We use cookies to improve our service and to tailor our content and advertising to you.More infoYou can manage your cookie settings via your browser at any time. To learn more about how we use cookies, please see our cookies policy.
I heard this study reported on the Today programme on Radio 4. What effective dissemination! I applaud the inclusion of SES as a covariate in this study, and I agree that the correlational effects reported are of interest. Nevertheless, I believe that one important aspect of the interpretation of the study is omitted from the discussion. This is the small effect sizes.
I heard this study reported on the Today programme on Radio 4. What effective dissemination! I applaud the inclusion of SES as a covariate in this study, and I agree that the correlational effects reported are of interest. Nevertheless, I believe that one important aspect of the interpretation of the study is omitted from the discussion. This is the small effect sizes.
In Table 2, standardized regression coefficients relating to 140 different analyes are presented. Ignoring the sign, these range from values of 0.00 to 0.16. Squaring a standardized regression coefficient gives a value of R2, which is an estimate of the amount of variance explained by the regression model. Therefore the largest amount of variance explained by any of these models is the square of 0.16, which equals 0.026, or 2.6%. Cohen (1988) has defined small, medium and large effect sizes for values of R2. He defined small effect size as R2 = 0.02, medium effect size as R2 = 0.13, and large effect size as R2 = 0.26. Therefore, all of these are small effects.
The reported p values add credibility to these data, however, one must bear in mind the large sample size employed (n = 5352). As sample size increases, so study power increases, leading to increased sensitivity to detect, and label ‘significant’ even the smallest of effects. Therefore, this ‘superpowered’ study has, not surprisingly, detected many small, and smaller, effects. This is OK in itself, however, in my opinion, some comment as to their ‘real world’ significance should have made. Additionally, the lack of consideration of likely Type I error inflation due to multiple hypothesis testing is a conspicuous omission.
Dr Nishiura [1] accentuated that caution must be exercised in using mathematical models to ascertain the recent SARS epidemics.
The key issue, as we believe, is to understand the model and its results for what they are and, more importantly, for what they are not. It is especially true with the basic reproductive number R0, or its variant the effective reproductive number at time t Rt, which has bee...
Dr Nishiura [1] accentuated that caution must be exercised in using mathematical models to ascertain the recent SARS epidemics.
The key issue, as we believe, is to understand the model and its results for what they are and, more importantly, for what they are not. It is especially true with the basic reproductive number R0, or its variant the effective reproductive number at time t Rt, which has been estimated for the recent SARS outbreaks in Beijing, Hong Kong,, Toronto, Taiwan, and Singapore in several recent articles (e.g. [2-6]). R0, the average number of secondary infections caused by an infective person upon entering a totally susceptible population, is a useful tool to gauge the initial trend of an epidemic. It is also often misunderstood and misused. Indeed, a recent news feature in Nature [7] described the basic reproductive number R0 as "A measure of a disease's infectiousness" corresponds to how many people, on average, are infected by each patient in the absence of any control measures, which erroneously left out the important requirement that the patient must be an index case in that population, i.e. all possible contacts of that person are susceptible to infection.
The effective reproductive number at time t Rt =R0 x(t), where x is the susceptible proportion of population at time t, measures number of infections caused by a new case at time t.[3] It is more important as a mean to understand the progression of the epidemic, taking into consideration the control measures, behavior changes, and climate as they have all been proven to be important in the case of SARS. Moreover, one can approximate the average growth rate of an epidemic over a given time interval while the epidemic is underway from the cumulative case data. From which one could then estimate the "mean effective reproductive number of the observed time period" R*, i.e. the average number of secondary infections caused by one infective person during the observed time interval. The precise definition gives the public officials a clear chronology of progression (or cessation) of the epidemic, albeit retrospectively.
