NIH support for PhD training (TRR-IV)

Sweeping problems under the rug may not work

Sweeping all those problems under the rug can leave tell-tale bumps and edges

Today we start bearing down on truly nitty-gritty issues raised by the Tilghman-Rockey report (TRR; see 1). Our trek through the TRR has dealt with only two of its 19 recommendations so far. We deal today with three more—recommendations 3, 4, and 5—all related to PhD training, and numbered in order of appearance in BiomedWatch (see Table and 2). A  quick look at the Table tells us that the TRR’s framers were concerned that: too much graduate training is funded by NIH research project grants (RPGs) rather than by training grants; PhD training takes too long; we know too little about the subsequent career paths taken by our PhDs. Each of these concerns is legitimate and important. Unfortunately, the same cannot be said of TRR recommendations 3-5.

By way of motivating recommendation 3, the report notes that the vast majority of PhD students who receive NIH support are funded by RPGs, rather than training grants, “and yet the NIH has no influence of the quality of the training of these individuals. Training grants uniquely provide the NIH with a mechanism for peer review of training, and permit the NIH to require attention to issues such as outcomes, diversity, and professional ethics training.” Striking data (3), identifying RPGs as the principal NIH funding mechanism for training PhD students, raise many problems; since 1980 the number of basic biomedical PhD degrees awarded each year increased almost 5-fold, along with a 3.4-fold increase in PhD students supported by RPGs, and minimal increases in the numbers of graduate fellowships and traineeships. Moreover, the yearly number of biomedical PhD awards is tightly coupled to variations in the size of NIH’s budget, with a built-in delay that reflects the average 6.5-year duration of graduate training. Thus students who entered PhD training in 2001, at the middle of the budget-doubling “boom,” now represent an abundant “boom” in number—but as they complete their postdocs they enter a “bust” job market (3). Finally, since 1980 the proportion of the NIH budget devoted to training decreased from 15% to 3% (3).

Could the change in funding since 1980 have exerted significant influences on the quality and direction of graduate training? In my opinion, the answer must be yes. Instead of focusing on such influences, however, the TRR simply says that RPG-based funding prevents the NIH from tracking the progress of graduate students and from regulating the quality of their training. (Elsewhere the report does cite evidence that PhD students supported by NIH training grants or fellowships ultimately do better in their careers than students who are supported by other mechanisms; see 4).

Ensuring high-quality graduate training furnishes the explicit rationale for recommendation 3, which is short and simple: the NIH should “increase the proportion of graduate students supported by training grants and fellowships [relative to RPGs], without increasing the overall number of graduate student positions.” Like many others in the TRR, this recommendation is unlikely to affect graduate training in any useful way, because it sweeps key numbers and issues under a rug. Such issues include:

  • How many trainees do we have now, and where does their support come from?
  • How many PhD graduates does the US need?
  • How much should we change the proportion supported by NIH training grants?
  • Where will the NIH find the money to support more graduate trainees?
  • Increasing the proportion of students on training grants without increasing total student number means funding fewer students with RPGs. How should such a change be implemented? Over what period of time?
  • Graduate programs typically use NIH training grants to fund students in years 1-2, so students working in the lab get RPG support (5). Should this practice continue, or should students receive training grant support for the entire training period?

These nitty-gritty questions will eventually creep out from under that rug, so why not shed light on them now? They carry serious implications for how to implement the transition to more training grant support, and also for the effects we expect the transition to bring. Rather than shy away from knotty complexities, the TRR should have explicitly identified them and discussed how to cope with them. With a few facts, we can try to do so ourselves. So, here’s what I could find: at a cost of about $750M per year, NIH funds training slots for about 8,000 biomedical PhD students, representing about 10% of all 83,000 biomedical PhD students in US graduate programs (6). I could not find a reliable number for graduate students supported by NIH RPGs, but a reasonable estimate would be between 20,000 and 25,000 (7). Thus the present training grant-RPG (TG-RPG) ratio is probably a bit less than 1:3. (Note that this ratio refers only to biomedical graduate students funded by training grants and RPGs awarded by the NIH. Non-federal sources support PhD training for about 42% of such students; see 7.)

