Epi in the News

Haha excellent!! I read this news article shared in another forum and was thinking that probably this will be interesting to discuss in Epi in the News!.... and here we are :-))

In case anybody wishes to read the main study manuscript, this is the open access link to the study-
https://academic.oup.com/jnci/advance-article/doi/10.1093/jnci/djac165/6759686

1) These are incidence figures coz women start off without cancer and then develop uterine malignancy over a time period. The denominator is the number of women aged 35-74 entering the study (group 1 would be those who have never used products, and group 2 would be frequent users of products), the numerator would be the number with uterine cancer.

2) RR here= risk in frequent users/risk in never users = 4.05/1.64= 2.47 times higher risk in frequent users

3) Exposure here is use of hair straightening products.... we have never users, less frequent users and frequent users. Such exposures (that is based on personal lifestyle of people) would most likely be self-reported via interviews or questionnaires...

4) The measure of association here is odds ratio ... was that the question?

5) the baseline risk here was 1.64%- so regardless of their usage of hair products, this proportion would have developed uterine cancer. In frequent users that was increased to 4.05%. So, the attributable risk here is 4.05-1.64 = 2.41%- ; attributable risk fraction is 2.41/4.05= 60%-- meaning 60% of cancer in the frequent use group can be attributed to the frequent use of hair products (provided of course that there is a causal association)

5) The mediator mentioned here would be "endocrine-disrupting chemicals"

6) They looked at the association between strata of frequency of product use (strata 1- >4 times in last year, strata 2- <4 times in the last year). They have looked at odds ratios here- in fact the association is noted to be beyond chance only in the >4 times users

7) Considering that its definitely a prospective study with exposure info collected upon entry into the study, the exposure misclassification is going to be non-differential to the outcome (since the outcome has not yet happened). Coming to the misclassification as such- yes its possible because the exposure is self-reported- they have asked for use over past 12 months and this is subject to recall bias- which means there could be some women in never used group who actually might belong to using group... Since this is nondifferential, the association will be biased towards the null...
In addition to this, there are so many products and devices to straighten hair from different brands and having different contents... and this would mean not all women have the same level of exposure. This could probably underestimate the effect of the truly harmful chemicals (as the group would be diluted by products with contents not carcinogenic).

8) Arguments for a causal relationship that the authors could use-
- Biological plausibility- coz the hair products do have potential carcinogenic content so its biologically possible
- Strength of association and biological gradient- both have been demonstrated in this study- there is a strong association noted (beyond chance) and its much more with more frequent use of hair products
- Temporality- being a prospective cohort study, the exposure clearly happens before the outcomes and there is no issue with reverse causality
- Bias and confounding- this study did take into consideration various confounding factors - that included other risk factors for uterine cancer. adjusted analysis retained the strength of the evidence
- Coherence- in the sense these findings are not really surprising as many cosmetic items have been linked to different malignancies previously and these products are known to contain endocrine-disrupting chemicals

I hope this is enough to get everybody thinking...

Fathima

Hello, sorry for late reply. Let me try!!

1. Are these incidence or prevalence figures? Who is in the denominator and numerator for these calculations?
-> Incidence because they followed cohort around 11 years
-> Total a new case of uterine cancer in people who used strengtheners frequently over 11 years (numerator) / Total population who used strengtheners frequently over 11 years (denominator)

2. Calculate a relative risk figure.
-> 4.05% / 1.64% = 2.47
-> Relative to people who used hair strengthener frequently, people who used hair strengthener frequently had 2.47 times the incidence of uterine cancer over the follow up period.

3. What are the exposure categories implied in the article? How do you think the researchers measured exposure?
-> Frequent hair strengthener usage
-> Measured by survey, maybe, with questionnaire. It can measure as times/duration of time or rated with frequent and never.

4. What specific type of relative risk did the researchers calculate?
-> Incidence risk ratio

5. Assuming the researchers correctly estimated the causal effect here, what proportion of uterine cancer experienced among people who use straightening products frequently was attributable to the hair straightening? What proportion of these people would have experienced uterine cancer anyways?
-> 4.05% - 1.64% = 2.41 = attributable
-> Proportion attributable risk % = 2.41/4.05 = 59.51
-> 59.51% of people who used hair strengthener frequently will be attributable to have uterine cancer

I am not returning student but I will try to answer question 7.