For illustration, we used the cumulative number of probable SARS cases in Taiwan by onset date from March 12 to June 15,[8] exponential curve fitting with first-order autocorrelation in the error structure,[9] and the period of SARS infectivity of 29.03 days (i.e. time from onset to death or discharge) estimated from [10] to obtain the mean effective reproductive numbers for the five distinct periods during March 12 - June 15 (Table 1). A chronology of relevant events of importance is given as a footnote of Table 1. Figure 1 paints a clear picture of slowly growing epidemic in the beginning, to the outbreak kindled by the admission of first SARS patient to Ho Ping Hospital, the site of first hospital cluster infections, on April 9. The peak period of infections (4/11-4/26) ended with the shutdown of Ho Ping Hospital on April 24. The series of hospital clusters in Taipei and subsequently in the southern port city of Kaohsiung finally subsided with the May 11 shutdown of Chang Gung Hospital in Kaohsiung, due to successful intervention efforts to stop nosocomial infections, the last of which occurred shortly before June 9 the onset date of the last hospital infection in Taiwan. The result clearly points to the important lesson from the outbreak in Taiwan shutdown of hospitals where cluster infections have occurred had been a crucial step in breaking the local chains of transmissions. The effect of quarantine measures, however, is less clear and requires further study, perhaps with mathematical modeling. Clearly, retrospective mathematical modeling is an important reference for public health policy makers intending to contain possible future outbreaks with the most effective intervention measures as long as we understand them for what they are and what they are not.
3/18 –
Implementation of Level A quarantine.
4/09 – Admission of first SARS patient to Ho Ping Hospital.
4/24 – Shutdown of Ho Ping Hospital.
4/28 – Implementation of
Level B quarantine.
5/11 – Shutdown of Chang Gung Hospital.
6/15 – Onset date
of the last hospital infection.
References
(1) Nishiura H. Mathematical modeling of SARS: cautious in all our movements. J Epidem Com Health 2003; In Press.
(2) Riley S, Fraser C, Donnelly C, Ghani AC, Abu-Raddad LJ, Hedley AJ, et al. Transmission dynamics of the etiological agents of SARS in Hong Kong: Impact of public health interventions. Science 2003; 300: 961-66 (20 June 2003) Published online 23 May 2003 (10.1126/science.1086478)
(3) Lipsitch, M, Cohen T, Cooper B, Robins JM, Ma S, James L, et al. Transmission dynamics and control of severe acute respiratory syndrome. Science 2003; 300: 1966-70 (20 June 2003) Published online 23 May 2003; 10.1126/science.1086616
(4) Zhou G, & Yan G. Severe Acute Respiratory Syndrome epidemics in Asia. Emerg Infect Dis 2003; 9(12), In Press.
(5) Hsieh YH, Chen CWS, & Hsu SB. The Severe Acute Respiratory Syndrome outbreak in Taiwan: Lessons to be learned. Emerg Infect Dis 2003; To Appear.
(6) Chowell G, Fenimore PW, Castillo-Garsow MA, & Castillo-Chavez C. SARS outbreaks in Ontario, Hong Kong, and Singapore: the role of diagnosis and isolation as a control mechanism. J Theoret Biol 2003; 224: 1-8.
(7) Pearson H, Clarke T, Abbott A, Knight J, & Cyranoski D. SARS: what have we learned? Nature 2003; 424(6945):121-6. Nature 424, 121: 126(2003) (10 July 2003).
(9) Hsieh YH,. & Chen CWS. Severe Acute Respiratory Syndrome: Numbers don¡¦t tell the whole story. British Medical J 2003; 326: 1395-1396.
(10) Donnelly C, Ghani AC, Leung GM, Hedley AJ, Fraser c, Riley S, et al. Epidemiological determinants of spread of causal agent of severe acute respiratory syndrome in Hong Kong. Lancet 2003; 361(9371): 1761-66. (May 24 2003) Available at http://image.thelancet.com/extras/-3art4453web.pdf.
The focus group research by Thomson and colleagues is, as they
themselves suggest, not strictly Health Impact Assessment (HIA). Whilst it
gives valuable insights into the perceived health impacts on a local
community of closing an existing leisure facility, this was not a joint
exercise with service providers with a view to lessening any negative
impacts.
The focus group research by Thomson and colleagues is, as they
themselves suggest, not strictly Health Impact Assessment (HIA). Whilst it
gives valuable insights into the perceived health impacts on a local
community of closing an existing leisure facility, this was not a joint
exercise with service providers with a view to lessening any negative
impacts.
To achieve optimum benefits for local residents, HIA should be
prospective and collaborative. Otherwise it runs the risk of becoming a
post hoc evaluation with little benefit for the participants. In this case
a collaborative prospective assessment could have predicted some of the
less obvious negative impacts of closure, enabling residents and planners
to take avoiding action.
