CONFIDENTIAL DRAFT 2-18-2000

Annals of Propecia

Background

In April of 1998 I was contacted by NBC for a possible appearance on a DATELINE segment about

Propecia. Propecia is the new name for the drug finasteride which has been marketed under the name

Proscar as a cure for benign prostatic hyperplasia. Proscar has been prescribed for older men with

urinary problems on the supposition that the cause is an enlarged prostate. Propecia I was told was

designed for younger men who wish to grow hair. Since I had never studied this use of finasteride, I

declined. I knew, of course, that finasteride would promote hair growth and I thought it would be

interesting to follow this up.

The legal approval of a drug for a new use is based on the data submitted by the manufacturer to the

FDA. Not finding the approval request for Propecia on the FDA website, although the data submitted for

approval for drugs approved at a later date were already there, I decided to obtain the Propecia materials

using the Freedom of Information Act. After some months, the fiches appeared and DATELINE was

helpful in printing them out and sending the three pounds of material for me to study.

There I discovered the beautiful data on testosterone to dihydrotestosterone conversion. It is that

conversion that is the physical basis for hair growth. I have good experience in searching the medical

literature. (I had found 45 papers in my web search of finasteride as is described in one of my earlier

papers) 1, so I was puzzled. Therefore, I initiated phone and fax communications with Keith D. Kaufmann

of Merck relating to Propecia. He affirmed that the testosterone data had not been published. When I

asked why it had not been published but had been sent to the FDA, he said that I must surely know how

difficult it is to get papers published these days.

That response convinced me to take time off from my other research efforts to investigate this bizarre

response. I diligently studied the huge amount of data, wrote up my results, and sent a prepublication

copy to the FDA ombudsman for transmission to the appropriate FDA persons responsible for the drug's

approval.

In August of 1998 I submitted the paper ``Study of the Food and Drug Administration Files on Propecia

'' to the Archives of Dermatology. It was given three peer reviews, all positive, re-formatted as a

Commentary, and accepted after minor revisions. It finally appeared about a year after submission, in the

March 1999 issue.

I am not a physician but a physicist, with over a half century of experience in data analysis, so I thought

my examination of the data might be of interest. I was pleased at the plethora of emails supporting my

material after it appeared in print. But I was concerned that the FDA had not done an appropriate

assessment of the most quantitative data in the submission.

I did not criticize Merck but raised questions about the FDA's approval and travelled to Washington to

speak with the Director, Jonathan Wilkin, and the chief physician responsible for the approval, Hon-Sum

Ko.

I was troubled that not a single FDA refereee had commented on the following key pieces of the data:

a) The conversion of T to DHT fell to a low level at .05 milligrams and stayed the same up to 5 milligrams

b) The efficacy had not been established below a dosage of 1 mg.

Further, although side effects at 1 mg were not frequent, effects of finasteride on both semen ejaculate

volume and a large depression of the Prostate Specific Antigen (PSA) score, (a test for prostate cancer),

appeared in the data submitted to the FDA.

I again returned to my other physics interests until a dermatologist friend asked me in Dec. 1999 whether

I had seen three comments from Merck personnel denigrating my commentary in the Archives of

Dermatology.

It is common in most scientific journals for the editors to peer review articles that comment unfavorably on

a peer-reviewed article that they have already published. Also the author of the original article is asked to

publish replies in the same issue. The Archives of Dermatology did not follow this usual procedure

although the same issue carried a criticism of another article and the author's reply. To the credit of the

Archives, my article appeared in an issue that carried a full page advertisement for Propecia.

This present ``Annals'' deals with several issues. First, is a response to the Merck letters that appeared in

the Archives. Most of the comments are so incorrect or lacking in logic that I will simply quote them and

rebut them in turn. Second, I will quote and respond to the FDA Director of Dermatology's comments on

my work. Third, I will touch on the new role of drug companies in ``promoting'' drug use rather than what

used to be called ``filling prescriptions''. Finally I will suggest activities that organisations, such as the

Institute of Medicine of the National Academy of Sciences or Physicians for Social Responsibility, can

initiate to help improve the control of drugs that are now flooding the USA markets.