Taking those numbers, I would amend recommendation 3 in several respects. First, if a TG-RPG ratio of 1:3 is too low, what ratio would suffice to improve the quality of PhD training? Doubling the number supported on training grants (and concomitantly reducing the number on RPGs) would increase the ratio to about 1:1, but would that increase make an appreciable difference? To my mind, the best answer is also conceptually the simplest: no NIH-funded biomedical PhD training should be funded by NIH RPGs, and all NIH-funding for trainees should come from institutional training grants or individual training fellowships, for an TG-RPG ratio of 1:0. (Because more than 40% of biomedical graduate students are funded by non-federal sources (7), even at the 1:0 ratio slightly less than three of every five PhD students would be funded by NIH training grants.) The total cost to the NIH would remain about the same, with a four-fold increase in its training budget and a 10-15% decrease (about $3 billion) in its “research” budget. This apparent decrease would not hinder research, because the students would contribute to NIH research pretty much as they do now.

This no-RPG-graduate-funding plan is indeed the simplest conceptually, but implementing the transition will not be at all simple—an issue to which we’ll return. First, though, let us review the advantages to be derived from successful transition to the 1:0 ratio. This move would significantly stabilize the biomedical workforce, because complete separation of NIH-supported RPG dollars from training dollars would:

  • By dramatically increasing the relative number of students on training grants, enhance the value of excellent training, as perceived by both faculty and students.
  • Make it easier for PIs and institutions to insulate policy, practice, and training standards from conflicting motivations that inevitably arise when young scientists are valued primarily for contributing ideas and data to a lab’s publications.
  • Increase the ability of training programs to track and monitor training quality, as well as ultimate career outcomes of all trainees.
  • Later, once we have accumulated better outcome information, enable the NIH to regulate the training “pipeline” in relation to national needs.
  • Damp instabilities produced when transient funding booms increase RPG dollars and hence graduate student numbers, subjecting those individuals later to bust cycles and tight job markets (as is presently the case; see 3).
  • Eliminate instabilities that are bound to follow a partial increase in training dollars (and decrease in RPG dollars), as PIs and institutions dispute the relative values of each kind of support and the students who receive them.

So, given the advantages of transitioning away from RPG support for PhD students, and the fact that NIH would not have to pay our additional dollars, why might this transition prove so difficult? After setting a starting date, the transition could be gradual if. after that date, (i) each entering graduate student funded by the training grant continues to receive that support for a maximum of six years, and (ii) every RPG-supported PhD student who graduates or leaves graduate school frees up a “slot,” for which the funds would be transferred from the PI’s RPG into the appropriate graduate program, providing support for a new entering student. The transition would be finished in seven years, because PhDs are awarded, on average, after 6.5 years.

In reality, such a transition will be hard to pull off, for two sets of reasons. The first is that, as always, real devils abound in every detail. But even if those devils are exorcised (see “Devils in Details,” below), NIH will have to make a deal with the US Congress. This second problem is truly hard. Congress jealously guards its powers to control key details of the NIH budget, like the total dollars devoted to “research” and “training.” A representative or senator may find it hard to grasp that we can increase the quality of “training” by consolidating it under one funding mechanism, and find it easier to interpret such a move as a quiet attempt to steal dollars from research and transfer it to the greedy pockets of young people who will loll about, talking and studying. We can easily imagine why the TRR swept this one under the rug, but doing so creates another problem. The NIH, PIs, and universities need to agree on a course of action and then, together, persuade Congress to permit them to make it happen. But now the mysteriously vague recommendation 3 casts a dense fog over the path these stakeholders need to take, making it harder to convince Congress to go along with any decisive course of action. Even now, however, a vigorous NIH Director could probably organize his forces and successfully persuade Congress to help separate training funds completely from RPGs. (The task of persuading Congress will be even harder if the NIH seeks a small increase in the TG-RPG ratio, because such a move is less likely to prove useful, needlessly complicated, and harder to sell than complete separation.)