We have carried out two collaborative HIAs, one on a proposed new
road and one on land remediation options and both of these, together with
our 'toolkit' appear on the Health Development Agency's website
(www.hiagateway.org.uk). Local people contributed to both assessments,
together with planners and public health specialists, before action was
initiated. If HIA is to be an effective instrument for improving health
and reducing health inequalities, it needs to be both prospective and
collaborative.
Dr Bernard CK Choi and Dr Anita WP Pak recently developed a
simple approximate mathematical model to predict the cumulative incidence
and death.[1] Although it’s certainly easy to understand and to use as
they stated, every users must be cautious about misunderstanding the real
applications and evaluations for SARS epidemics. This problem originates
in their too rough assumptions.
Dr Bernard CK Choi and Dr Anita WP Pak recently developed a
simple approximate mathematical model to predict the cumulative incidence
and death.[1] Although it’s certainly easy to understand and to use as
they stated, every users must be cautious about misunderstanding the real
applications and evaluations for SARS epidemics. This problem originates
in their too rough assumptions.
Firstly, since the model is based on simplification of an epidemic
model, so called Kermack & McKendrick model,[2] the fate of an
epidemic depends on threshold theorem: that is, R0 > 1 or not. They
used a single value of the basic reproductive number, R0, throughout the
epidemic. This, however, is not an accurate description of actual
transmission dynamics because R0 is likely to decrease after the onset of
an epidemic is detected and announced. Considering threshold host
densities in case of simple circumstances,[3] R0 would be written as R0 =
N/NT. Here N is the population size, and 1/NT is the proportionality
constant whose value would be determined by all manner of biological,
social, and environmental aspects of transmission. In the real SARS
epidemic, the latter value might be largely affected by social reactions
as well as environmental factors (including weather conditions as the
authors mentioned). It is well known that the potentials for panic and
social stigma are often much greater than the risk for the disease such as
SARS. It is also notable that great variability in R0 should be a key to
consider SARS epidemic. To say that a crude picture of SARS can be
estimated with the model is quite different from saying that the model can
be used to evaluate the success of interventions.
Secondly, as authors stated ‘may not’, homogenous mixing would not be
a correct depiction of actual population interactions of SARS
transmission. Although we are still in the face of so many unknowns
including super-spreading events (SSEs), the small number of transmissions
in most of the countries that experienced SARS occurrences suggests that
the close contact required for transmission did not occur.[4] For
instance, Japan has so far not experienced domestic transmission even
though it had experienced the entrance of an SARS-CoV infected person,
whereas one index case, not SSEs, caused an epidemic originated from
Hospital in Toronto.[5] When we note that 76% of the infections in
Singapore were acquired in a health-care facility,[6] it is evident SARS
can easily be spread by close personal contact. Those facts were
relatively known at that time when authors developed their model.
Therefore, it is too optimistic to apply their assumption to the real data
that every infected person will pass the disease to exactly R0 susceptible
individuals.
I agree that user-friendly mathematical models must be developed and
the mathematical models should be kept as simple as possible so that
public health officers as well as students could predict and learn more
effectively. Such models and recommendations for general usage, however,
may send the wrong message to the local health officers and/or public. The
models for the public must be based on clear knowledge, especially of
epidemiological determinants and ecology of pathogens, when it comes to
general usage. In order to avoid misleading, the mathematical
epidemiologists cannot be too careful to describe their limitations of own
models.
References
(1) Choi BCK, & Pak AWP. A simple approximate mathematical model to
predict the number of SARS cases and deaths. J Epidem Com Health. 2003; In
Press.
(2) Kermack WO, & McKendrick AG. Contributions to the mathematical
theory of epidemics – 1927. R Stat Soc J 1927;115:700-721. (Reprinted
in Bull Math Biol 1991;53:33-55).
(3) Anderson RM, & May RM. Infectious Diseases of Humans: Dynamics
and Control. Oxford: Oxford University Press, 1992.
(4) Arita I, Kojima K, & Nakane M. Transmission of Severe Acute
Respiratory Syndrome. Emerg Infect Dis 2003; 9: 1183-4.
(5) Dwosh HA, Hong HHL, Austgarden D, Herman S, & Schabas R.
Identification and containment of an outbreak of SARS in a community
hospital. CMAJ 2003; 168: 1415-20.
