Microbiological Laboratory of the Stockholm County Council. Indications for testing for RSV were the same throughout the period, during which the same pediatrician was responsible for the care of patients with infectious diseases. Occurrence of severe symptoms in children with a predicted stay of more than 2 days indicated testing for RSV to facilitate cohort care.
Indications for rotavirus diagnostic testing were less stringent since the hospital time is usually too short to make cohort care relevant. About 30% of children hospitalized with community-acquired diarrhea were tested for possible etiology. Among cases with diarrhea that were tested for the presence of rotavirus, 50% were positive. Rotavirus testing was always done in patients with severe and long-lasting symptoms and also for investigation of clusters, e.g. in day-care centers.
The Department of Pediatrics of St Göran's Hospital is a tertiary referral center with neonatal and pediatric intensive care. The hospital has its own catchment area, with about 8600 infants less than 1 year of age in 1990.
Samples
Nasopharyngeal samples were sent to the virus laboratory and analyzed on the same day. Fecal samples were most often analyzed by electron microscopy on the same day. The number of nasopharyngeal samples was increased during the period, along with the increase in the number of children in the area, and also varied during larger or smaller seasons. The number of fecal samples analyzed varied with regard to different outbreaks.
Detection of RSV and influenza virus infections
Fixed, exfoliated cells from nasopharyngeal aspirates were stained directly by FITC-conjugated monoclonal antibody against RSV (Ortho Diagnostic Systems Inc., NJ, USA). An indirect immunofluorescence technique was applied for detection of influenza virus by monoclonal antibodies, obtained from WHO. Influenza virus was also detected after virus isolation using MDCK cells, a canine kidney cell line, especially suitable for the isolation of influenza viruses. The season of influenza A virus was defined as the period between the first and last case of influenza in the Stockholm area. The same person was responsible for the virologic techniques and routines during the study period.
Electron microscopy for the detection of rotavirus infections
Fecal specimens were examined by electron microscopy using negative staining. A 10% suspension of feces was incubated on carbon/Formvar-coated grids for 1 min, stained with 2% phosphotungstic acid and finally examined at a magnification of × 40 000 in a JEOL 100C electron microscope. The specimens were considered negative if no virus was detected after 10–15 min of examination. The same person was responsible for the technical quality.
Weather reports
The Swedish Meteorological and Hydrological Institute supplied data about the monthly precipitation and wind force, the monthly means of temperature, cloudiness (lack of sunshine) and humidity. They were recorded at 1.00 pm at Bromma Airport, which is situated within the catchment area of St Göran's hospital. This observation moment was chosen instead of other time periods, such as morning or evening, because the sunlight at this time would vary less over the year. Cloudiness was estimated by observation of the sky by meteorologically trained personnel. We arbitrarily defined the cut-off for cloudiness as 75%.
Statistics
The start of an epidemic was defined as the week with at least two cases of RSV. The peak was defined as the middle of the 2-week-period with the highest number of cases. The Wilcoxon rank sum test, Fisher's exact test, and the runs test of randomness were used. ‘Early’ versus ‘late’ epidemics were pooled and compared by one-way analysis of variance, with week number from 9 September as the dependent variable.
RESULTS
Alternating pattern of RSV seasons
The interrelationship of the start, peak of number of cases and alternating pattern of seasonal size of RSV epidemics was documented (Table 1). The seasonal pattern varied every other year, with late (peaks in weeks 13–18) epidemics followed by early (peaks in weeks 49–5) ones (p<0.001, accounting for approximately 50% of the variance). The number of detected cases was significantly greater, around twice as many, during early than during late epidemics. The peak was reached faster during the early seasons as compared to the late ones; 8.6 versus 11.4 weeks (p < 0.05, using the Wilcoxon rank sum test).
| Season | Starting week | Week of peak | No. of RSV cases | Size compared to previous one |
|---|---|---|---|---|
| 1984–85 | 9 | 18 | 64 | < |
| 1985–86 | 43 | 3 | 132 | > |
| 1986–87 | 1 | 13 | 45 | < |
| 1987–88 | 49 | 5 | 109 | > |
| 1988–89 | 5 | 15 | 92 | < |
| 1989–90 | 49 | 5 | 178 | > |
| 1990–91 | 7 | 13 | 127 | < |
| 1991–92 | 49 | 5 | 211 | > |
| 1992–93 | 7 | 17 | 74 | < |
| 1993–94 | 39 | 49 | 236 | > |
In the catchment area, as well as in the rest of Sweden, there was a 40% increase in the birth rate from 1983 to 1991. There was a stable seasonal variation of 40 % with the lowest rate in November and the highest in March. Although the total number of cases of RSV increased from 1984 to 1992, in agreement with the increase in birth rate during this period, the alternating pattern of small and large epidemics remained the same and was statistically significant (p< 0.025, using the runs test of randomness).
RSV seasons and relationship to influenza virus and rotavirus seasons
The distribution over time of the individual cases of RSV is shown in Figure 1, as well as the distribution of rotavirus infections and the influenza virus season. Peaks occurred from November through April. Few seasons had cases in August and September, but cases were absent only from one calendar week, no. 35. Early epidemics usually occurred during the influenza virus season and late ones after. Apart from this, the relation to influenza virus seasons showed no consistent pattern.
