The word science comes from the Latin word scire, "to know." The practice of modern medicine is scientifically based, although the scientific process is not always smooth or easy.

Knowledge enters into the practice of medicine largely through the published literature. Two anti-intellectual demons, opinion and speculation, may fill the vacuum where scientific knowledge does not exist and may acquire a life of their own if not stopped early.

How can doctors be sure that what they "know" is really the "truth"? Published scientific studies come in many styles, some appropriate for answering certain types of questions and inappropriate for other questions. There are strengths and weaknesses to these study styles.

Anecdotal case reports. These publications typically present between one and 10 patients with an interesting problem which may never have been reported before. Treatment of the problem is usually discussed. Previously unrecognized surgical complications are frequently reported in this way. New information can be presented relatively quickly in the literature, since such reports usually speak for themselves and do not require statistical analysis or a formal study to be set up. This information may be important in treating a condition.

These articles can't necessarily gauge how commonly an event occurs. They may over-emphasize a rare, but spectacular occurrence which causes physicians to become overly cautious in treating a certain condition. Thus, they may be severely biased and misleading. Sometimes an entire field of knowledge can be affected for decades by such reports, which may be literally "out of control."

Biopsy-control. If a certain disease is being studied, it would be important to know that all patients had the same disease.

This can be ensured by biopsy of a lesion to make sure all patients under study had the disease. If the disease is diagnosed by visual criteria only, the accuracy of the visual diagnosis is important. If the disease under study has one appearance which is easily identified in all patients, then no problem may exist. However, if a disease has myriad appearances and several mimics, this may allow inclusion of patients who don't have the disease or exclusion of others who do.

Historical controlled trials (HCT). The effectiveness of a new or established treatment can be compared to another treatment to try to see which is better. Sometimes the treatment under study is compared to previously published treatment outcomes, or historical controls. Such reports depend heavily on the accuracy of the published results since the investigator is accepting the published results as valid. If the published results are in error for some reason, they can't be changed retrospectively, and the error may not even be obvious. Thus, an ineffective treatment may seem effective, or an effective treatment may seem ineffective.

Controlled trial. (CT) A new or established treatment in a group of patients (the study group) is compared to another treatment applied to another group of patients (the control group) being treated at the same time by the same investigators. This allows a similar selection process to be applied to enter patients into both the study group and the control group. This helps to ensure that the groups are similar except for the treatment under study. Thus, differences in outcomes can be ascribed to the differences in treatment rather than some unknown variable which may be operating by chance.

However, as the number of selection criteria increase, the study results may be interpreted as being applicable only to patients with similar findings. CT's are expensive and require lots of organization. They are also time-consuming. By the time results are available, a treatment may be obsolete or irrelevant for some other reason. Trivial or unimportant findings may be overemphasized by the results of a CT.

Randomized controlled trial. (RCT) Similar to a controlled trial, except that similar patients are assigned to study or control groups by a random number table. This is thought to eliminate bias as completely as possible in selection of patients.

For instance, a controlled trial which entered patients into the study group if they were seen in the clinic on Saturday and into the treatment group if seen on a Tuesday might generate large differences in these groups based on religion, employment status, number of children at home, etc. These differences might bias the treatment results and would not occur if patients were assigned at random.

An RCT is not the only way to indicate whether a treatment is effective, but is thought to be the "gold standard" for proof of superiority of one treatment over another. This carries certain risks.

If an undiscovered flaw exists in a study, misleading information is generated which carries much more weight because it came from an RCT. There is an unfortunate tendency to discount findings of other types of studies if they are not RCT's. This allows a destructive intellectual elitism to enter the scientific process: anyone can take a cheap shot at any study that is not an RCT, belittling its results in the process.

Prospective longitudinal cohort study. A group of patients undergoes a treatment and is followed for a certain length of time to determine outcome. This type of study can give very good insight into fundamental strengths or weaknesses of a certain type of treatment. Such a study may be followed by an RCT if the treatment appears promising. This type of study is the basis of the current popularity of outcome analysis, one means by which the effectiveness of treatments is judged.

Operator bias. In surgical trials, not all surgeons may have the same skill level, so the surgical "dose" given to a patient is difficult to control. By comparison, studies of medical therapies involve reproducible doses which can be the same from patient to patient around the world. This seeming precision of medicine lends itself well to statistical number crunching which sometimes empowers the results with importance exceeding true clinical usefulness.

Number of patients. The larger the number of patients studied, the more accurate and reproducible the results might be to other patients.

Adequacy of follow-up. If a large number of patients is studied, but a lot of them are lost to follow-up, the results can be misleading since the outcome of those lost to follow-up might differ from those successfully followed.

Length of follow-up. The longer the period of follow-up, the clearer the picture about long term effectiveness.

Conclusions

Readers of the scientific literature must have some knowledge about the strengths and weaknesses of the study methods employed. There are many pitfalls between conceptualization of a study and its publication in the literature, so a general truth is that readers must interpret the published literature cautiously in light of their own experience and common sense. For patients with endometriosis, a study of the history of scientific publication regarding the disease is revealing. Much of our misunderstanding about endometriosis is based on overemphasis of early anecdotal reports, errors of visual identification, lack of biopsy control, and no follow-up or poor follow-up in small numbers of patients. Against this background, the pursuit of truth sometimes may seem like chasing flies, but progress is occasionally made.