7.Do you think the RR results were affected by misclassification of exposure? Would this be differential or non-differential?
-> Outcome is the same because they tend to use nation data registry. So, it depends on misclassification of exposure measurement. There was reporting bias and recall bias which could affect an answer of participants.
-> I would say it is a differential misclassification so RR is unpredictable. It can toward or away.
** I am not sure ><

Harisd

1. Incidence risk figures as it is a prospective cohort study.
Denominator: population at risk at start of study (N = 33,947). Numerator group 1 = number of women developing uterine cancer over 11-year study period who never used hair straightening chemicals. Numerator group 2 = number of women developing uterine cancer over study period who used hair straightening chemicals frequently. [Frequent use defined as application of chemicals > 4 times in previous 12-month period]

2. Relative risk = 4.05/1.64 = 2.47

3. Exposure categories: (1) No use of hair straightening chemical in previous year (2) Use of chemicals <= 4 times in previous year (3) Use of chemicals > 4 times in previous year (frequent use). Exposure most probably self-reported on questionnaire.

4. Incidence risk ratio

5. Attributable risk = risk (exposed) - risk (unexposed) = 4.05% - 1.64% = 2.41%. 1.64% would have developed uterine cancer anyway (background risk).

6. Mediator (causal pathway) = endocrine-disrupting chemicals (putative carcinogens)

7. Effect modification: incidence risk stratification into two groups based on frequency of hair chemical use in previous 1-year period (group 1: <=4 times in previous year; group 2: > 4 times) - each group compared to baseline risk group (never used chemicals).

8. Misclassification of exposure: possible. Non-differential if exposures were documented prior to development of outcome (as we assume would happen in a cohort study). Was the questionnaire administered at enrolment (specifically where they ask about frequency of use of hair chemicals *in the previous 12-month period*??).

8. Arguments for causal relationship as per Bradford-Hill criteria:

Temporality: exposure precedes outcome
Strength of association: incidence risk ratio
Biological gradient: progressive increase in outcome with increasing exposure (compare data in three strata)
Biological plausibility: endocrine disrupting chemicals are known carcinogens
Analogy: endocrine disrupting chemicals associated with higher risks of breast and ovarian cancer

We need to know what exposures were investigated and adjusted for (potential confounders).

“We estimated that 1.64% of women who never used hair straighteners would go on to develop uterine cancer by the age of 70, but for frequent users, that risk goes up to 4.05%,” the study leader, Alexandra White of the US National Institute of Environmental Health Safety (NIEHS), said in a statement.
1. Are these incidence or prevalence figures? Who is in the denominator and numerator for these calculations? These are incidence figures as women who did not have uterine cancer at baseline were enrolled in the study.
Num: women (35-74 yrs) who developed uterine cancer by the age of 70yrs and never used hair straighteners
Denom: women (35-74 yrs) enrolled in the study who never used hair straighteners
2. Calculate a relative risk figure.
RR: 4.05/1.64 = 2.47 (risk in frequent users/risk in never users)
3. What are the exposure categories implied in the article? How do you think the researchers measured exposure?
Exposure categories: Use of hair straightener products Never users, less frequent users (<1 but <4 in pev. Year) and frequent users (>4times in prev. year)
Self-reported via a questionnaire (not mentioned if it was interviewer or self administered)
...the odds of developing uterine cancer were more than two and a half times higher for women who had used straightening products more than four times in the previous year.
4. What specific type of relative risk did the researchers calculate? Incidence risk ration (Dose response)
5. Assuming the researchers correctly estimated the causal effect here, what proportion of uterine cancer experienced among people who use straightening products frequently was attributable to the hair straightening? What proportion of these people would have experienced uterine cancer anyways?
PAR= (4.05-1.65)= 2.4%
=2.4/4.05= 0.59, Hence 59% of cancer in the frequently HS using group can be attributed to the use of hair products.
1.64% of women who never used hair straighteners would go on to develop uterine cancer by the age of 70 years. Hence, Proportion who had cancer even if they did not use HS product was 1.64%
FOR RETURNING STUDENTS
5. What mediator (causal pathway) is mentioned in the article? Via endocrine-disrupting chemicals
6. The researchers assessed effect modification. What does this mean in terms of the RRs they estimated? They explored the role of Race as an effect modifier. Though they did not find a difference by race but as black women use HS products early in life and more frequently, this could be an important bearing for them.
7. Do you think the RR results were affected by misclassification of exposure? Would this be differential or non-differential?
As this is a prospective study with exposure info collected at entry into the study, the exposure misclassification will be non-differential to the outcome. However, there is a possibility of misclassification because the exposure is self-reported which has been asked for the past 12 months, which might involve recall bias in terms of frequency of use or reporting bias for alternate products.
8. How could the researchers argue that they found a causal relationship? accounting for confounders,
The investigators have considered Biological plausibility and analogy- as HS products have potential carcinogenic and role explored by previous studies with other cancer types
The study design being cohort (prospective) in nature confirms temporality as exposure occurred before outcome.
Assuming bias and confounders have been accounted for in design and analysis
And Strength of association and biological gradient- greater risk among those who use HS products frequently with a dose response relationship.

Thank you!

“We estimated that 1.64% of women who never used hair straighteners would go on to develop uterine cancer by the age of 70, but for frequent users, that risk goes up to 4.05%,” the study leader, Alexandra White of the US National Institute of Environmental Health Safety (NIEHS), said in a statement.

1. Are these incidence or prevalence figures? Who is in the denominator and numerator for these calculations?

cumulative incidence risk. number who developed uterine cancer during the 11y of the study/total number in the study
2. Calculate a relative risk figure.

4.05/1.64 = 2.47 cumulative relative risk of devloping uterine cancer in frequent compared to never users of hair straighteners

3. What are the exposure categories implied in the article? How do you think the researchers measured exposure?

frequent use >4x a year
use <=4x a year
never use

...the odds of developing uterine cancer were more than two and a half times higher for women who had used straightening products more than four times in the previous year.

4. What specific type of relative risk did the researchers calculate?

hazard ratios

5. Assuming the researchers correctly estimated the causal effect here, what proportion of uterine cancer experienced among people who use straightening products frequently was attributable to the hair straightening? What proportion of these people would have experienced uterine cancer anyways?

1.64% of never users developed uterine cancer
4.05% of frequent users developed uterine cancer


1.64/4.05 = 40% of frequent users of hair straighteners who developed uterine cancer would have developed uterine cancer anyway.
60% of the uterine cancer in frequent users of hair straighteners was attributable to the hair straighteners

FOR RETURNING STUDENTS

5. What mediator (causal pathway) is mentioned in the article?

endocrine disrupting chemicals

6. The researchers assessed effect modification. What does this mean in terms of the RRs they estimated?

There was no statistically significant difference between the incidence risk ratio in African American/black and non-Hispanic white women in this study. It is likely this test was underpowered as there were only 17 African American and 7 Hispanic women who were frequent users of straighteners who developed uterine cancer. The p value for effect modification was 0.09.



7. Do you think the RR results were affected by misclassification of exposure? Would this be differential or non-differential?

Exposure was measured in the year before baseline and outcome assessed 11 years later. There is an assumption that exposure in the year before baseline is associated with lifetime usage. This is likely non-differential misclassification. Recall bias is also likely to be non-differential.

8. How could the researchers argue that they found a causal relationship?

Bradford Hill criteria
Temporality - seen as this is a cohort study
Biological plausibility - effects of the chemicals in straighteners being carcinogenic making cause more likely
Dose response effect (which is seen)
Other studies having the same findings
Analogy - chemicals can be carcinogenic, although these chemicals would presumably have to be absorbed through the skin in the scalp to have an effect on the ovaries.
Adequate adjustment for confounding by ethnic group:
Only 7% of women in the study were African-American/Black but 60% of all the women who had ever used hair straighteners were African-American/Black. Black women are known to be at higher risk of uterine cancer, than non-Black women (eg CRUK website). It is possible that this is related to hair straighteners, but there may be another cause. It would be interesting to look at uterine cancer rates by ethnicity in countries where hair straighteners are not used. This study gives rates adjusted for ethnicity, and the adjustment increases the strength of the association, but we only see the hazard ratio after adjustment for multiple variables so we do not know if ethnicty is a positive or a negative confounder of the association.

I note that only women who did not have breast cancer and had a sister who did have breast cancer were invited to participate in this study. These women are therefore not representative of the target population as they are likely to be at higher cancer risk than other women.

It is an interesting and important study. I wonder if hair straighteners should carry a warning in the light of this evidence?

Judith