(6) Center for Disease Control and Prevention. Severe Acute Respiratory
Syndrome-Singapore, 2003. MMWR 2003; 52: 405-11.
NOTE:Conflict of Interests including financial interests: Nil
The report by Jousilahti and colleagues in the 2003 September's issue of JECH adds to growing evidence of a consistent association between serum inflammatory
markers – particularly fibrinogen – and social position. [1-3]
These
authors interpret their data as suggesting that the fibrinogen-social
position link is not merely a reflection of the social patterning of
prevalent disease, smoking and obesi...
The report by Jousilahti and colleagues in the 2003 September's issue of JECH adds to growing evidence of a consistent association between serum inflammatory
markers – particularly fibrinogen – and social position. [1-3]
These
authors interpret their data as suggesting that the fibrinogen-social
position link is not merely a reflection of the social patterning of
prevalent disease, smoking and obesity (all of which are positively
associated with increased serum fibrinogen and lower social position)
since a strong trend of increasing fibrinogen with decreasing social
status survived statistical adjustment for these covariates. Fibrinogen,
they conclude, is therefore a promising candidate for the “missing link”
between social position and cardiovascular health.
The authors’ reasoning implicitly accepts that fibrinogen is a cause
of coronary heart disease (CHD). However this runs contrary to recently
published evidence utilising the principle of “Mendelian randomisation”
(the situation where a particular genetic polymorphism strongly influences
exposure level of a putative disease risk factor, and should in turn be
related to increased risk of disease if the risk factor is indeed a
cause).[4,5] In fact we discussed this evidence in our commentary on
psychosocial explanations of health inequalities in last month’s edition
of the JECH (fibrinogen has been shown to be “independently” associated
with some psychosocial exposures leading to claims that it may mediate
relations between these exposures and health).[6] Plasma fibrinogen
levels are related to a polymorphism in the beta-fibrinogen gene, with
presence of the “T” allele being associated with higher levels. Amongst
controls of a recent large case-control study, fibrinogen increased by
0.12g/l per T allele present. Comparing cases with controls, a 0.12g/l
rise in fibrinogen was associated with a relative risk of CHD of 1.20 (95%
CI 1.13-1.26). If increased fibrinogen actually caused heart disease then
a similar per allele relative risk of CHD should be seen. In fact the per-
allele relative risk of CHD was 1.03 (0.96-1.10). Individuals whose
genotype would have subjected them to chronically elevated plasma
fibrinogen experienced no increased risk of heart disease, suggesting that
observed associations between fibrinogen and CHD risk are not causal. This
finding is in keeping with evidence from RCTs that suggests that drugs
lowering fibrinogen do not decrease the risk of CHD.[7]
Fibrinogen probably predicts cardiovascular events because of reverse
causation (atherosclerosis is an inflammatory condition and raises
circulating fibrinogen levels) and because of confounding – smoking,
abstaining from alcohol, not exercising and being poor are all associated
with elevated fibrinogen and themselves increase the risk - or are markers
for factors that increase the risk - of cardiovascular disease
The data presented by Jousilahti and colleagues illustrate how easily
one can misattribute causality to associations in social epidemiology.
Many factors are socially patterned and thus appear as possible candidates
for a causal role in the processes that generate any disease outcome that
is also socially patterned.[6,8] Demonstrating apparent statistical
independence of associations between such exposures and outcomes does
little to infer their causal basis as it is often likely to reflect issues
of residual confounding and measurement imprecision of correlated
covariates.[9] Strategies such as Mendelian randomisation can help
untangle these issues. Where feasible, they should be more widely adopted
in epidemiology.
References
(1) Jousilahti, P Salomaa, V Rasi, V Vahtera, E, T Palosuo T.
Association of markers of systemic inflammation, C reactive protein, serum
amyloid A, and fibrinogen, with socioeconomic status J Epidemiol Community
Health 2003;57:730-733.
(2) Wilson TW, Kaplan GA, Kauhanen J, et al. Association between
plasma fibrinogen concentration and five socioeconomic indices in the
Kuopio Ischemic Heart Disease Risk Factor Study. Am J Epidemiol
1993;137:292–300.
(3) Brunner E, Davey Smith G, Marmot M, Canner R, Beksinska M, O’Brien
J. Childhood social circumstances and psychosocial and behavioural
factors as determinants of plasma fibrinogen. Lancet 1996;347:1008-13.
