Clinical Trials for Daleks

Welcome to the Dalek News of 2022.Show more

Thank you for watching.

We're sorry that you don't have anything better to do.

A statistical trial compares an event or the absence of an event with an exposure to an effect,
or the absence of that exposure. The numbers of the resulting occurrences A,B,C,D for the various
groups can be formed into a 2 by2 square.

Odds ratio, or OR equals Odds of the event in exposed group divided by Odds of the event in the non exposed group.
Odds ratio equals A over B divided by C over D. The values A+B = the number of subjects exposed to the event,

and C+D = the numbers not exposed to the event.

If the cohorts can be found of equal size then A+B = C+D.

It is not usual in a human trial that all of the groups can be followed after any event,
or maintained under the discipline needed to continue an experiment.

A similar value, the Risk Ratio, RR is the relative proportion RR=A/(A+C) compared to C/(B+D).
If Gaussian statistics is assumed, the 95% confidence intervals can be written down.

The parameter of 1.96 comes from encompassing 95% of the Gaussian above or below the mean.
If a different confidence interval was chosen, the parameter 1.96 would change.
For example a 99.5% confidence interval would have a parameter of 2.807.

As the values of A, B, C, D tend toward very large numbers the Confidence Interval limits tend
towards the Odds Ratio. The confidence interval is sometimes written as [Lower – Upper].
The confidence interval indicates that if the Gaussian statistics were applicable and the test was repeated, 95% of the repeats would fall in the confidence interval range.
In a ten year study reference 1, questionnaires were sent out to a large group of medical professionals who lived in England or Wales in 1951.
A total of 59,600 questionnaires were mailed out in October 1951.
The useable replies were 40,637 (68%) with 34,445 male and 6,192 female responders.
Breaking down the results into representative cohorts reduced the reportable numbers further.
Death certificates filed from the Registrar General of the United Kingdom November 1951 to October 1961 were obtained.
A total of 4,597 deaths were reported in the 10 year period, with 157 of the deaths attributed to lung cancer,
with 153 cases confirmed. Smoking history could only be obtained for 136 lung cancer deaths.

The rate ratio for lung cancer death for various cigarette consumption brackets compared to non smokers varied
from 8 to over 30, with the average rate ratio over all groups of over 13.
As an example, a cohort of 1000 smokers might have 133 lung cancer deaths, and a matched cohort,
adjusted for age health etc, of 1000 non smokers may have 3 lung cancer deaths.

It can be noted from the data that the Odds Ratio is large (over 50) and the sample size of 1000 in each cohort still
leads to a reasonable spread in the confidence interval. If the sample size was 10,000 in each group,
the confidence interval might come down to [42.4 – 61.4] or 51+/- 9.5.
The results presented were for lung cancer death and smoking, which is significant, noticeable and not controversial.
Similar results can be got from other data sets in other countries. Statistical significance can be judged by the fact
that the confidence interval does not include the null hypothesis value. In this case the null hypothesis value would
be 1.0, indicating the odds of death were independent of the smoking status. In the case that the 95% confidence
interval is 28.4 to 91.6, the odds ratio is statistically significant.
Predictably then, less settled arguments will have statistics associated with their experiments that are not as
statistically significant. The confidence intervals will encompass the null or negative result, indicating that
he events measured may not be represented by the exposure or non-exposure to an event.

In a study of influenza among vaccinated and unvaccinated people 65 and older in Northern California in reference 2,
3,044,531 person years and 9 years of influenza and vaccination status were analyzed.
115,823 deaths occurred in total out of the population, and 20,484 died during the laboratory defined flu season.
As a crude estimate of vaccine effectiveness, consider the all cause deaths only through the virus period with two
randomly selected groups of 100,000. The total death rate was defined as 683 out of 100,000 during the influenza period,
with 326 vaccinated and 357 unvaccinated.

The case control Vaccine Effectiveness VE, reference 3 can be stated as 100% times (1 minus OR) = 8.7% in this case.
Alternatively, a cohort defined VE can be written as VE = 100% times (1 minus Incidence in vaccinated divided by Incidence in unvaccinated).

In practice the sample size was larger than the 100,000 so the statistical significance
(smaller Confidence interval range) of the study reference 2 was larger than estimated here. In the case of the
influenza vaccination, the 95% confidence interval is significant for this vaccine.
Naturally, initial health status, absolute age and the restriction to merely the
laboratory defined influenza period influence the results. One of the important factors is
the nature of the vaccine and influenza virus, for example the 2015 to 2016 northern hemisphere
vaccine contained a specific strain.

