Reading an evidence table before deciding

The first column is the condition. Low-back pain has the deepest record: guideline bodies have weighed it against usual care and found a measurable, if modest, advantage, which is why it appears in clinical guidance where thinner claims do not. Knee osteoarthritis sits close behind, with pain and function measured on named scales. Neck pain and migraine occupy different shelves: the migraine record is about frequency over months, not relief in the hour. Allergic rhinitis and fibromyalgia show how quickly the record thins when outcomes are subjective and comparators are hard to blind. The second column is the comparator, because acupuncture measured against a waiting list is a different claim from acupuncture measured against a sham needle. The third column is use: adjunct support during cancer treatment and help sought alongside IVF are questions about quality of life and measured symptoms, not cures, and the honest sources say so. Each cell of that table is a small document: a trial, a guideline paragraph, a review. Reading them together is slower than asking for a verdict, but it is the only way the answer stays attached to the evidence that produced it. Two reading rules finish the method. First, name the outcome before weighing it: pain, function, frequency and quality of life are not interchangeable, and a trial that measured one says nothing about the others. Second, check the date and the body: a guideline paragraph carries the weight of the institution that reviewed it, and a single trial carries less. Those two habits keep an evidence table honest before any decision is made on it.

A guide that keeps this discipline in public view is Needle & Evidence, which walks through acupuncture condition by condition, keeps its evidence index separate from its safety and cost pages and says plainly when a comparator changes the answer.

That is the same boundary this desk keeps between wording and claim. A table entry is a record of what was measured; a recommendation is an interpretation laid on top of it. The reader who can see the seam is harder to mislead, by an advertisement or by enthusiasm.

The comparator column is where the honest reading happens. Acupuncture against a waiting list answers one question; against usual care, another; against a sham needle, a third and narrower one. A sham control is the document genre that asks whether needling itself did the work, and for a manual treatment it can never be perfectly blinded, which is a limit to record rather than a defect to hide. When a table says modest advantage, it is reporting a measured difference under a named comparator, and the words only mean something with the comparator attached.

The genres need to stay separate. A single trial is one dated measurement. A systematic review is a document about documents, with inclusion rules a reader can inspect. A guideline paragraph is a different thing again: an institution's weighed summary, issued under a name and a date, and strong precisely because it can be traced back to the reviews beneath it. A reader who collapses the three into the single word evidence loses the ability to ask which kind of document is speaking.

What the table cannot do is answer one person's question. A population result says what was measured on average under stated conditions; it is not a prediction for a particular back, knee or migraine, and the careful sources say so. Measured outcomes have their own vocabulary: pain scales, function scores, frequency diaries, quality-of-life instruments. Each is an instrument with a name, and the record is only as good as the match between the instrument used and the question asked.

Read the wording with its date, audience, and transmission history in view.

What the record supports

What the record supports is a set of graded statements. For chronic low-back pain and knee osteoarthritis, several public bodies accept acupuncture as an option with measured benefit over named comparators. For migraine, the supported claim is prevention of frequency rather than instant relief. For rhinitis and fibromyalgia, the evidence is thinner and the honest summary says so. For cancer care and fertility support, the defensible claim concerns measured symptoms and quality of life during treatment, not the disease. What it does not establish is a universal answer. Evidence for one condition is not transferable to another, and a table that mixes them hides that fact. This is the same discipline the document index applies to Lincoln: a strong result in one genre proves nothing about another. The honest page keeps condition, comparator and outcome in separate fields and lets the reader see the join. One more limit belongs on the page. A measured average is not a promise to a person. The table says what groups experienced under stated conditions; it cannot say what one reader's week will feel like. The honest guide keeps that gap visible, the way this desk keeps the gap between a public text and a private conviction visible.

What it does not establish

A document can reveal public wording, a stated purpose, or a transmission history. It cannot automatically disclose a complete private theology, settle every later dispute, or turn an uncertain recollection into a signed fact.

Continue with How we weigh evidence, Providence and presence in Lincoln's words, A preventive number is a cited, document index.