(4) Youngman LD, Keavney BD, Palmer A et al. Plasma fibrinogen and
fibrinogen genotypes in 4685 cases of myocardial infarction and in 6002
controls: test of causality by “Mendelian randomisation”. Circulation
2000;102(suppl II):31-32.
(5) Davey Smith G, Ebrahim S. “Mendelian randomisation”: can genetic
epidemiology contribute to understanding environmental determinants of
disease? Int J Epidemiol 2003;32:1-22
(6) Macleod J, Davey Smith G. Psychosocial factors and public health:
a suitable case for treatment? J Epidemiol Community Health 2003; 57: 565-
570.
(7) Meade T, Zuhrie R, Cook C, Cooper J. Bezafibrate in men with lower
extremity arterial disease: randomised controlled trial. BMJ 2002;325:1139
-41
(8) Macleod J, Davey Smith G, Heslop P et al. Are the effects of
psychosocial exposures attributable to confounding? Evidence from a
prospective observational study on psychological stress and mortality. J
Epidemiol Community Health 2001;55:878-84.
(9) Phillips AN, Davey Smith G. Bias in relative odds estimation owing
to imprecise measurement of correlated exposures. Statistics in Medicine
1992;11:953-961.
“A leader is best when people barely know he exists, not so good when
people obey and acclaim him, worse when they despise him....But of a good
leader who talks little when his work is done, his aim fulfilled, they
will say, "We did it ourselves.”
Clearly John Ashton’s aphorism mirrors Lao Tzu’s thoughts on
leadership, and is thus hard to argue against. However I think public
health practitioner...
“A leader is best when people barely know he exists, not so good when
people obey and acclaim him, worse when they despise him....But of a good
leader who talks little when his work is done, his aim fulfilled, they
will say, "We did it ourselves.”
Clearly John Ashton’s aphorism mirrors Lao Tzu’s thoughts on
leadership, and is thus hard to argue against. However I think public
health practitioners should indeed be “bothered… if we don’t get the
credit for our own ideas.” It precisely this self effacing stance that has
led to the current situation where public health is grossly undervalued
and under-resourced. It is hard enough trying to promote a negative – the
disease outbreak that didn’t happen – we only make it worse for ourselves
if we allow others to claim responsibility for all the visible successes.
Why should public health practitioners be employed if the successes are
due to the ‘exertions’ of politicians and generic bureaucrats? It is time
that we assembled the evidence of our many successes and market public
health and its practitioners aggressively. We need to attract the
attention of those who procure sickness services at ever increasing
expense and show them the folly of neglecting the protectors, promoters
and the preventers. Time to dispense with the bushel and let our lights
shine!
I have read with high interest the comments made by Petronis and
Anthony on my editorial.[1] I have also read their forthcoming article,[2] and
I believe they apply an analytical approach that seems to be, in my
opinion, a step in the right direction for research on contextual
influences and health that focus on investigation of clustering. I will be
very pleased of writing a larger comment and send i...
I have read with high interest the comments made by Petronis and
Anthony on my editorial.[1] I have also read their forthcoming article,[2] and
I believe they apply an analytical approach that seems to be, in my
opinion, a step in the right direction for research on contextual
influences and health that focus on investigation of clustering. I will be
very pleased of writing a larger comment and send it for consideration and
possible publication in the Journal.
Measuring clustering with the aim of obtaining substantive scientific
information is so far an uncommon approach in social epidemiology and most
multilevel analyses have in fact been plain “contextual analysis”[3] focused
on measures of association. This seems be true not only for studies using
GGE techniques but also for studies using multilevel hierarchical
regression. However, the original standpoint of multilevel hierarchical
regression analyses is the investigation of complex patterns of variation [4]
rather than dealing with residual correlation.
Regarding the use of “GEE2” and measurement of clustering, the
comment of Petronis and Anthony is certainly right.
From an epidemiological point of view the most interesting question
is the conceptual rather than the mathematical approach used. I agree with
Petronis and Anthony in their conceptual approach and I believe that they
put context back in epidemiology using “GEE2” techniques.[5] The Parwise
Odds Ratio and other techniques for measuring neighbourhood heterogeneity
and clustering like the Median Odds Ratio and the Interval Odds Ratio [6,7]
deserve more development and spreading.
