The serologic diagnosis of Q fever was performed at the National Reference Center by using the immunofluorescence reference technique as previously described (29). The titers of immunoglobulin (Ig) G, IgM, and IgA against phases I and II of C. burnetii were determined. A case of evolving Q fever (acute or chronic) was diagnosed when the phase II IgG titer was >200 and the phase II IgM titer was >50. A diagnosis of chronic Q fever was made when the phase I IgG titer was >800 (29).

Marseille is a city located in southern France with 1 million inhabitants. About 40 km northwest of Marseille is a large, natural, saltwater lake called "Etang de Berre" (Figure 1). Northwest of Etang de Berre is a 600-km², semidesert plain region called "La Crau," which is the only steppe in Western Europe. La Crau is limited by the Alpilles Mountains (north), the Mediterranean Sea (south), the Etang de Berre (west), and the Rhône River (east). The irrigated northern area (humid Crau) is devoted to hay cultivation (120 km²), whereas the stony and dry southern part is devoted to sheep grazing.

Data on sheep breeding were obtained from the "Chambre d’Agriculture" in Aix en Provence. These data included the approximate number of sheep and the features of sheep breeding in the area.

Meteorologic data were obtained from the Meteo France weather station at Istres in the form of cumulative wind speed and direction compass cards from January 1996 to December 1999 and monthly compass cards and data sheets for November and December between 1996 and 1999. Wind speed and direction are measured every 3 hours, providing about 240 data points per month. The tables provided by Meteo France show the wind direction and speed in three meter per second (m/s) speed ranges of 2–4, 5–8, and >8. We considered only the mistral, which is the most common and strongest wind and blows for several consecutive days, in sunny and dry conditions. It mostly originates from between the west-northwest and the north and blows through La Crau before reaching the study area. Winds that occur in other local climatic conditions, such as rainfall and high humidity, come primarily come from other directions, mainly the east and south, and blow for short periods only. These winds were not considered in our study. Monthly cumulative precipitation data between 1996 and 1999 in Istres, and daily precipitation for October, November, and December, 1998 and 1999, were also obtained from Meteo France (www.meteo.fr).

Using the serologic criteria described above, Coxiella burnetii–positive patients were identified from the database of the National Reference Center between 1996 and 1999. The eastern part of La Crau, along the Etang de Berre, was considered the study area. All patients had conditions diagnosed when they were inpatients or outpatients at the general hospital in Martigues or in private laboratories in the cities of Martigues, Fos-sur-Mer, Istres, Sausset-les-Pins, and Châteauneuf-les-Martigues (outpatients who were sent to the laboratory by their general practitioner). Medical practices in this area send all serum specimens for Q fever diagnosis to the National Reference Center or, if they test serum specimens by immunofluorescence with phase II antigen provided by the National Reference Center, send positive serum specimens to the center for confirmation. All serologic diagnoses of Q fever are therefore ultimately made at the National Reference Center with the reference serologic technique. Therefore, we may assume that all diagnosed cases are included in our database. For each patient with a diagnosis of Q fever, a questionnaire was filled out, which provided administrative, epidemiologic, clinical, and biologic data. Patient data were included in the study only if the patient’s place of residence (as recorded in the hospital or laboratory files) was in the study area.

All data were managed by using EpiInfo 6 (Centers for Disease Control and Prevention, Atlanta, GA). Pearson’s chi square or two-tailed Fisher exact test was used to compare frequencies of qualitative data. A difference was considered significant when p < 0.05.

Between 1996 and 1999, acute Q fever was diagnosed in 73 patients in the study area. Figure 2 shows the cumulated monthly distribution of these cases (based on date of onset of symptoms) and indicates the usual main peak in incidence of infection in late spring (May–June). When this distribution was considered by year (Figure 3), an unusual peak was observed in December 1998 (5 cases) and January 1999 (2 cases). The proportion of cases occurring in December 1998, in relation to the total number of cases in 1998 (5/20), was significantly higher than the proportion of cases occurring in December in the other study years (1/53) (p < 0.006, Fisher exact test).

The seven patients whose diagnosis was made in December 1998 and January 1999 were interviewed: they were six men and one woman, with a mean age of 34 years. Two case-patients had cardiac valve disorders. Four of them reported living in rural areas. One of them lived in La Crau (Entressen), and six lived on the opposite (southeastern) shore of Etang de Berre (Châteauneuf-les-Martigues, Saint Mitre, Marignane, and Gignac) (Figure 1). None reported an occupational exposure. Two case-patients reported a usual exposure to sheep; one had been exposed to parturient sheep. Two persons owned a cat, and four owned a dog. None of these pets was a newborn or had given birth recently. No patients reported consuming unpasteurized milk or cheese.

Sheep have been bred in La Crau for centuries. The main lambing season (80% of births) is in October and November (Figure 2) and takes place either indoors or outdoors. In the latter case, birth products (mainly placentas) are left on the ground where they desiccate and can be a source of fomite spread of C. burnetii. Between October and February, sheep are allowed to graze hay in the northern humid part of La Crau. A second lambing season (20% of births) occurs in March. Between March and June, sheep graze in the dry southern part of La Crau. In June, the sheep are moved by trucks to their summer pastures in the Alps so sheep are generally absent from La Crau until September.

During the 4 years of the study, the mistral represented 44.9% of all winds >2 m/s, 59.4% of winds >5 m/s, and 80% of winds >8 m/s, which confirms the importance of the mistral in terms of frequency of occurrence and speed. To explain the unusual peak of Q fever cases in December 1998 and January 1999, we studied the wind frequencies 1 month before, i.e., in November and December. Table 2 shows the percentage of northwest wind for the three defined speed ranges, by year; p refers to the comparison between the year under consideration and 1998. In each speed range, the frequency of the mistral was higher in November and December 1998 than in any other year: for the speed range >2 m/s, it was 60.9% in November to December 1998, whereas it was 41.3, 40.5, and 52.4 in November to December 1996, 1997, and 1999, respectively. This unusual frequency was even more notable for the strongest winds (which are more consistent with the mistral), representing 93.5% of the wind in November to December 1998 but only 63.1%, 47.7%, and 85% in November to December 1996, 1997, and 1999, respectively. All these differences were statistically significant, except those between 1998 and 1999 in the speed range >5 m/s.

In terms of precipitation (Figure 4), October and November 1999 were much more rainy (190.6 mm and 54.4 mm of rainfall, respectively) than October and November 1998 (34.8 mm and 14.2 mm, respectively). The rainfall amount of December 1999 (7.2 mm) was lower than that of December 1998 (56.8 mm). Exceptionally high amounts of precipitation (156.6 mm) were noted in October 1999. The global precipitation of December 1998, primarily depends on that of December 31 (43.4 mm of 56.8 mm). Figure 5 shows the mean monthly rainfall from 1995 to 1999. The main lambing in autumn occurs during the period of the most rainfall, whereas the spring secondary lambing occurs during a dry period.