- Email: [email protected]
MORTALITY FROM PINK DISEASE IN 1923-47
608
MORTALITY FROM PINK DISEASE IN 1923-47 WILLIAM
P.
D.
LOGAN
M.D., B.Sc. Glasg., D.P.H.
From the
General
Register Office INTEREST in this relatively rare ’and
setiologically
obscure disease of childhood has been stimulated by the suggestion of Watkany and Hubbardthat mercury administered in teething powders or in ointments may be the causal agent ; out of 20 cases of pink disease mercury was found in the urine of 18. It may be opportune, therefore, to review the mortality of pink disease in England and Wales during the past twenty-five years. The synonyms for pink disease include erythroedema,
erythrcedema polyneuritica, dermato-polyneuritis, and acrodynia, but their use on death certificates is being steadily given up in favour of the simpler name. Table i. gives’the number of deaths and the death-rate from pink disease in each year from 1923 to 1947, and fig.1 illustrates the trend of the’death-rate over those years. Starting with 1 death in 1923 (a rate of 0-3 per million) the recorded deaths and the death-rate progressively 1. Warkany,
J., Hubbard, D. M.
Lancet, 1948, i, 829.
SEX AND AGE, AND ANNUAL DEATH-RATES PER MILLION CHILDREN AGED UNDER 5 YEARS, IN ENGLAND AND WALES IN 1923-47
TABLE
I-ANNUAL
DEATHS
BY
(ALL AGES),
as the disease became more widely recognised, up to 1936, when there were 88 deaths (31-4 per million). For the next three years the numbers fluctuated, dropped to a much lower level in 1940 (49 deaths, 17-0 per million), and remained practically constant at this lower level during the next six years. It is difficult to believe that this was a chance sequence, and it cannot be explained by changes in the birth-rate, which behaved dissimilarly. In 1947, however, deaths rose sharply ; the number (103) was the highest yet recorded, and the death-rate (29-2 per million) was almost as high as in 1936. The figures, it should be noted, have not been corrected for the changes in classification that came into force in 1940. But application of the appropriate conversion factor would only partially reduce the drop in 1940 and would not materially alter the picture. The association here revealed between the mortality of pink disease and the war may or may not indicate a real temporary change in the behaviour of thedisease. The circumstances of war may have altered the standard
increased,
of death cer-
of
tification
pink disease; or may have its reduced
prevalence fatality. As table
or i
shows, variations in
the
predominating sex occurred
randomly from year to In year. of numbers deaths each sex
predomi-
nated in twelve of the
twenty-five
I-Annual death-rates from
disease
Fig. pink per years, wheremillion children under 5 years of age in England and Wales in 1923-47. as in 1945 the deaths were divided equally between the sexes. The female death-rate exceeded the male in thirteen years. Table 11 shows that, aggregating years 1923-47, there was no significant difference between the sexes : 49-2% of deaths in males, and 50-8% in females. The percentage for each sex differs from expectation (the 25-year average proportion of each sex in the child population : males 50-8% and females 49-2%) by 1-6%, with a standard error of 1-5%. Most deaths occurred under the age of 1 year, and In four of the earlier years none over the age of 9 years. (1923, 1924, 1928, and 1932) deaths at ages over 1 year were more numerous than those under 1 year; but after 1932 the lowest age-group has persistently shown an excess ranging from 7 to 32, except in 1940, when the excess was only 1 death. Table 11 shows that, of all the deaths in 1923-47, about 61% were in children under 1 year, 38% at TABLE II-DISTRIBUTION OF DEATHS IN ENGLAND AND WALES IN BY SEX AND AGE
1923-47,
Though 9 deaths are included at ages 5-10, these death-rates calculated on the basis of the 0-5 age’group populations, a course which -was considered more appropriate here than to exclude the 9 deaths or to base the rates on the 0-10 age-group populations. *
were
609 (FOR
TABLE III—DEATHS
1
YEAR
AND
1940-:-46
COMBINED
UNDER 5 YEARS
SEXES)
IN
AT
ENGLAND
AGES
UNDER.