References
(1) Merlo J. Multilevel analytical approaches in social epidemiology:
measures of health variation compared with traditional measures of
association. J Epidemiol Community Health 2003;57:550-2.
(2) Petronis KR,.Anthony JC. A different kind of contextual effect:
geographic clustering of cocaine incidence in the US. J Epidemiol
Community Health 2003;in press.
(3) Diez Roux AV. A glossary for multilevel analysis. J Epidemiol
Community Health 2002;56:588-94.
(4) Goldstein H. Multilevel Statistical Models. London: Hodder Arnold, 2003.
(5) Carey V, Zeger SL, Diggle P. Modelling multivariate binary data
with alternating logistic regressions. Biometrika 1993;80:517-26.
(6) Larsen K, Petersen JH, Budtz-Jorgensen E, Endahl L. Interpreting
parameters in the logistic regression model with random effects.
Biometrics 2000;56:909-14.
(7)Larsen K,.Merlo J. Appropriate assessment of neighborhood effects
on individual health -integrating random and fixed effects in multilevel
logistic regression. Working paper 2003.
This is indeed a strange disease. The epidemiology suggests it to be
of relatively low infectivity, but high severity.This in itself is odd,
especially if the causative agent is a virus and the principal mode of
spread by coughing/droplet.Also odd is the undoubted existence of
"superspreaders", who can infect very many of their contacts - I can't
think of any parallels to this in respiratory virology....
This is indeed a strange disease. The epidemiology suggests it to be
of relatively low infectivity, but high severity.This in itself is odd,
especially if the causative agent is a virus and the principal mode of
spread by coughing/droplet.Also odd is the undoubted existence of
"superspreaders", who can infect very many of their contacts - I can't
think of any parallels to this in respiratory virology.
Perhaps the SARS virus obeys the usual rules of droplet-transmitted
respiratory infections, and is of high infectivity. However, due to
shared antigens, a proportion of the population has an acquired resistance to the new virus, having already been exposed to another, relatively
innocuous, virus that provides immune protection. It is possible that the
proportion of humanity immune or partially immune to SARS could be as high as, say, 95% if the second virus were a very common one, e.g. one of the
coronaviruses that causes coryza.This would explain the seemingly low,
unexpectedly so, infectivity of the SARS agent.
Maybe this also explains "superspreaders". Picture humanity divided
into two categories:
1). Those who have met a common related coronavirus,
and consequently have a degree of immunity to SARS, say for the sake of
argument 95% of the population.
2). Those who have not met it, and have no immunity, 5%. If the defences of the first group are overwhelmed by
exposure to a huge SARS virus inoculum, perhaps they would contract a
modified form of the disease, quickly recruit their immune systems to
produce antibodies to a recognised infectious agent, be likely to recover,
not shed large amounts of virus, not be all that infectious.The second
group would get the disease in an exuberant form, excrete quantities of
infectious material, be likely to succumb before their immune system could
meet the challenge.....the superspreaders.
I believe that a coherent model of the SARS epidemic could be
constructed
from the above theory. This of course would not necessarily lend it
validity,
but it may be worth looking at.
We appreciate the comments from Cope et al on our paper reporting the
association between smoking cessation and smoking reduction and subsequent
risk of myocardial infarction (1). Specifically, Cope et al propose that
the lack of a beneficial effect of reduced smoking - in contrast to
smoking cessation - could be due to inaccuracy (underreporting) of the
self-reported tobacco consumption. In addition, Cope et al raise the...
We appreciate the comments from Cope et al on our paper reporting the
association between smoking cessation and smoking reduction and subsequent
risk of myocardial infarction (1). Specifically, Cope et al propose that
the lack of a beneficial effect of reduced smoking - in contrast to
smoking cessation - could be due to inaccuracy (underreporting) of the
self-reported tobacco consumption. In addition, Cope et al raise the
important question of which measurement method most accurately reflects
tobacco exposure in the individual smoker.
