Epidemiology-
Apart from questions 12-15, all questions are about this systematic review:
Amegah et al. Household Air Pollution from Solid Fuel Use and Risk of Adverse Pregnancy Outcomes: A Systematic Review and Meta-Analysis of the Empirical Evidence. Plos One 2014
Questions 12-15 are about this article, which is included in the systematic review:
Thompson et al. Impact of Reduced Maternal Exposures to Wood Smoke from an Introduced Chimney Stove on Newborn Birth Weight in Rural Guatemala. Environ Health Perspect2011;119:1489-1494
The questions are about birth weight as a continuous outcomeand as a binary outcome - low birthweight (LBW. The international definition of LBW is birthweight < 2500 grams. The word birthweight is used throughout the assignment and in the systematic view when referring to the continuous outcome, and LBW for the binary outcome.
For the systematic review, you can ignore:
- Information about the Newcastle-Ottawa quality scale
- Information about stillbirth, preterm birth, intrauterine growth retardation and miscarriage
- Publication bias and how the authors attempted to deal with it
- Meta regression
You will not need to read all of the Thompson article. Look at Questions 12-15 to see what sections you will need to read.
Format of your assignment:
- Briefly explain all your answers
- Do not copy the questions or introductions, since this increases the similarity score in Turnitin.
- Format your assignment in portrait mode, not landscape.
- If you directly quote material from either of the papers, please put it in quotes and include the reference
- Use the author year (e.g., Amegah et al, 2014) system of referencing.
Research question-
Item 4 in the PRISMA checklist states: "Provide an explicit statement of questions being addressed with reference to participants, interventions, comparisons, outcomes, and study design (PICOS)."
1) For LBW only, complete the following table, which contains the elements of a PICOS research question. Where the authors of the review have provided sufficient information to fill in the blank, do so. If they haven't provided the information, put "not stated" in the table. Use only the information in the Introduction of the article to answer this question.
2) The authors stated that their manuscript complied with this item on the checklist. Do you agree with their claim?
Search strategy
Item 8 in the PRISMA checklist states: "Present full electronic search strategy for at least one database, including any limits used, such that it could be repeated."
3) Would you be able to replicate the electronic search strategy for any of the databases used by the authors?
Summary effect estimate for low birthweight
For low birthweight (LBW), the authors refer to the pooled estimate from the meta-analysis as a "summary effect estimate (EE)" without indicating what the measure of association actually is. For all except two of the studies included in the review, the measure of association reported was the odds ratio. For the other two studies, it was the risk ratio. Given the design of the studies, you can assume that the odds ratios are estimates of the risk ratios. Thus, the "EE" can be considered to be a risk ratio.
4) Write a sentence reporting the pooled risk ratio for LBW and its 95% CI, making it clear that it is a risk ratio.
Selection bias
At the top of page 14, Amegah et al. state: "Selection bias was generally minimized in all the included studies as the studies were largely representative of their source population and reported high response rates."
5) Potential for selection bias in relation to "response rates"
a) Separately for cohort studies and case-control studies, what information on "response rates" is necessary to help determine whether selection bias is an issue? [Hint: for cohort studies, the initial response rate at baseline is not the issue.]
b) Do the authors provide you with sufficient information to make this judgement for yourself?
6) Potential for selection bias in relation to choice of controls in the case-control studies
a) What information does table 2 provide on the source of cases and controls for each of the studies?
b) What additional information should it contain in order for you to assess whether the control groups were appropriate?
Information bias
In all the observational studies, the exposure was measured by interviewing the study participants, so misclassification of the exposure is likely. These questions address this misclassification only (i.e., ignore any potential misclassification of the outcome).
For questions on information bias, assume that the cross-sectional studies are case-control studies, because they are very similar with respect to misclassification and information bias.
7) For the prospective cohort studies, what direction of bias would you expect due to misclassification of the exposure?
8) For the case-control studies, can you predict what direction of bias you would expect due to misclassification of the exposure?
9) Do you think that the positive association between solid cooking fuels and the two outcomes can be fully explained by information bias? [Make use ofyour previous answers about information bias and the results in Table 6]
Confounding
Consider the outcome birthweight for which the pooled estimate of the mean difference was 86 g higher birthweight for unexposed infants.
