Analysis of the causes of spawning of large-scale, severe malarial epidemics and their rapid total extinction in western Provence, historically a highly endemic region of France (1745-1850)

Abstract

Background: The two main puzzles of this study are the onset and then sudden stopping of severe epidemics in western Provence (a highly malaria-endemic region of Mediterranean France) without any deliberate counter-measures and in the absence of significant population flux.

Methods: Malaria epidemics during the period from 1745 to 1850 were analysed against temperature and rainfall records and several other potentially relevant factors.

Results: Statistical analyses indicated that relatively high temperatures in early spring and in September/October, rainfall during the previous winter (principally December) and even from November to September and epidemics during the previous year could have played a decisive role in the emergence of these epidemics. Moreover, the epidemics were most likely not driven by other parameters (e.g., social, cultural, agricultural and geographical). Until 1776, very severe malarial epidemics affected large areas, whereas after this date, they were rarer and generally milder for local people and were due to canal digging activities. In the latter period, decreased rainfall in December, and more extreme and variable temperatures were observed. It is known that rainfall anomalies and temperature fluctuations may be detrimental to vector and parasite development.

Conclusion: This study showed the particular characteristics of malaria in historical Provence. Contrary to the situation in most other Mediterranean areas, Plasmodium falciparum was most likely not involved (during the years with epidemics, the mean temperature during the months of July and August, among other factors, did not play a role) and the population had no protective mutation. The main parasite species was Plasmodium vivax, which was responsible for very severe diseases, but contrary to in northern Europe, it is likely that transmission occurred only during the period where outdoor sporogony was possible, and P. vivax sporogony was always feasible, even during colder summers. Possible key elements in the understanding of the course of malaria epidemics include changes in the virulence of P. vivax strains, the refractoriness of anophelines and/or the degree or efficiency of acquired immunity. This study could open new lines of investigation into the comprehension of the conditions of disappearance/emergence of severe malaria epidemics in highly endemic areas.

Figure 1

Cassini’s map of the western Provence (1770–1776). The geographic coordinates of this map are approximately 43.15 to 44.0 N and 4.20 to 5.30 W. The three studied areas have been surrounded by a black line; they carry the name of the underlined cities (i.e., Avignon, Arles and Berre). In the south-east, the area of Arles is bounded by the western arm of the Rhone. The positions of the towns and villages of the three studied areas mentioned in the text have been postponed. As meteorological data were measured in Marseille, this city has also been labeled. Data other than names of the towns or villages are in italics. Areas identified as wetlands by Cassini were stained in light blue, but the floodplains have not been indicated to not overload the figure. This map shows the great expanses of marsh and swamplands. Adapted from [9].

Figure 2

Number of malaria epidemics in western Provence. The number was given by area and by decade from 1541 to 1850.

Figure 3

Marseille climate diagram for the 1745–1850 period. Average temperatures (solid line) and precipitation (broken line). The horizontal blue and red lines indicate minimal temperatures required for P. vivax (15°C) and P. falciparum (18°C) development, respectively. These figures are not precise thresholds, but they illustrate a climate above which temperatures are generally favourable for transmission.

Figure 4

Mean temperatures in June and September at Marseille. The horizontal red line indicates minimal temperature required for P. falciparum (18°C) development.

Figure 5

Principal Component Analysis for monthly mean temperatures, with the binary epidemic factor as supplementary variable. For November and December, the mean temperatures of the previous years were used. On the one hand, the horizontal axis, accounting for 17% of the total variance, gives a temperature gradient. Hence, the more a year is situated on the left, the warmer it is globally. On the other hand, the vertical axis, accounting for 13% of the total variance, permits to identify the years with warm intermediate seasons (situated at the bottom of the graphic). Years with and without epidemic are indicated with red circles and green squares, respectively. Thus epidemics were observed in years that were generally warmer, especially for intermediate seasons. Moreover, years with epidemic seem to keep a consistent profile with respect to monthly temperatures, since the associated ellipse is smaller.

Figure 6

Total rainfall by season at Marseille. The last great epidemic period (1772 to 1776) has been underlined in grey. The seasons were interpreted as follows: winter (December of the previous year, January and February), spring (from March to May), summer (from June to August) and fall (from September to November). For reasons of readability, only the period from 1750 to 1790 is shown.

Figure 7

Theoretical probability of observing at least one epidemic according to the total amount of precipitation. The generalized linear model with a Poisson error distribution was used to compute the fitted probabilities. The fitted values are plotted against the amounts of precipitation (from previous November to current September). The more it rains in the previous winter, current spring and summer, the more probable epidemics occur.

Figure 8

Boxplot for the mean of the temperatures from July to October. The box plot gives a representation of the concentration of the mean temperature (from July to October) over the period 1745 to 1850. The years 1774 and 1756 are pointed out. The middle 50% of the years fall inside the grey box, whereas the 25% warmest and 25% coldest years are above and below the box, respectively. The black line in the rectangle corresponds to the median temperature.

Figure 9

Multivariate analysis to illustrate the interrelationships between the favourable factors. Abbreviations for variables: RainD, rainfall during the December month of the previous year; RainNtoS, rainfall from November to September; TempPrevSummer, mean temperature of the summer of the previous year; TempM, mean temperature of March; TempSO mean temperature of September-October; PrevEpid, presence of an epidemic during the previous year. On the one hand, the horizontal axis, accounting for 27.2% of the total variance, gives a precipitation gradient. Hence, the more a year is situated on the left, the more rainy it is globally. On the other hand, the vertical axis, accounting for 26.4% of the total variance, summarizes the factors due to temperature and to the presence of an epidemic the previous year. The binary epidemic factor is used as a supplementary variable. Years with and without epidemics are symbolized by red circles and green squared symbols, respectively. Thus epidemics were more frequently observed in years that were warmer and rainier (the convenient months).

Figure 10

Rainfall in the December month before and after the breakpoint year (1776). The mean monthly total rainfall before and after this date is indicated by brown and green horizontal lines, respectively.

Figure 11

Study of the fluctuations of the mean temperature from May to August using a Gaussian white noise model. The year 1776 was the epidemic breakpoint. An Arima model was used to plot prediction intervals at confidence levels of 80% and 95% in light grey and grey, respectively.