We agree that nowadays almost every study of smoking habits apply one
or more measurement of biochemical marker of smoking in addition to self-
report. It is also correct that in our paper the study participants are
divided into the different smoking categories on basis of self-reported
smoking and changes in smoking. However, as mentioned in the discussion,
measurements of expired carbon monoxide (CO) and serum cotinine were
undertaken in one of the follow-up examinations. We found increasing
levels of CO (Table 2) and cotinine (not shown) with increasing self-
reported tobacco consumption, indicating that underreporting of smoking
alone cannot explain our results, but clearly misclassification cannot be
ruled out in this observational study. Furthermore, a previous review and
meta-analysis (2) concluded that self-reported smoking is an accurate
measure of tobacco exposure in population based studies, whereas this is
not the case in intervention and clinical studies. Our data were based on
a sample of the general population; participants with known coronary heart
disease prior to study entrance were excluded. In addition, information on
smoking habits and changes in smoking were part of a large questionnaire
initiated in the late 1970's and the 1980's, thus minimizing the risk of
"social desirability bias" in this study.
Cope et al describe a pilot study using a urine cotinine test for
measuring nicotine intake. There are various methods of validating tobacco
intake including biochemical markers, and cotinine is one of the better
due to its relatively long half-life and the possible linear relationship
with number of cigarettes smoked. However, cotinine is not very useful in
smoking reduction studies since most of the participants in these trials
are supplied with nicotine replacement medications. Interestingly, the
intervention studies of smoking reduction all report that despite nicotine
replacement the percentage decline in amount of tobacco is followed by a
smaller decline in biochemical markers of smoking exposure.
Evidence of the effects of reduced smoking on risk of coronary heart
disease is limited. The few ongoing smoking reduction trials report
favorable changes in blood analyses of parameters of arteriosclerosis up
to two years after smoking reduction. Unfortunately, these studies have a
very high "drop-out" percentage, but it will be interesting to see the
clinical results of a long-term follow-up in this type of "risk reduction"
study.
In summary, we believe that the self-reported smoking habits in our
study are fairly precise. However, biochemical verification of smoking is
necessary in intervention and clinical studies although there are no ideal
markers of tobacco exposure specifically with respect to assessment of
smoking reduction.
(1) Godtfredsen NS et al. Smoking reduction, smoking cessation, and
incidence of fatal and non-fatal myocardial infarction in Denmark 1976-
1998: a pooled cohort study. J Epidemiol Community Health 2003;57:412-416.
(2)Patrick DL et al. The validity of self-reported smoking: a review
and meta-analysis. Am J Public Health 1994;84:1086-1093.
The recent editorial entitled "Multilevel analytical approaches in
social epidemiology: measures of health variation compared with
traditional measures of association" [1] offers an interesting critique of
the generalized estimating equations (GEE) analysis of a paper published
in the same issue of JECH (August 2003). In the editorial, the author
notes that the paper's GEE analysis treats "the intr...
The recent editorial entitled "Multilevel analytical approaches in
social epidemiology: measures of health variation compared with
traditional measures of association" [1] offers an interesting critique of
the generalized estimating equations (GEE) analysis of a paper published
in the same issue of JECH (August 2003). In the editorial, the author
notes that the paper's GEE analysis treats "the intra-neighbourhood
correlation as a 'nuisance' that needs to be adjusted in the analysis but
not explicity investigated" (p. 550).
The editorial then becomes a call for an alternative, more innovative
approach in social epidemiology: "Estimation of the extent to which
individuals within a given neighbourhood are correlated with one another
in relation to health (that is, the concept of intra-neighbourhood
correlation) has value in the context of ideas about the efficacy of
focusing intervention on places instead of people" (p.551). The finale is
a logical conclusion that studies of intra-neighbourhood correlation may
"...present themselves as a new epidemiological approach that may prove
very useful in social epidemiology" (p. 551).
The author is appartently speaking of first-order GEEs but the JECH
readership may not appreciate that second order GEEs (GEE2) treat the
intra-neighbourhood and inter-neighbourhood correlations into deliberate
objects of study and estimation [2]. Although we ourselves deserve
absolutely no credit for biostatistical innovations, the "alternating
logistic regressions" (ALR) approach [3] we employ in our forthcoming
article in JECH [4] is a computationally efficient alternative to GEE2 in
the case of a binary outcome. As such, it estimates the pairwise odds
ratio (PWOR), which quantifies the degree to which health conditions,
behaviors, or perceptions might cluster within neighbourhoods (or other
nested structures of community life) to a degree other than one might
expect if these health conditions, behaviors, or perceptions were
distributed at random across neighbourhoods.
