Acronyms are meant to telegraph meaning. More often, they hide it. Some acronyms are useful—SARS is so instantly recognizable, we may forget what it originally stood for. The use of GWAS (genome-wide association studies) seems to be taking root, and it's probably only a matter of time until we all are saying GEE-wass.
But then there are the million other possibilities. Some acronyms are a problem because they are obscure. Others are a problem because they are well known—but in different ways in different contexts. NHS stands for the National Health Service in Britain, while among cohort epidemiologists, NHS is the Nurses' Health Study. MSM may be “marginal structural models” to methodologists, but in the world of sexually transmitted diseases, this is “men who have sex with men.”
At Epidemiology we have a soft ban on acronyms. We figure that the space required to spell out a phrase is a small cost for making papers more accessible. You might ask how any acronym can achieve the convenience of familiarity if editors do not permit its use. We recognize the pitfall of self-fulfilling prophesy, and we allow some acronyms that are at the border of familiarity, for example HRT and FFQ and JEM. (Quick test: How many of these can you translate?) Each one is instantly recognizable in its specialty area, and yet for new epidemiologists or those in other specialties, these may mean nothing.
For acronyms we allow, we follow the standard editorial practice of defining them at first usage. If first usage happens to be in a part of the manuscript you skip (and don't pretend you don't), the definition may not be much help. Some journals also provide translations of acronyms at the beginning of the paper, or at the end, or in a box somewhere in between. As readers, we find this just one more obstacle. Scientific papers are hard enough to read without having to hunt down the decoder for phrases we actually could understand if they were spelled out. The approach at Epidemiology, when we allow acronyms of borderline familiarity, is to intersperse them with the full phrase so that a casual reader won't have to go far to find the translation.
The most risky are acronyms invented by authors themselves. We recently rescued 2 papers from abbreviating “black smoke” (in one instance) and “basic stratum” (in the other), assuming that the authors did not really want to spend the rest of their papers discussing the properties of BS. This does not even consider possibilities in other languages. What may sound benign in Liverpool could produce snickers in Finland. Consider the fate of The National Association of Local Government Auditors (NALGA) (http://www.governmentauditors.org/). They quietly dropped “National” from their name after discovering that their initials spelled “butt” in Spanish.