Certain times of year are abstract-deadline-time. A resident or fellow does a super job of analyzing data and drafting an abstract. With fear and trembling, that person sends it out to the group. If the group is doing its job, “track changes” and comments proliferate. It’s easy to come of out this feeling kind of bruised. And maybe to wonder - what makes a great abstract anyhow? Here are some tips.
• First of all start with a really interesting (but answerable) question. What is an answerable question? Well, “What is the natural history of ductal carcinoma in situ (DCIS)?” is a very important question, but one you are not likely to be able to answer in the finite time you have available for research. On the other hand, you might be able to formulate a question that you can answer with time and resources available to you. This often means taking a small part of a large problem. So, for DCIS, you might design a retrospective study of your own institutional experience with a particular diagnostic or therapeutic modality. Or, you might work with an epidemiology colleague to analyze some large national datasets. Unfortunately, the more “answerable” questions are not always the most interesting.
• Implicit in all this is a careful review of current literature, including material presented at big national meetings (which may not yet be published).
• Next, formulate a hypothesis. Don’t simply do a retrospective study and hope that something of interest will emerge. In the DCIS example, you might have noticed that a large percentage of elderly patients from rural areas were having lumpectomy without radiation. You might hypothesize that these patients did (or did not) do as well as those who did have radiation. Either one is a hypothesis which you may be able to test.
• Determine whether your hypothesis is testable. Enlist a biostatistician and do a preliminary power analysis. This is a statistical analysis that tells you if you have enough patients to answer the question. If you do not have enough patients to answer the question (i.e. your study is underpowered) then find something else to study. Work with your biostatistician as you design your data collection and analysis strategy.
• Do the research and analyze the data. Did you prove or disprove your hypothesis? Did something else emerge? List your key findings.
• Follow the instructions. Generally the structure will be: Introduction, Methods, Results, Conclusions. There will be a word limit or a size limit. Tables and graphs may or may not be allowed. It never hurts to look at old program books to see what the abstracts for that meeting “looked like.”
• Decide which findings are best illustrated by numbers in the text, by graphs, or by tables. Summarize some of the least important findings. Which findings are not important? Well, if you are comparing two groups of patients it is important to see how well-matched the groups are for age, gender, and so on. A simple statement such as “the groups were well-matched for x, y, and z” will suffice.
• Make sure that your conclusions are supported by the data. “Data to be provided” is a sure recipe for rejection.
• As soon as you send off the abstract, start working on a manuscript for submission to a journal. That way you benefit from the momentum gained during abstract submission. For more details on manuscript preparation, see my book on Medical Writing or any standard text.
• First of all start with a really interesting (but answerable) question. What is an answerable question? Well, “What is the natural history of ductal carcinoma in situ (DCIS)?” is a very important question, but one you are not likely to be able to answer in the finite time you have available for research. On the other hand, you might be able to formulate a question that you can answer with time and resources available to you. This often means taking a small part of a large problem. So, for DCIS, you might design a retrospective study of your own institutional experience with a particular diagnostic or therapeutic modality. Or, you might work with an epidemiology colleague to analyze some large national datasets. Unfortunately, the more “answerable” questions are not always the most interesting.
• Implicit in all this is a careful review of current literature, including material presented at big national meetings (which may not yet be published).
• Next, formulate a hypothesis. Don’t simply do a retrospective study and hope that something of interest will emerge. In the DCIS example, you might have noticed that a large percentage of elderly patients from rural areas were having lumpectomy without radiation. You might hypothesize that these patients did (or did not) do as well as those who did have radiation. Either one is a hypothesis which you may be able to test.
• Determine whether your hypothesis is testable. Enlist a biostatistician and do a preliminary power analysis. This is a statistical analysis that tells you if you have enough patients to answer the question. If you do not have enough patients to answer the question (i.e. your study is underpowered) then find something else to study. Work with your biostatistician as you design your data collection and analysis strategy.
• Do the research and analyze the data. Did you prove or disprove your hypothesis? Did something else emerge? List your key findings.
• Follow the instructions. Generally the structure will be: Introduction, Methods, Results, Conclusions. There will be a word limit or a size limit. Tables and graphs may or may not be allowed. It never hurts to look at old program books to see what the abstracts for that meeting “looked like.”
• Decide which findings are best illustrated by numbers in the text, by graphs, or by tables. Summarize some of the least important findings. Which findings are not important? Well, if you are comparing two groups of patients it is important to see how well-matched the groups are for age, gender, and so on. A simple statement such as “the groups were well-matched for x, y, and z” will suffice.
• Make sure that your conclusions are supported by the data. “Data to be provided” is a sure recipe for rejection.
• As soon as you send off the abstract, start working on a manuscript for submission to a journal. That way you benefit from the momentum gained during abstract submission. For more details on manuscript preparation, see my book on Medical Writing or any standard text.