Because we believe our work is responsive to the author's call for a
new approach in social epidemiology that measures intra-neighbourhood
correlation, we would welcome an editorial comment on the potential value
(and possible shortcomings) of the GEE/ALR approach we used in our
forthcoming article in JECH on clustering of cocaine incidence in the
United States. We hope you will
concur that our application of the ALR approach is a step in the right
direction for research on contextual influences and health.
References
(1) Merlo J. Multilevel analytical approaches in social epidemiology:
measures of health variation compared with traditional measures of
association. JECH 2003;57:550-552.
(2) Liang K-Y, Zeger SL. Regression analysis for correlated data.
Annu. Rev. Pub. Health;14:43-68.
(3) Carey V, Zeger SL, Diggle P. Modelling multivariate binary data
with alternating logistic regressions. Biometrika;80:517-26.
(4) Petronis KR, Anthony JC. A different kind of contextual effect:
geographic clustering of cocaine incidence in the U.S. JECH, in press.
Dear Editor
I heard this study reported on the Today programme on Radio 4. What effective dissemination! I applaud the inclusion of SES as a covariate in this study, and I agree that the correlational effects reported are of interest. Nevertheless, I believe that one important aspect of the interpretation of the study is omitted from the discussion. This is the small effect sizes.
In...
Dear Editor
Dr Nishiura [1] accentuated that caution must be exercised in using mathematical models to ascertain the recent SARS epidemics.
The key issue, as we believe, is to understand the model and its results for what they are and, more importantly, for what they are not. It is especially true with the basic reproductive number R0, or its variant the effective reproductive number at time t Rt, which has bee...
Dear Editor
The focus group research by Thomson and colleagues is, as they themselves suggest, not strictly Health Impact Assessment (HIA). Whilst it gives valuable insights into the perceived health impacts on a local community of closing an existing leisure facility, this was not a joint exercise with service providers with a view to lessening any negative impacts.
To achieve optimum benefits for local re...
Dear Editor
Dr Bernard CK Choi and Dr Anita WP Pak recently developed a simple approximate mathematical model to predict the cumulative incidence and death.[1] Although it’s certainly easy to understand and to use as they stated, every users must be cautious about misunderstanding the real applications and evaluations for SARS epidemics. This problem originates in their too rough assumptions.
Firstly,...
Dear Editor
The report by Jousilahti and colleagues in the 2003 September's issue of JECH adds to growing evidence of a consistent association between serum inflammatory markers – particularly fibrinogen – and social position. [1-3]
These authors interpret their data as suggesting that the fibrinogen-social position link is not merely a reflection of the social patterning of prevalent disease, smoking and obesi...
Dear Editor
“A leader is best when people barely know he exists, not so good when people obey and acclaim him, worse when they despise him....But of a good leader who talks little when his work is done, his aim fulfilled, they will say, "We did it ourselves.”
Clearly John Ashton’s aphorism mirrors Lao Tzu’s thoughts on leadership, and is thus hard to argue against. However I think public health practitioner...
Dear Editor
I have read with high interest the comments made by Petronis and Anthony on my editorial.[1] I have also read their forthcoming article,[2] and I believe they apply an analytical approach that seems to be, in my opinion, a step in the right direction for research on contextual influences and health that focus on investigation of clustering. I will be very pleased of writing a larger comment and send i...
This is indeed a strange disease. The epidemiology suggests it to be of relatively low infectivity, but high severity.This in itself is odd, especially if the causative agent is a virus and the principal mode of spread by coughing/droplet.Also odd is the undoubted existence of "superspreaders", who can infect very many of their contacts - I can't think of any parallels to this in respiratory virology....
We appreciate the comments from Cope et al on our paper reporting the association between smoking cessation and smoking reduction and subsequent risk of myocardial infarction (1). Specifically, Cope et al propose that the lack of a beneficial effect of reduced smoking - in contrast to smoking cessation - could be due to inaccuracy (underreporting) of the self-reported tobacco consumption. In addition, Cope et al raise the...
Dear Editor,
The recent editorial entitled "Multilevel analytical approaches in social epidemiology: measures of health variation compared with traditional measures of association" [1] offers an interesting critique of the generalized estimating equations (GEE) analysis of a paper published in the same issue of JECH (August 2003). In the editorial, the author notes that the paper's GEE analysis treats "the intr...
Pages