AND
WALES,
, -
(a) Under 1 year
-
about regional populations at ages during the war years, total live births in each region, aggregated for the- same years, have been used as a base for the calculation of "death-rates." Such rates, it should be noted, do not take account of migrational movements of the child population during the war. There are highly significant differences between 117-89. P< the rates in the twelve regions in North i The rate was recorded 0-0001%). highest (201-4 per million), followed by Wales 11 (168-7) The lowest rate occurred in the and North ill (114-3). South-west region (24-1), followed by, the South-east -(38-3), and then by Greater London (40-9). These regional differences’ are illustrated, and may be com- -pared region with region, in fig. 2. The regional distribution of deaths in 1947 is shown in the last column of table 111. It will be seen that the
(&khgr;2
ages 1-4 years, and less than over.
There
was
1%
no significant
=
at 5 years of age or difference in age-
distribution between the sexes (’1.2 2-63, P > 10%). For the years 1940-46 table ill provides a further analysis of deaths by age : 63.5% of all deaths under 5 years were in children aged under 1 year, and 33-2% between 1 and 2 years. Among deaths under 1 year over 85% took place in the second 6 months of life. There were few deaths under 6 months or over 2 years of age. Discussing the age-distribution of morbidity in pink disease, Fisher,2 who had formed the impression that with the paSsage of time the " age of attack" was becoming lower, found, on analysis of his figures, no evidence of any significant age-shift. In our present series of deaths, on the other hand, there is evidence of a significant age-shift between 1923 and 1947 from the higher to the lowest age-group. (By Spearman’s rank correlation technique, p 0 63, significant by Student’s t test -at a level of between 0-1%.) The deaths, aggregated for the years 1940-46, in each of the twelve main regions of England and Wales are contained in table iv. In the absence of information =
-
=
2.
Fisher, T. N.
TABLE IV-REGIONAL
AND
IN
1947,
AND
LIVE BIRTHS IN
Brit. med. J. 1947, i, 251.
DISTRIBUTION OF DEATHS IN REGIONAL DEATH-RATES PER
1940-46,
1940-46
2-Regional death-rates from pink disease per million- live births in 1940-46, with abbreviated confidence intervals to test whether differences are 5 % significant.
Fig.
-
Method of Construction The abbreviated confidence interval technique has been devised by a. statistician of the General Register Office. The abbreviated intervals are obtained by reducing, in the proportion ! : 2, the confidence intervals of Neyman. The diagram provides a useful approximation to the results that would be obtained by carrying out 66 tests of significance to compare each possible pair of regions. Before using this technique it is essential, as has been done here, first to establish a significant general heterogeneity of the regions by means of a X2 or other equivalent test. °
-
-
MILLION
IN ENGLAND AND WALES
Method of Use Note whether the rectangles representing any chosen pair of regionsoverlap on the vertical scale. If they overlap-e.g., North III and Eastthere is no significant difference between the two regions. If they do not overlap—e.g., North III and Midland i-there is a significant difference between them. A significant difference can be taken as one which is more than twice its standard error, and this criterion of significance was adopted in constructing the diagram. -
-
increase in deaths during that year was not concentrated in any one part of the country but was shared among the regions in about the same proportions as in previous years. It is conjectural whether these differences in death-rate = between the regions are due to differences in morbidity, in fatality, or in standards of diagnosis. It would be interesting to know whether there exist regional differences of habit in the administration of mercurial medicaments to young children.
large
_
,
SUMMARY
The death-rate of pink disease fell to a constant low level during the late war but rose sharply in 1947. There was no significant difference in sex-mortality over the years 1923-47. About two-thirds of the deaths were in children aged under 1 year, especially at ages 6-12 months. Of the other third practically all were between the age of and 2 -years. ’-
1 -
’
-
-
*For the geographical constitution of the regions, see- the Registrar-General’s Statistical Review (any recent year).