Biweekly frequencies of rotavirus and RSV detections during 10 consecutive seasons. Each x-axis covers two seasons. It starts on week 37 (9 September), breaks and ends on week 26 (1 July). Vertical dotted lines are positioned at 1 January. RSV is represented by black bars, and rotavirus by gray bars. Rectangles show yearly influenza A occurrence.
Cases with rotavirus infections during a season were detected for a longer time than cases of RSV infections (Figure 1). Although less evident than for RSV, the rotavirus seasons also alternated with early peaks at weeks 49–6 and later ones at weeks 11–17 (p< 0.001, accounting for approximately 20% of the variance). Early RSV seasons appeared with late rotavirus seasons, and vice versa.
Influence of weather components
There was no obvious relationship between RSV peaks and weather components such as precipitation (as rain and/or snow) or wind force. In years with early RSV peaks, these coincided with winter temperatures but late peaks occurred with rising spring temperatures (Figure 2). The study period included winters with relatively low temperatures (such as 1984–85 and 1986–87) as well as with relatively high temperatures (such as 1988–89 and 1989–90), without any change in the overall pattern of alternating late and early RSV peaks. The increased relative humidity during winter months coincided only with early RSV peaks. Late peaks appeared with decreasing relative humidity.
Climate variation in relationship to RSV infections over 8 years. The number of RSV cases is as shown in Figure 1, but accumulated over a month and not represented by a vertical axis. Empty squares represent RSV cases, half-filled squares represent cloudiness (%), and empty circles represent temperature. As a reference, the mean temperature over the years 1961–90 (triangles) is given in the top left graph (1984–85).
The only weather component that coincided with both early and late RSV peaks was cloudiness. Thus, 10 of 23 months with at least 10 cases of RSV were cloudy, as compared to 15 of 72 months with less RSV present (p = 0.02, using one-sided Fisher's exact test). During the 1986–87 season, which had little cloudiness, there was only a minor peak of RSV.
DISCUSSION
The epidemiologic characteristics of RSV and rotavirus infections have been observed over a decade. There was a regular pattern of RSV seasons. Later-appearing seasons had relatively few patients, and early ones usually more cases. Early and late rotavirus seasons were also observed with a pattern that was the reverse of that seen with RSV. Rotavirus seasons lasted longer than RSV seasons.
Although some previous studies of RSV epidemiology have suggested an alternating pattern with early and late start of the season, as observed by us in the Stockholm area [14–16], this was not found in a recent study, including several laboratories, in the USA [17]. The large geographic areas that were studied in the USA might have concealed an alternating pattern in smaller regions. In addition, we demonstrated that epidemics that started early involved more patients, as suggested by others [14,16].
Since the majority of children in this area are seropositive for RSV by 2 years of age (85% in a study by Forsgren and Eriksson, unpublished), the transmission rate of RSV during a given season must be great. The hospitalized cases with rotavirus infections occur in somewhat older children than do those with RSV infections. The relative role of maternal immunity in this age difference is not fully known.
Immunity following the first infection is not long-lasting, and we saw reinfection in 10 hospitalized cases of RSV and two of rotavirus, respectively. Repeated rotavirus infections improve the immunity both to the types that the subject has been infected with and to other types [18]. Symptoms also decrease, so that after two rotavirus infections the children are protected against severe diarrhea [18]. Similarly, with regard to RSV, there seems to be an increased immunity after a second RSV infection, irrespective of whether this was caused by the homologous type or heterologous type of RSV, i.e. type A or B of RSV [19].
RSV infections are mainly transmitted by large droplets of secretion, containing RSV. The main sources are older children, especially siblings [20]. Aerosols seem to increase the efficiency of spread [21]. Rotavirus is most often transmitted by the fecal-oral route but the respiratory route has also been suggested [22,23]. During the studied years, simultaneous RSV and rotavirus infections were only seen in hospitalized RSV patients with nosocomial rotavirus infection.
It is well known that infections with rotavirus and RSV occur in peaks in the winter months, at least in the temperate parts of the world [6–10]. Long-term seasonal distribution has been best studied for RSV [7], while rotavirus infection has only been studied for a few seasons [6]. Several hypotheses have been investigated with regard to the seasonal variation of RSV, such as climate and viral interference [8,12,14,24]. Climate has been discussed for variation of rotavirus infections as well [10,11]. Our study of several weather components showed that weather most probably plays a modest or no role in the alternating pattern of seasons. Cloudiness (lack of sunshine around noon) might have influenced the size of the peak. This may be because more children stay indoors and have closer contacts with other children.
For rotavirus infections a geographic seasonal distribution has been described in the USA, starting in the south-west and spreading to the north-east [25]. For RSV, variation in geographic distribution has been reported between neighboring countries, with high peaks in Norway and England when there were no cases in Sweden and Finland [24].
With regard to the management problem associated with the care of patients when their number increases rapidly, such as during early RSV epidemics, the results presented here ought to be valuable for allocation of medical resources, as well as prevention studies in the Stockholm area.
Acknowledgment
We wish to express our gratitude to the late Ms Kerstin Brebäck for expert technical assistance in the electron microscopy work. We also want to thank Mr Pascalis Beoglou, Ms Anne Qvist and Ms Yvonne Öberg for excellent technical assistance throughout this period at the former Central Microbiological Laboratory of the Stockholm County Council, from where these laboratory results were obtained.
