Geographically selective assortment of cycles in pandemics: meta-analysis of data collected by Chizhevsky

Abstract

In the incidence patterns of cholera, diphtheria and croup during the past when they were of epidemic proportions, we document a set of cycles (periods), one of which was reported and discussed by A. L. Chizhevsky in the same data with emphasis on the mirroring in human disease of the ~11-year sunspot cycle. The data in this study are based on Chizhevsky’s book The Terrestrial Echo of Solar Storms and on records from the World Health Organization. For meta-analysis, we used the extended linear and nonlinear cosinor. We found a geographically selective assortment of various cycles characterizing the epidemiology of infections, which is the documented novel topic of this paper, complementing the earlier finding in the 21st century or shortly before, of a geographically selective assortment of cycles characterizing human sudden cardiac death. Solar effects, if any, interact with geophysical processes in contributing to this assortment.

Fig. 1.

Graph of A. L. Chizhevsky on cases of cholera recorded in Moscow between 1823 and 1923, published by Sigel [23]. In this graph, the original data are folded over the ∼11-year solar activity cycle. The data show a peak incidence coinciding with maximal solar activity (left panel). Taking these folded data off the published graph, an analysis by cross-correlation of cholera incidence vs. Wolf numbers (WN) yields the largest correlation coefficient at a lag of zero (right) [33]. (© Halberg.)

Fig. 2

[colour online]. Plot of yearly mortality data from cholera in India from 1901 to 1961, shown as original values (solid curve) and after removal of a linear trend (dashed curve). Data from Chizhevsky (1901–1924) and from WHO reports (1920–1961) during 1920–1924 are the same in both sources. * Prediction of outbreaks by Chizhevsky (two correct). (© Halberg.)

Fig. 3.

Least squares spectrum of yearly detrended data on mortality from cholera in India during 1901–1961. The ∼11·7-year cycle is detected with statistical significance and it is validated by nonlinear least squares, the period [and its confidence interval (CI)] being estimated as 11·668 (95% CI 10·44–12·895) years, similar to the ∼11-year solar activity cycle, as suggested by Chizhevsky. (© Halberg.)

Fig. 4.

In order to visualize the waveform of the ∼11·7-year cycle characterizing mortality from cholera in India during 1901–1961, the yearly detrended data were stacked over an idealized 11·7-year scale (after removal of a linear trend), using five bins (classes). The ∼11·7-year component was found to be statistically significant by one-way analysis of variance (P = 0·024). Data from Chizhevsky (1901–1924) and from WHO reports (1920–1961) during 1920–1924 are the same in both sources. (© Halberg.)

Fig. 5

[colour online]. Plot as a function of time of yearly data on mortality from cholera in Russia for 104 years (1823–1926) (solid curve), compared to changes in solar activity, gauged by Wolf numbers (dashed curve). Peaks in the incidence of cholera correspond to every other peak in solar activity, suggesting the presence of a ∼22-year (Hale) cycle. Data from Chizhevsky [3]. (© Halberg.)

Fig. 6.

Least squares spectrum of yearly data on mortality from cholera in Russia (1823–1926), revealing the presence of several peaks corresponding to components with periods of ∼21, 9, and 5·6 years. Whereas the ∼9-year cycle only reaches borderline statistical significance by nonlinear analysis, the other two components are validated with statistical significance. Period estimates and their CIs (given in parentheses) are listed above their respective spectral peaks. Nonlinear result and 95% confidence interval (CI); for 1823–1924, Chizhevsky's data [3] were used and compared with the data of Pollitzer [5] and added for 1924–1926. * Found for Chizhevsky originally by Vladimir Shostakovich. Borderline significance (non-overlap of zero by CI of amplitude only with one-parameter CI). (© Halberg.)

Fig. 7

[colour online]. (a) In order to visualize the waveform of the ∼20·7-year cycle characterizing mortality from cholera in Russia during 1823–1926, the yearly data were stacked over an idealized 20·714-year scale, using 11 bins (classes). This component is validated with statistical significance both by cosinor (P = 0·004) and by one-way analysis of variance (P < 0·001). This component corresponds to a spectral peak and is validated nonlinearly, based on the ‘conservative’ CI (see earlier footnote for explanation of conservative CI). (b) In order to visualize the waveform of the ∼5·6-year cycle characterizing mortality from cholera in Russia during 1823–1926, the yearly data were stacked over an idealized 5·62-year scale, using six bins (classes). This component is validated with statistical significance both by cosinor (P = 0·002) and by one-way analysis of variance (P = 0·015). This component corresponds to a spectral peak and is validated nonlinearly, based on the ‘conservative’ CI (see earlier footnote for explanation of conservative CI). (c) In order to visualize the waveform of the ∼9-year cycle characterizing mortality from cholera in Russia during 1823–1926, the yearly data were stacked over an idealized 8·95-year scale, using five bins (classes). This component is validated with statistical significance both by cosinor (P = 0·019) and by one-way analysis of variance (P = 0·039). This component only reached borderline statistical significance by nonlinear analysis. For 1823–1923 data from Chizhevsky [3] were compared with data from Pollitzer [5] and added for 1924–1926. (© Halberg.)