1940-46 ther& were highly significant difference regional death-rates.
During between
.
MORTALITY FROM PINK DISEASE IN 1923-47 WILLIAM
P.
D.
LOGAN
M.D., B.Sc. Glasg., D.P.H.
From the
General
Register Office INTEREST in this relatively rare ’and
setiologically
obscure disease of childhood has been stimulated by the suggestion of Watkany and Hubbardthat mercury administered in teething powders or in ointments may be the causal agent ; out of 20 cases of pink disease mercury was found in the urine of 18. It may be opportune, therefore, to review the mortality of pink disease in England and Wales during the past twenty-five years. The synonyms for pink disease include erythroedema,
erythrcedema polyneuritica, dermato-polyneuritis, and acrodynia, but their use on death certificates is being steadily given up in favour of the simpler name. Table i. gives’the number of deaths and the death-rate from pink disease in each year from 1923 to 1947, and fig.1 illustrates the trend of the’death-rate over those years. Starting with 1 death in 1923 (a rate of 0-3 per million) the recorded deaths and the death-rate progressively 1. Warkany,
J., Hubbard, D. M.
Lancet, 1948, i, 829.
SEX AND AGE, AND ANNUAL DEATH-RATES PER MILLION CHILDREN AGED UNDER 5 YEARS, IN ENGLAND AND WALES IN 1923-47
TABLE
I-ANNUAL
DEATHS
BY
(ALL AGES),
as the disease became more widely recognised, up to 1936, when there were 88 deaths (31-4 per million). For the next three years the numbers fluctuated, dropped to a much lower level in 1940 (49 deaths, 17-0 per million), and remained practically constant at this lower level during the next six years. It is difficult to believe that this was a chance sequence, and it cannot be explained by changes in the birth-rate, which behaved dissimilarly. In 1947, however, deaths rose sharply ; the number (103) was the highest yet recorded, and the death-rate (29-2 per million) was almost as high as in 1936. The figures, it should be noted, have not been corrected for the changes in classification that came into force in 1940. But application of the appropriate conversion factor would only partially reduce the drop in 1940 and would not materially alter the picture. The association here revealed between the mortality of pink disease and the war may or may not indicate a real temporary change in the behaviour of thedisease. The circumstances of war may have altered the standard
increased,
of death cer-
of
tification
pink disease; or may have its reduced
prevalence fatality. As table
or i
shows, variations in
the
predominating sex occurred
randomly from year to In year. of numbers deaths each sex
predomi-
nated in twelve of the
twenty-five
I-Annual death-rates from
disease
Fig. pink per years, wheremillion children under 5 years of age in England and Wales in 1923-47. as in 1945 the deaths were divided equally between the sexes. The female death-rate exceeded the male in thirteen years. Table 11 shows that, aggregating years 1923-47, there was no significant difference between the sexes : 49-2% of deaths in males, and 50-8% in females. The percentage for each sex differs from expectation (the 25-year average proportion of each sex in the child population : males 50-8% and females 49-2%) by 1-6%, with a standard error of 1-5%. Most deaths occurred under the age of 1 year, and In four of the earlier years none over the age of 9 years. (1923, 1924, 1928, and 1932) deaths at ages over 1 year were more numerous than those under 1 year; but after 1932 the lowest age-group has persistently shown an excess ranging from 7 to 32, except in 1940, when the excess was only 1 death. Table 11 shows that, of all the deaths in 1923-47, about 61% were in children under 1 year, 38% at TABLE II-DISTRIBUTION OF DEATHS IN ENGLAND AND WALES IN BY SEX AND AGE
1923-47,
Though 9 deaths are included at ages 5-10, these death-rates calculated on the basis of the 0-5 age’group populations, a course which -was considered more appropriate here than to exclude the 9 deaths or to base the rates on the 0-10 age-group populations. *
were
609 (FOR
TABLE III—DEATHS
1
YEAR
AND
1940-:-46
COMBINED
UNDER 5 YEARS
SEXES)
IN
AT
ENGLAND
AGES
UNDER.