TRR recommendation 4 (see Table) proposes a cap of five years on NIH support (whether via training grants for RPGs) of PhD training. The motivation for this cap is that biomedical scientists take their first tenure-track position at about age 37, which is later than in other fields (e.g., chemistry; 8). But facts presented in the TRR do not indicate that changes in the duration of PhD training account for this difference, or for the later age at which biomedical scientists are awarded their first grant, which appears to be as much as four years later than in 1980 (9). In part, the relative tardiness is due to students’ entries into graduate school (now 1.5 years later) and to an increase of 1-1.5 years in time between PhD award and first permanent job; the duration of PhD training, however, has increased by less than half a year since 1980, from about 6.2 to 6.5 years (8,9). Moreover, the NIH already imposes a five-year limit on training grant support (but not RPG support) for graduate students. (Note: my plan for replacing RPG support by training grants increased this to six years. This time limit would markedly reduce the need for RPG support, which at present allows graduate training to elude all limits. It would also hint that graduation time is nigh, because subsequent support would have to come from a non-federal source, such as the university or a non-federal grant.) While I may agree that PhD training takes too long, the data demonstrate neither that this is so nor that long PhD training accounts for the increasing age of new tenure-track biomedical scientists. Recommendation 4 is not justified, in my opinion, but the TRR is right to urge the NIH to provide funds for pilot programs to determine whether the PhD years can be shortened without reducing the quality of newly minted PhDs. (As indicated in the Table, this recommendation is a codicil to recommendation 2; see 1)

TRR recommendation 5 proposes that peer review of training grant applications include considering “outcomes of all students in the relevant PhD programs at those institutions, not only those supported by the training grant” (1). Here, for a change, I completely agree that this would be a great improvement. The TRR points out that institutions do not keep track of RPG-supported students as well as they do with those on training grants. (The same, they fail to note, is true of the 42% of US biomedical PhD students supported on funds from non-federal sources, which the university itself as well as foundations, etc.) Society, however, very much needs to know the outcomes of training forall students (or at least for a larger proportion thereof). My only cavil: cash-strapped universities are loath to expand their responsibilities in this regard, because collecting outcome information costs time and money. If—as the TRR says, and I wholeheartedly agree—outcome information is vital and should be provided to applicants for PhD training as well as peers reviewers of graduate programs, then it should have asked the NIH to help universities foot the bill for expanding coverage of outcomes. It will help, of course, if NIH’s support for students by RPG support is replaced by support from training grants, but that will also cost host institutions more money. In addition, it is imperative to include the 40+% of students who receive no federal support, and therefore essential for the NIH to help with the extra cost.

At this stage I’m feeling bad about giving the TRR such a hard time, as indicated in the Table. As this series of posts on the TRR continues, I hope readers will compare the report and my criticisms, come to their own conclusions, and (hope, hope, hope!) share with us their written comments—pro, con, and otherwise.

Beginning with this post’s fifth paragraph, I listed a number of difficult and critical issues and claimed that the TRR had tried to sweep them under the rug. The present post has not dealt with the most difficult of these: how many new PhD graduates per year does our biomedical research enterprise need? (We may also ask, how many can it absorb?) It is true that the TRR punted on this one, so the next post will tackle it, head-on.

Devils in Details

My proposal to gradually replace all NIH RPG support for graduate students with training grant support will require measures to deal with myriad nasty but unavoidable details. Because even patient readers balk at swallowing these little monsters, I shall confine my discussion to the two varieties—administrative challenges and attempts to “game” the new system.