Replies to the Merck Representatives

A. Keith D. Kaufman Comments ( AOD August 1999 )

1) Kaufman is correct in noting that my conclusions were ``based on serum and scalp DHT

measurements we made on balding men treated with finasteride ...''. Those were the data supplied by

Merck to the FDA and on which the approval was based. The data showed the percent change in

testosterone (T) to dihydrotestosterone (DHT), was 46% going from zero dosage to .05 mg and staying

flat to 5 mg. Any physician who has taken a pre-med course on elementary physics, especially one with a

laboratory component, who examines the error bars on the FDA plots that I published, would come to

the same conclusion.

2) Kaufman disputes the FDA data by referring to a paper (his ref 3) which is ``in press'', so neither I nor

the FDA could possibly have known about it in April of 1998 nor can your readers yet assess it. He

quotes the difference in T to DHT conversion for 1.0 and 0.2 mg samples as 68.5 +/- 1.4% and 61.2 +/-

1.7%, (a difference of 7.3). These are not the data given to the FDA. Nor did I get a preprint of these

results from Kaufman although he knew about my interest as the result of our phone calls.

3) Kaufman further makes claims that the efficacy was better for the 1.0 mg than for the 0.2 mg dosage

quoting his ref. 4, also ``in press''. That is not data on which the FDA acted.

4) The hair count study: The number of persons in the sample were only 101 (1 mg) and 99 (0.2 mg). If

Kaufman bothered to read the material that had been submitted by his company to the FDA, page 23 of

NDA 20-788, he would see that Merck had properly given the data as mean change from baseline: 68.7

+/- 17.3 and 54.9 +/- 17.3. These are the widths of the distributions at the 95% confidence interval. At

the risk of boring the reader, we present in Appendix 1 the standard statistical analysis of the data. Yet

Kaufman suggests in his letter in the Archives ( March 1999 issue) (his references 6 and 7) that a

physicist with over 50 years of publishing data in refereed journals, complete with both ``statistical''' and

``systematic errors'', 2 needs to read a book on statistics.

5) Kaufman argues that I should have known that he claimed to have data that proved that 1.0 mg was

superior to .2 mg in hair growth. To show this he gives his reference3 (3 and 4) which he lists as ``in

press''. There is very remarkable use of the calendar in his remarks, a kind of new time-reversal

invariance concept.

B. Janet L. Roberts Comments ( AOD August 1999 issue)

1) Roberts says ``commentaries are not held to the same rigorous scrutiny and high scientific standard as

peer-reviewed articles" This does injustice to the editors of the AOD since the article was peer-reviewed

by three physicians. The Roberts letter was not peer-reviewed.

2) Roberts tries to sell the superiority of the 1mg dose by referring to her reference (1) That reference is

also ``in press''. I do not claim clairvoyance.

3) Conclusion: Roberts has introduced a free advertisment for Propecia in the Comments Section. My

article never said Propecia did not grow hair.

C. Diane Thiboutot Comments (AOD Nov. 1999)

) Thiboutot says: ``Dr. Frankel insinuates that collaboration occurred between physicians and subjects as

regards questionaires and global assessments.'' Absolutely nothing in my paper does so. The placebo

effect is a remarkable effect connecting mind and body. Although it is routinely used in double blind

studies, the effect is rarely understood in its complexity. It is recommended that interested readers study

the recent book The Placebo Effect edited by Anne Harrington of Harvard Medical School.3

Roberts is apparently not familiar with the huge literature on placebo effects or she would be able to

distinguish patient and physician assessments based on a ``Patient Hair Growth Questionaire'' or a

physician ``Investigator Assessment'' vs. real data based on the counting of hairs from photographs. The

former are qualitative opinion polls and are usually only used when real measurements of hair counts are

not available. This was not the case here. What Roberts has not understood is that one placebo effect

arises because the patient would like to see the treatment grow hair. So would the physician. To deny this

effect for physicians makes an unflattering statement suggesting that physicians, unlike their patients, do

not have the same human responses. I thought it was worth mentioning, from my examination of the data,

that there was nice evidence here for the placebo-like overestimate of the hair growth for both types of

observers. None of the FDA referees commented on this result.

