Challenges in quantifying the clear effect of malnutrition on diarrhoeal deaths.
Challenges in quantifying the clear effect of malnutrition on diarrhoeal deaths.
复制标题
量化营养不良对腹泻死亡的明显影响面临挑战。
DOI：
10.1016/s2214-109x(19)30553-4
复制
发表时间：
2020
期刊：
影响因子：
--
通讯作者：
McQuade,ElizabethTRogawski
中科院分区：
文献类型：
--
作者：
McQuade,ElizabethTRogawski
腹泻和营养不良之间的恶性循环早在几十年前就有描述，1现在仍然是一个难以解决的问题，在资源匮乏的环境中，儿童的双重负担仍然很重。2腹泻和营养不良之间相互作用的一个突出表现是营养不良对儿童死亡率的影响：患腹泻的急性营养不良儿童的死亡风险远远高于营养较好的腹泻儿童。2，3在《柳叶刀全球健康》中，柯克比D Tickell及其同事4通过评估特定病原体与急性营养不良和非急性营养不良儿童的肠道结局之间的关系，为我们理解营养不良与肠道发病率和死亡率之间的相互作用增加了重要的细微差别。营养不良相关易感性的影响可能因病原体而异，这是合理的，而且这项分析确实发现，肠道细菌感染与中度至重度腹泻和死亡之间的关联在营养不良儿童中通常更强，但肠道病毒感染之间的关联在营养不良儿童中通常较弱。文章还强调了一个经常被误解和考虑不足的问题的重要性，这个问题围绕着加法和乘法相互作用之间的差异，或者同样，绝对（风险差异）和相对（风险比或比值比）尺度上的相互作用。对群体一级相互作用的评估依赖于统计模型，并取决于所用模型的规模。非直观地，加性相互作用的证据可以定性地不同于乘法相互作用。由于加性相互作用直接估计与相互作用因素相关的观察到的病例数量的过多或减少，因此该量表与改善公共卫生更相关。5，6然而，由于乘法模型（例如，逻辑回归）是常用的，当在模型中包含交互作用项时，乘法尺度通常是默认选择。尽管长期以来一直有方法用乘法模型和病例对照研究来评估加法相互作用，但仍然存在一种只关注乘法相互作用的不幸趋势。[5，7，8]在Tickell及其同事的研究中，4选择加性交互作用还是乘性交互作用对结果的解释有重要影响。尽管对中度至重度腹泻相互作用程度的估计并不精确，但对于超过一半（17种中的9种）评估的病原体，两个量表之间明显相互作用的方向不同。例如，在乘法量表上，志贺氏菌与营养不良儿童中的中度至重度腹泻的关联较弱，而在加法量表上，志贺氏菌与营养不良儿童中的中度至重度腹泻的关联较强。这种不一致性的最突出的例子是隐孢子虫属和营养不良之间的相互作用与死亡率的关系，其中加性相互作用表明强大的公共卫生协同作用，但乘法相互作用表明拮抗作用。为了更清楚地显示尺度之间的这种冲突，表中显示了隐孢子虫属物种和死亡率的乘法和加法相互作用的粗略分析。尽管营养不良组中这种感染与死亡之间的估计风险比小于营养较好组，
The vicious cycle between diarrhoea and malnutrition was described decades ago1 and remains a refractory problem with continued high dual burden among children in low-resource settings. 2 An acute manifestation of the interaction between diarrhoea and malnutrition is the effect of malnutrition on diarrhoeal mortality: the risk of death is far higher among acutely malnourished children with diarrhoea compared with better nourished children with diarrhoea. 2, 3 In The Lancet Global Health, Kirkby D Tickell and colleagues4 add important nuance to our understanding of the interactions between malnutrition and diarrhoeal morbidity and mortality by assessing the relationships between specific pathogens and diarrhoeal outcomes in children with and without acute malnutrition. It is plausible that the effects of malnutrition-related susceptibilities could differ by pathogen, and this analysis indeed found that the associations between infections with enteric bacteria and moderate-tosevere diarrhoea and death were generally stronger among malnourished children, but the associations for infections with enteric viruses were generally weaker among malnourished children. The Article also highlights the importance of an often misunderstood and under-considered issue surrounding the difference between additive and multiplicative interaction, or equally, interaction on the absolute (risk difference) and relative (risk ratio or odds ratio) scales. Assessments of interaction at the population level rely on statistical models and are dependent on the scale of the model used. Non-intuitively, evidence for additive interaction can qualitatively differ from that for multiplicative interaction. Because additive interaction directly estimates excesses or reductions in the number of observed cases related to the interacting factor, this scale is more relevant for improving public health. 5, 6 However, because multiplicative models (eg, logistic regression) are commonly used, the multiplicative scale is often the default choice when including interaction terms in models. There remains an unfortunate tendency to focus only on multiplicative interaction, despite the long-standing availability of methods to assess additive interaction with multiplicative models and in case-control studies. 5, 7, 8 In the study by Tickell and colleagues, 4 the choice of additive versus multiplicative interaction has an important effect on the interpretation of results. Although estimates of the magnitude of interaction on moderate-to-severe diarrhoea were imprecise, the direction of apparent interaction differed between the two scales for more than half (nine of 17) of the pathogens assessed. For example, Shigella spp had a weaker association with moderate-to-severe diarrhoea among malnourished children on the multiplicative scale and a stronger association with moderateto-severe diarrhoea among malnourished children on the additive scale. The most striking example of this incongruence was the interaction between Cryptosporidium spp and malnutrition with respect to its association with mortality, in which additive interaction suggested strong public health synergy but multiplicative interaction suggested antagonism. To show this conflict between scales more clearly, a crude analysis of multiplicative and additive interaction for Cryptosporidium spp and mortality is shown in the table. Although the estimated risk ratio linking this infection to death was smaller in the malnourished group than in the better nourished group, the estimated