AND
WALES,
, -
(a) Under 1 year
-
about regional populations at ages during the war years, total live births in each region, aggregated for the- same years, have been used as a base for the calculation of "death-rates." Such rates, it should be noted, do not take account of migrational movements of the child population during the war. There are highly significant differences between 117-89. P< the rates in the twelve regions in North i The rate was recorded 0-0001%). highest (201-4 per million), followed by Wales 11 (168-7) The lowest rate occurred in the and North ill (114-3). South-west region (24-1), followed by, the South-east -(38-3), and then by Greater London (40-9). These regional differences’ are illustrated, and may be com- -pared region with region, in fig. 2. The regional distribution of deaths in 1947 is shown in the last column of table 111. It will be seen that the
(&khgr;2
ages 1-4 years, and less than over.
There
was
1%
no significant
=
at 5 years of age or difference in age-
distribution between the sexes (’1.2 2-63, P > 10%). For the years 1940-46 table ill provides a further analysis of deaths by age : 63.5% of all deaths under 5 years were in children aged under 1 year, and 33-2% between 1 and 2 years. Among deaths under 1 year over 85% took place in the second 6 months of life. There were few deaths under 6 months or over 2 years of age. Discussing the age-distribution of morbidity in pink disease, Fisher,2 who had formed the impression that with the paSsage of time the " age of attack" was becoming lower, found, on analysis of his figures, no evidence of any significant age-shift. In our present series of deaths, on the other hand, there is evidence of a significant age-shift between 1923 and 1947 from the higher to the lowest age-group. (By Spearman’s rank correlation technique, p 0 63, significant by Student’s t test -at a level of between 0-1%.) The deaths, aggregated for the years 1940-46, in each of the twelve main regions of England and Wales are contained in table iv. In the absence of information =
-
=
2.
Fisher, T. N.
TABLE IV-REGIONAL
AND
IN
1947,
AND
LIVE BIRTHS IN
Brit. med. J. 1947, i, 251.
DISTRIBUTION OF DEATHS IN REGIONAL DEATH-RATES PER
1940-46,
1940-46
2-Regional death-rates from pink disease per million- live births in 1940-46, with abbreviated confidence intervals to test whether differences are 5 % significant.
Fig.
-
Method of Construction The abbreviated confidence interval technique has been devised by a. statistician of the General Register Office. The abbreviated intervals are obtained by reducing, in the proportion ! : 2, the confidence intervals of Neyman. The diagram provides a useful approximation to the results that would be obtained by carrying out 66 tests of significance to compare each possible pair of regions. Before using this technique it is essential, as has been done here, first to establish a significant general heterogeneity of the regions by means of a X2 or other equivalent test. °
-
-
MILLION
IN ENGLAND AND WALES
Method of Use Note whether the rectangles representing any chosen pair of regionsoverlap on the vertical scale. If they overlap-e.g., North III and Eastthere is no significant difference between the two regions. If they do not overlap—e.g., North III and Midland i-there is a significant difference between them. A significant difference can be taken as one which is more than twice its standard error, and this criterion of significance was adopted in constructing the diagram. -
-
increase in deaths during that year was not concentrated in any one part of the country but was shared among the regions in about the same proportions as in previous years. It is conjectural whether these differences in death-rate = between the regions are due to differences in morbidity, in fatality, or in standards of diagnosis. It would be interesting to know whether there exist regional differences of habit in the administration of mercurial medicaments to young children.
large
_
,
SUMMARY
The death-rate of pink disease fell to a constant low level during the late war but rose sharply in 1947. There was no significant difference in sex-mortality over the years 1923-47. About two-thirds of the deaths were in children aged under 1 year, especially at ages 6-12 months. Of the other third practically all were between the age of and 2 -years. ’-
1 -
’
-
-
*For the geographical constitution of the regions, see- the Registrar-General’s Statistical Review (any recent year).
1940-46 ther& were highly significant difference regional death-rates.
During between
.