Epi in the News

Hello everyone,

This was quite tricky and I had a few ideas. Let me present these to begin with:

1. the null hypothesis is that there is no difference in the mean quantity of methane emitted between cows fed with seaweed compared to those on the standard diet.

2. I think a randomized controlled trial will yield the highest level of evidence. Some cows could be randomized to seaweed feeding and another group to the normal feeds and comparisons made in the average amount of methane excreted at different at time points (for example at start, 1, 2, 3, 4, 5 and 6 months) in both groups.

3. The exposure here is seaweed feeding which could be measured as a binary variable (fed with seaweed or not)
The outcome would be measured as a continuous quantitative variable.

4. I would use the 2-sample t-test for independent samples

I look forward to reading your comments.
Cheers,
Colette

1. The null hypothesis is that there is no association between cattle eating seaweed and cattle methane emissions.

2. A cluster randomised controlled trial would be most appropriate as adding seaweed to the cattle’s diet is an intervention. It would probably be hard to assess cows for confounding variables so randomisation would be helpful to manage this. Cattle usually live in groups on farms, so taking each farm as a cluster would be best, perhaps matching intervention/control clusters for size of farm, location etc.

3. Exposure= “small amount of seaweed” (asparagopsis spp) was added to the cattle’s diet. We aren’t provided with a scale in the article. I presume its a binary variable (seaweed in diet; yes/no).

Outcome= methane output which is a quantitative variable.

4. A t-test as the exposure was binary (did they get seaweed in their diet; yes/no) and outcome is quantitative (methane emissions)

Mags

Haha.. I love the title...

1) The null here would be: " there is no difference between the methane emissions in the burps and flatulence of the cows in the control and intervention groups, i.e the difference between the outcome variable in the control and the intervention group is zero.

2) I agree with Colette and Margaret... a randomized trial would be the best design. I did read the abstract of the published article. The study actually used cattle from the same farm, of the same age and weight, and randomized them to receive the different diet options- this kinda makes sense because having cattle from the same farm would help with any confounders- and increase internal validity... If they are planning a larger more generalizable study, I agree with Margaret- a cluster randomized trial would be quite good.

3) The exposure here is diet mixed with seaweed. Upon reading the guardian article you would think that its a binary variable yes/no- 'yes' being the intervention group and 'no' being the control group. (However, in the actual study they do have 3 groups- control, a low seaweed and a high seaweed group- so basically 1 control and 2 intervention groups). It would then be classified as an ordinal variable- since there is an order to the categories.
The outcome has to be a quantitative variable- the article says the intervention cows belch out 82% less methane- which means they have quantified the amount of methane.

4) Now the thing is I read the original paper and know that the cows were in 3 groups. So, I would be thinking about doing the ANOVA (rather than t-test) since we need to compare means across 3 groups, or else a linear regression model if there is a concern for any confounding.

Anybody interested in reading the original paper, here is the link...
https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0247820

Fathima

Hi everyone:
Thanks for the link to the article Fathima
1. Null hypothesis: There is no difference in percentage of methane emitted between cows fed on standard diet and those fed on seaweed diet after 5 months.

2. I think an RCT will be appropriate because the introduction of the seaweed into the diet is an intervention. Also, to reduce the impact of possible confounders the cows will be similar in other parameters such as breed, weight and age and will be randomly allocated to the various groups. Thanks to the link Fathima gave, it showed that they were from the same farm and the same breed so I believe a cluster RCT will be necessary if a more generalisable study were to be conducted using different breeds, from different farms with differences in other characteristics.

3. The exposure from the article is binary (seaweed included diet vs no seaweed included), however in the study it is categorical as there were three groups, one control and two with different levels of seaweed included in the diet. The outcome which is methane emission is continuous.

4. I'm not too sure about this bit but I'll go with a one way ANOVA because my independent variable has 3 levels while the dependent is a continuous variable.

Thank you Sujit, It sounds a very useful intervention, but grim reading for a vegan. Thanks also Colette, Margaret, Fathima and Afua for your replies which I found helpful.

1. Null hypothesis is that the amount of seaweed in the diet of steers has no effect on their methane production.

2. Randomised controlled trial. Each animal was put inside a special machine for measuring its methane production for 140 minutes at a time. In this study there were 6-7 animals in each of the 3 groups.

3. Ordinal categorical exposure - no, low or high seaweed diet. Continuous quantitative outcome - methane exposure

4. Chi squared test for trend.
Also a one way ANOVA
I don't think they actually used either of these methods in the study itself. They say "Data was analyzed as repeated measures with an autoregressive 1 correlation structure". I didn't know what that meant.

Judith