Louis's numerical study of bloodletting

In 1833 alone France imported 42 million leeches for medical use. Much of the demand came from the teaching of François Broussais, who held that fevers were inflammations of organs and treated them by bleeding; a patient suspected of pneumonia might have the chest covered with leeches. Pierre-Charles-Alexandre Louis, who had practised in Russia before returning to Paris in 1820, doubted the theory.

Louis had years of case records and autopsies from the Paris hospital La Charité. His question, as he put it, was whether, other things being equal, patients bled in the first four days of illness recovered more readily than those bled later. He chose 77 patients with the same well-defined pneumonia, all in perfect health before it began, and fixed as closely as he could the day each had fallen ill. Grouped by the timing of the first bleeding, 41 had been bled on days 1 to 4 and 36 on days 5 to 9. Their average ages were similar, 41 and 38.

The early group was ill about three days less on average. But three sevenths of them died, against one fourth of those bled later, a result Louis himself found startling and seemingly absurd. He pointed out that nearly twice as many patients over 50 had been bled on the first day as later, which must have affected the deaths. He concluded that bloodletting had narrow limits of usefulness, though it should not be neglected in severe cases.

The work first appeared as an article in 1828 and was revised into a book in 1835. C. G. Putnam's English translation, with a preface by James Jackson of the Massachusetts General Hospital, was published in Boston in 1836. Writing for the James Lind Library in 2004, Alfredo Morabia noted that however Louis's figures are analysed, the patients bled in the first four days did worse (p = 0.07).