2) Roberts goes into well known conditions on pricing that are irrelevant to the drug dosage approval and

which are well known. I should not have made the remark about solubility in ethanol since the FDA

actually blacked-out, i.e. censored, the actual solubility. I apologize for not telling the dermatologist

readers that there is a trivial way to subdivide the pill: Crush it between two spoons and transfer the very

fine powder to a small clear glass bottle of water. (The containing coating is easy to remove.) An

extremely fine suspension is obtained. Shaking the bottle and removing a tablespoon of liquid is the simple

way to reduce the dose.

I cannot understand why saving money ``cannot be condoned'' when the cost is approximately $1000 per

year, not covered by insurance, and the drug must be taken forever in order to preserve any hair growth.

There are no data proving that the lower dosages are ineffective while the physical T to DHT conversion

measurements are flat from .05 to 5 mg. Considering the cost to take the drug for the rest of one's life and

that the drug is too expensive for most young men, an intelligent user might want to do a few month trial at

lower dosage. Nothing to lose but some fuzz.

Comments on the FDA performance

13) While my article preceded my discussions with the FDA it seems useful to report now on comments I

received in writing from the FDA:

a) The FDA denigrates the data on the T to DHT measurements as ``surrogate'' even though there were

no low dosage measurements on efficacy. That beautiful data cannot be ignored to justify the FDA

dosage approval. Surrogate can only be applied to physical data if the efficacy data are done over the

same dosage range. Otherwise ``surrogate'' has no scientific meaning.

b) ``Dr. Frankel is minimizing this trial by saying the number of participants was only 100. In fact 466

patients were enrolled and 382 completed (92-98 patients completing in each arm).'' Can it be true that

the director of the FDA does not know that the number of patients receiving zero or .01 mg is irrelevant

to the comparison of the patients receiving 1.0 and 0.2 mg. They have nothing to do with the statistics of

the 1.0 and 0.2 mg comparison.

By the way, the Propecia data still today (Feb. 2000), a half year after my written request, is not on the

FDA web.

Perfortmance of The Archives of Dermatology

It is customary in most journals for articles on medical matters, such as my commentary on the FDA

supplied data on Propecia, to be refereed, if the editors reading the article judge it to be of interest to their

readers. The editors play a crucial role, especially since they may reject an article they have read without

sending it to referees. My article on the finasteride data was read by either Kenneth Arndt or Robert

Stern. the AOD editors. It was deemed sufficiently interesting so that it was sent to three referees, all of

whom approved of publication.

Presumably the editors also read the three Merck letters to the editor and should be expected to have

enough competence to judge their scientific merit. They chose not to have any of the letters refereed, nor

did they follow the usual process of informing me of the fact that were going to publish the Merck letters

without them being refereed. Normally the primary author would have the opportunity to respond in the

same editions. Clearly they must have understood the special status of Merck associates and that the

Merck responses may not have been entirely scientific. Most scientific journals allow an author to

respond to his critics.

When I discovered the negative letters related to my study of the FDA data I called and wrote to the

editors to discuss this matter. My letter of Dec. 5, 1999 and subsequent phone calls were unanswered so

on Jan 5, 2000 I again wrote asking that an editor call me. That letter was ignored. So I decided to fedex

an earlier copy of this article to the AOD which did not have this section in it. In February I received a

rejection of that article. I again tried to call Robert Stern. After many attempts Robert Stern called me and

we had an interesting conversation, some of which is repeated below.

Stern responded to my questions about his unusual decision to neither inform me of the Merck letters or

give me a chance to respond by saying that he does not ask for referees of letters. But he had to have

approved of their publication so he knew their content. This happened a second time in the next issue

when the third letter was printed. So apparently he decided that the Merck letters were valid and made

no effort to find out whether they were not.