The table below shows results for the studies classified by the authors as having adequate control for confounding.
|
|
Mean difference in BW (g)
(Unexposed - exposed)
|
|
Boy 2002
|
63
|
|
Mishra 2004
|
175
|
|
Siddique 2008
|
82
|
|
Tielsch
|
104.5
|
|
Sreeramareddy 2011
|
39.9
|
|
Abusalah 2012
|
186
|
|
Epstein 2013*
|
78.1
|
|
Amegah 2012
|
Charcoal 243
Charcoal and LPG 109
|
* This study was reported as having adequate control, but the differences presented are unadjusted.
10) Do the results from studies the authors considered to have adequate adjustment generally show an association in the same direction and of similar magnitude as the pooled mean difference?
[You don't need to do a Forest plot or perform a meta-analysis.]
11) Does this strengthen or weaken the evidence that exposure to solid cooking fuels decreases birthweight?
The RCT (Thompson et al 2011)
In the lecture on causal inference, the scheme proposed by Weiss and Koepsell for assessing causality was discussed. In this scheme, the results of RCTs take precedence for making causal inferences.
12) Was this trial designed to assess whether exposure to pollution from solid cooking fuel reduced birthweight?
13) Was the analysis of the trial by intention to treat?
14) Comment on the possibility of bias due to "attrition" in this study with respect to measurement of birthweight within 48 hours of birth. ["Attrition" here meansthere are missing data on birthweight]
15) Given your answers to the previous three questions about the trial, do you consider that this trial provides higher quality evidence than the cohort studies?
Consistency of the association
As discussed in the lectures on systematic reviews and causal inference, I2 is a measure of the inconsistency (heterogeneity) between studies. An I2of 0 indicates that all of the variation between studies is due to chance and none is due to inconsistency and an I2of 100% indicates that all the variation between studies is due to inconsistency rather than chance. The larger the I2value, the greater the heterogeneity. The developers of I2stated: "we would tentatively assign adjectives of low, moderate, and high to I2 values of 25%, 50%, and 75%. (Higgins et al, 2003)
16) For each of the outcomes of birthweight and LBW, comment on the consistency of the findings. In your answer, state whether this is supportive of a causal interpretation of the association or otherwise.
17) For LBW, how similar are the pooled estimatesfor the different types of studies? Does this weaken or strengthen the argument for a causal interpretation of the evidence?
Temporality
18) Is there strong evidence that the exposure precedes the outcome (for both LBW and birthweight)?
Strength of the association
19) For interventions designed to increase birthweight, 100 grams is usually chosen as the minimum "clinically significant difference", i.e., the minimum difference of practical consequence to the health and welfare of infants. Given this, would you consider the observed mean difference in birthweight to be a strong association?
Dose-response relationship
20) Does the paper provide evidence on whether there is a dose-response relationship for either birthweight or LBW (i.e., whether the association is stronger for higher levels of exposure)?
Biological plausibility
21) Do the authors attempt to make a case for how these pollutants might affect foetal growth? [We are not expecting you to be experts in foetal growth and the likely impact of these kinds of pollutants - we just want to know whether you think they have outlined an argument that might convince an expert]
Attributable fraction (AF)
22) Calculate the attributable fraction for household use of solid fuel for cooking in relation to low birth weight using the "summary effect estimate" reported by the authors. Show the formula and your working.
23) Write a sentence interpreting the result.
24) What assumptions are necessary for this calculation to be valid?
Global population attributable fraction (PAF)
25) Calculate the global population attributable fraction for household use of solid fuel for cooking in relation to low birth weight. Show the formula and your working.
26) Write a sentence interpreting the result.
27) In addition to the assumptions for the AF to be valid, what additional assumption is necessary for the PAF to be valid?
Evidence to action
28) Based only on this systematic review and your responses to the previous questions, do you think the evidence that exposure to household air pollution from household solid fuel use for cooking reduces birthweight is sufficiently strong that action to reduce exposure is warranted? You should refer to the PAF in your summary. You should ignore other potential adverse effects of the exposure such as on risk of TB and of economic considerations. The word count must be no more than 120 words - please state the word count.
References - Higgins JPT, et al. Measuring inconsistency in meta-analyses. BMJ. 2003; 327: 557-560.
Word limit: 1500 words. Include a word count. For every 150 words (10%) you exceed the word limit.