Statistically insignificant effects can be found in the literature, and as an example consider Influenza vaccine and Seizures.
In reference 4, 251,600 children under 18 vaccinated between 1 Jan 1993 to 31 Dec 1999 were divided into two sample groups
for events up to 14 days after vaccination and 15 to 28 before and 15 to 28 after vaccination.
A total of 81 seizures were reported, and the authors found association statistically insignificant.
In reference 5 a study of 45,356 children 6 to 23 months old vaccinated between 1 Jan 1991 and 31 May 2003 found seizures
were not considered significant, reporting an Odds Ratio of 1.36 and a 95% CI [0.63 to 2.97].
In reference 6 a study of 13,383 children 6 to 23 months old reported a hazard ratio of 1.17 with a 95% CI [0.36 to 3.86].
These studies were enough for the committee involved to conclude:

The committee assesses the mechanistic evidence regarding an association between influenza vaccine and seizures as lacking.

Other studies, such as reference 7, using a sample of 226,889 children obtained an hazard ratio of HR 2.01 [1.09 to 4.42] for
the three days after vaccination and a HR of 142.8 and 95% CI [77 to 265] for the day of, which was a strong correlation.
The United Kingdom has several data sources, one of which is given in reference 10.
The footnotes for the table have two cautions. The first is that the case rates should not be used for vaccine
effectiveness against infection, as a properly designed clinical trial is needed for infection assessment.
The second foot note is shown:
Differences exist in the health consciousness, age structure, social interactions and natural immunity.
Conveniently, the death rates are divided by 100,000. If the population is considered a random sample,
in which the confounding items just mentioned are ignored, a clinical trial can be set up.
In this case the 2 by 2 square can be written down, and the data entered, using data from table 10 on
page 38 of the report in reference 10.

Results from reference ten suggest an Odds ratio of 2.658, and if the sample was repeated,
95% of repeated samples would be expected to produce odds ratios between 2.5 and 2.77.

Table10, Pg 38, Ref [10]
Under 18 4 0 3 0 0 1
18-29 14 0 9 0 1 4
30-39 53 0 43 0 1 9
40-49 113 3 60 0 4 46
50-59 244 5 116 1 10 112
60-69 453 5 141 0 9 298
70-79 725 8 169 1 22 525
≥80 1,350 13 241 4 37 1,055

COVID-19 vaccine surveillance report
Week 51, Published: 23 December 2021
Publishing reference: GOV-10820

[1] Mortality in Relation to Smoking: Ten Years' Observations of British Doctors
Doll R, Bradford A, Brit. Med. J., 1964, 1, 1460-1467
[2] Influenza Vaccination and Mortality: Differentiating Vaccine Effects From Bias
Fireman B, et al, Am. J. Epidemiol 2009; 170:650–656
[3] Evaluation of influenza vaccine effectiveness. A guide to the design and interpretation of
observational studies World Health Organization 2017
[4] France EK,et.al, Safety of the trivalent inactivated influenza vaccine among children:
A population-based study. Archives of Pediatrics and Adolescent Medicine. 2004; 158(11):1031-1036
[5] Hambidge SJ,et.al Safety of trivalent inactivated influenza vaccine in children 6 to 23 months old.
Journal of the American Medical Association. 2006;296(16):1990–1997.
[6] Goodman MJ et. al., The safety of trivalent influenza vaccine among healthy children 6 to 24 months of age. Pediatrics. 2006;117(5):e821–e826
[7] Febrile seizures after 2009 influenza A (H1N1) vaccination and infection: a nationwide registry-based study Bakken IJ et al BMC Infectious Diseases (2015) 15:506
[8] NCT004470427 Phase 3 randomized placebo controlled observer blind clinical trial.
[9] Effectiveness of COVID-19 Vaccines in Preventing Hospitalization Among
Adults Aged ≥65 Years — COVID-NET, 13 States, February–April 2021, MMWR,August 13, 2021, Vol. 70, No. 32

[10] COVID-19 vaccine surveillance report Week 51, Published: 23 December 2021
Publishing reference: GOV-10820 https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/1043608/Vaccine_surveillance_report_-_week_51.pdf