He said he had taken the time however to read my prior articles on finasteride. He then said that he had

examined my work on the frequency of prostate cancer due to finasteride and thought it was not

professional. I have never published any article on finasteride causing prostate cancer so this was some

confused invention to bolster his rejection of my article, Ännals of Propecia". He said that the article was

rejected by Dr. Arndt but that he concurred in the decision. The reason quoted was ënough is enough",

(setting a new style in editing). He also said one reason for rejection was that the journal had ä limited

number of pages". Also, he said "The problem is between you and the FDA and Merck and not with the

AOD."

Medical Journals

It is astonishing to scientists that medical journals take paid advertisements from drug companies. It puts

the editors into compromising situations that must be very unpleasant. I have seen this effect before. The

reader might want to read my article which appeared in Urology 4 relating to my dealings with the New

England Journal of Medicine. The editor rejected my finasteride article saying that it would have been

published if it had been submitted when the original Merck finasteride data were published. Apparently if

one finds that a published research claim is flawed one cannot point this out at a later time! (My

conclusions on Proscar were later verified in a double-blind comparison of Proscar, Hytrin and Cardura

carried out by Dr. Lepor at New York University.)

Drug Companies

There was a time when drugstores had drugists that ``filled prescriptions'' supplied by physicians. Now

the huge conglomerates, like Rite-Aid, send their customers materials obtained from a drug company

suggesting that one buy that company's drug. Of course there may be a better drug on the market so the

buyer is being given prejudicial information. I received the pitch for Proscar but there are other drugs used

for treating BPH.

Conclusions

I have examined the responses to my article on Propecia by persons representing Merck, as well as some

representing the FDA. I believe my responses will reveal much to the neutral reader. Physicians recognize

the role of drug companies driven by profit. They are aware of the underfunding and mediocre

performance of the FDA in the approval process. Very few articles by physicians appear on this subject.

I will soon submit a citizen's request for the FDA to reconsider its approval of the Propecia dosage and

plan to ask the many dermatologists from whom I have received email letters to join in with this request.

But what is really needed is for physicians, not physicists, to make the effort to improve medical practices.

Organizations like the Institute of Medicine of the National Academy of Sciences or like Physicians for

Social Responsibility need to study how to make the FDA a model federal agency.

Appendix I.

We reproduce the pertinent FDA data below:

8.1.2.4.2 Efficacy Parameters: Change from Baseline in Hair Count

Table 8.1.2.4.2A Least Squares Summary Statistics, Confidence Intervals, and Between-Group

Comparisons for Endpoints at Month 6 Placebo-Controlled Initial Study

Table 8.1.2.4.2A
Milligram Dosage	1 mg	0.2 mg	0.01 mg	Placebo
Number of Participants	N=101	N=99	N=97	N=96
Mean Change from Baseline+	68.7**	54.9**	-8.6	-15.4
95% Confidence Level	(51.4, 86.1)	(37.4, 72.5)	(-26.5, 9.3)	(-33.3, 2.4)

Note that in each case the change from baseline is closely +/- 17.5 . Since the sample size is large enough

(approximately 100 in each group) it is reasonable to compute s from the confidence level based on the

normal distribution as opposed to considering the t distribution. The standard deviation s is then 8.85 for

the 1.0 mg sample and 8.95 for the 0.2 mg sample. The difference in the means is then Ö2 ×8.9 = 12.89 .

If the distributions are normal with means m1-m2, we have the 95% confidence interval - difference in the

means - of (-10.87, 38.47). Since these values cover 0, there is no significant difference between the

sample means. Q.E.D.

written 1-4-2000 propcrap3 printed May 2, 2000

Footnotes:

1 Neurourology and Urodynamics, 14:619-24 (1995)

2 Apparently the medical profession almost never assigns systematic errors in their presentations and

conclusions, yet in many experiments the systematic errors are larger than the statistical errors.

3 ``The placebo Effect'', Harvard University Press, second printing, 1999

4 Urology 50 (3) 319-20 (1997)