Both host institutions and the NIH will face a daunting challenge in expanding training grant administration to accommodate a gradually increasing number of student slots on existing (and new) training grants. Small schools or programs (i.e., without training grants or with fewer than 20 PhD students, total) might be exempted from the transition altogether. For each larger school and/or program—hereafter termed “trainers”—it will be necessary to devote at least one year prior to the “start year” to arranging details of the transition, in collaboration with the NIH. During this period trainers will carefully enumerate the likely number of students to be admitted in the start year and the number of RPG-supported students likely to free up RPG support (during the start year and subsequent years, owing to graduation, quitting school, or completing the sixth year of graduate school). To help in estimating later years, trainers will document the past five years of NIH support for PhD students (on either RPGs or training grants), to set agreed-upon benchmarks for future years. (Increasing the numbers will be considered when training grants are renewed.)

The second variety of nasty details arises from the often unmentioned but inevitable tendency of some individuals and schools to “game” any system. For instance, in the run-up period to the transition, schools and PIs will be tempted to shift more students onto RPGs and use the freed-up training grant slots to recruit more students. Or transfer students from non-federal support to RPG support for that year, in a parallel attempt to swell the number of slots that can accumulate during the transition. Or puff up the numbers of NIH RPG-supported numbers by subterfuge or fraud. Or change locally established policies for absorbing/waiving/paying tuition and fees. I can’t imagine all the possibilities, but clever operators certainly will. Thus alert NIH training grant administrators will forbid or prevent as many of these practices as they can think of, and monitor a random sample of documented RPG-supported students at each institution and program.


1. Biomedical Research Workforce Working Group Report. Pdf here.

2. Recommendation numbers will be consistent in successive posts, but do not necessarily conform to the order of their appearance in the TRR. Moreover, I chose not to discuss one TRR recommendation regarding graduate students, which asks NIH to harmonize and simplify the very different graduate training grant requirements and policies of its multiple institutes. Future posts may mention but not focus on other TRR recommendations. Most of these are unexceptionable and/or relatively unimportant.

3. Data from pp. 13-18 of the TRR. As one example of this tight coupling between NIH research budget and PhD awards, examine Figure 1 (TRR, p 18), which shows that biomedical PhD awards were remarkably stable (at about 5,500 per year) from 1997 to 2004, but then (six years after the NIH budget began doubling) took a sharp upswing, which by 2008 had reached about 7,700. Allowing 6 years for PhD training and about six more years of postdoc before looking for research positions, biomedical PhD students who matriculated in 1999, 2000, or 2001 have to seek permanent employment in the much less rosy economic climates of 2011, 2012, and 2013.

4. While NIH training grants probably do improve training, the inference is confounded by the strong possibility (in my view, a certainty) that graduate programs request support for their very best students from training grants. Training grants that support very strong students are simply more likely to be renewed.

5. Students funded on training grants from the National Institute of  General Medical Sciences, for instance, tend to stay on the grant for the first two years, while training grants in some other institutes are used to fund students for longer.

6. Data from NIH RePORT, NRSA Training Grants/Fellowships, Pre-doc and post-doc FTEs. Pdf here.

7. The number of RAs supported by NIH RPGs is probably in this range, as estimated by Wallace Schaffer, a knowledgeable official in the Office of the Director of NIH, (telephone interview, July 30, 2012). The estimate is based on readjusting data in HH Garrison, K Ngo, Education and employment of biological and medical scientists 2011 (Microsoft Powerpoint presentation, here). From 2009 data compiled by the National Science Foundation, Garrison estimated that of 72,000 biomedical graduate students, support from research assistantships (RAs), “other” (including self), teaching assistantships, traineeships, and fellowships accounted for (respectively) 30,000, 20,000, 10,000, 5,000 and 6,000 students. Schaffer says that other data indicates that not all the RAs, traineeships, and fellowships are NIH-funded. For this reason the actual number of students presently supported on NIH RPGs is by no means definitive.


About biomedwatch
Professor Emeritus of Cellular and Molecular Pharmacology, University of California, San Francisco

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s

%d bloggers like this: